Exposure and response prevention versus stress management training for adults and adolescents with obsessive compulsive disorder: A randomized clinical trial

Abstract

Objective:

Though exposure and response prevention (ERP) is a well-proven treatment for OCD across the lifespan, prior RCTs have not studied adolescent and adult patients with the same ERP protocol relative to an active comparator that controls for non-specific effects of treatment. This approach assesses differences in the effect of OCD-specific exposures in affected adolescents and adults and in response to ERP compared to a stress-management control therapy (SMT).

Methods:

This assessor-blinded, parallel, 2-arm, randomized, ambulatory clinical superiority trial randomized adolescents (aged 12–18) and adults (24–46) with OCD (N = 126) to 12 weekly sessions of ERP or SMT. OCD severity was measured before, during and after treatment using the child or adult version of the Yale-Brown Obsessive Compulsive Scale (C/Y-BOCS), depending on participant age. We predicted that ERP would produce greater improvement in OCD symptoms than SMT and that there would be no significant post-treatment differences across age groups.

Results:

ERP (n = 63) produced significantly greater improvements on C/Y-BOCS scores at post-treatment than SMT (n = 63) (Effect size = − 0.72, CI = − 0.52 to − 0.91, p < .001). ERP also produced more treatment responders (ERP = 86%, SMT = 32%; ${\chi }^2=46.37$, p < .001) and remitters than SMT (ERP = 39%, SMT = 7%; ${\chi }^2=16.14$, p < .001). Finally, there were no statistically significant post-treatment differences in C/Y-BOCS scores between adolescents and adults assigned to ERP.

Conclusion:

A single ERP protocol is superior to SMT in treating both adolescents and adults with OCD. OCD-specific therapy is necessary across the lifespan for optimal outcomes in this highly disabling disorder, though non-specific treatments like SMT are still all-too-commonly provided.

1. Introduction

Obsessive-compulsive disorder (OCD), a condition involving unwanted, intrusive thoughts (obsessions) and associated compensatory behaviors (compulsions) is a common and highly impairing condition that often begins in childhood (American Psychiatric Association, 2013; Torres et al., 2017). Typical OCD symptoms include contamination obsessions with accompanying washing and cleaning compulsions, concerns about making a mistake presenting with co-occurring checking rituals, obsessions related to symmetry with associated straightening rituals, intrusive thoughts about harming others, etc. (Abramowitz et al., 2010). Cognitive-behavioral therapy, featuring exposure and response prevention (ERP), is the most extensively investigated psychosocial treatment for OCD in affected youth and adults (Reid et al., 2021), but it remains unknown whether response to ERP differs in younger compared to older patients.

ERP for OCD in affected youth and adults includes both therapist-assisted and patient-directed exposure exercises aimed at triggering distress and associated intrusive thoughts followed by attempts to block neutralizing behaviors or mental acts (response prevention) (Rowa, Antony, & Swinson, 2007). For example, persons with contamination obsessions are asked to handle dirty objects or surfaces without washing afterward. According to inhibitory learning theory, ERP provides the opportunity for patients to learn that disastrous consequences do not follow from prolonged encounters with challenging OCD-related stimuli (Craske et al., 2008; Jacoby & Abramowitz, 2016).

ERP for adults with OCD has been evaluated in randomized trials against a range of control conditions and comparison treatments including anti-obsessional medication and combined treatment approaches (e.g., medication + ERP) (Reid et al., 2021). In nearly all circumstances, ERP has been shown to be superior to control conditions (e. g., waitlist, relaxation, placebo pills) and at least equivalent to anti-obsessional medications and alternative psychosocial interventions designed to treat OCD (e.g., cognitive therapy) (Abramowitz, 1997; Olatunji, Davis, Powers, & Smits, 2013; Reid et al., 2021).

ERP for youth with OCD has also been evaluated in several controlled and treatment comparison studies (Öst, Riise, Wergeland, Hansen, & Kvale, 2016; Reid et al., 2021; Team, 2004). As with ERP for adults with OCD, most studies find ERP to be superior to control conditions (e.g., waitlist) and comparison treatments in affected youth (Ost et al., 2016; Reid et al., 2021). In addition to typical ERP approaches, ERP for youth with OCD often includes a special emphasis on family involvement with contingent rewards for treatment adherence and more extensive use of clinician and parent-guided ERP as compared to self-conducted ERP. ERP for youth with OCD also typically involves emphasis on coping self-statements (e.g., “it’s not me, it’s my OCD”) and externalizing OCD as an enemy to battle (Piacentini, Langley, & Roblek, 2007). These additional youth-focused approaches require specialized clinician training beyond what is typically offered for adult-focused ERP for OCD (Storch, McGuire, & McKay, 2018).

Finally, some studies of ERP for youth and adults with OCD indicate that participants with certain baseline characteristics (e.g., severe depression, personality disorders, family dysfunction) do not respond as well to ERP (Keeley, Storch, Merlo, & Geffken, 2008). Additionally, when common co-occurring conditions (e.g., generalized anxiety, depression) are present along with OCD at baseline, most studies find that these conditions improve alongside OCD symptoms when patients participate in ERP (Anholt et al., 2011; Rozenman et al., 2019).

Despite clear evidence that ERP is effective for both adults and youth with OCD, uncertainty remains as to whether there are differences in ERP efficacy at different ages. Two meta-analytic studies have compared differences in effect sizes between youth and adults given ERP. The first (Olatunji et al., 2013) found an advantage in effect size found for youth given ERP compared to adults. Conversely, other studies (Reid et al., 2021; Williams & Harris-Reid, 1999), failed to find a difference in effect size between youth and adults treated with ERP. It is important to note that the treatment provided in the outcome studies included in these meta-analyses involve versions of ERP that differ in content for adults versus youth (e.g., externalizing OCD as an enemy to battle and greater family involvement in ERP protocols for youth), thus hindering the ability to make firm conclusions about the relative effectiveness of ERP between youth and adults. Moreover, adults and youth were not typically studied in the same trial, which further limits the ability to understand the relative efficacy of ERP for youth versus adults with OCD. In our review of the literature, we were unable to locate an RCT testing the effect of ERP that includes both youth and adults exposed to the same ERP treatment protocol, within the same trial.

Comparing the relative efficacy of ERP for youth and adults is significant given the possibility of differential outcomes based on developmental stage. Indeed, some research suggests that development may impact the sensitivity of fear circuitry to make exposure-based treatments less effective in adolescents than adults (Casey & Lee, 2015; Meyer et al., 2023), whereas other work suggests that ERP is more effective in youth compared to adults for a range of potential reasons (e. g., more supportive family systems, more mailable cognitions and behavior among youth with OCD) (Olatunji et al., 2013). Thus, whether ERP effect varies with patient age remains an open question of relevance to clinical practice. Differential ERP response by age group would have important implications for treatment plan formulation and discussions with patients and families regarding prognosis. Moreover, age-related differences in response to ERP would have relevance for the tailoring of ERP techniques (e.g., more sessions, more therapist-assisted ERP) for patients within a certain age range.

To isolate the specific effects of ERP, some studies of adult OCD have randomized patients to ERP or stress management training (SMT) (Lindsay, Crino, & Andrews, 1997; Simpson et al., 2008). As a psychosocial comparator condition, SMT involves multiple psychotherapeutic strategies designed to reduce general tension and worry, while purposefully excluding OCD-specific exposures. SMT typically includes muscle relaxation, slow-paced breathing, non-OCD-related (e.g., financial challenges, school problems, friendship problems) problem solving training, and various healthy living strategies like sleep hygiene, exercise, healthy eating, and activity management (Lindsay et al., 1997). SMT has been shown to be ineffective in treating OCD (Ponniah, Magiati, & Hollon, 2013). Despite the ineffectiveness of SMT for OCD, many adults with OCD view SMT as a credible treatment approach for their OCD symptoms (Lindsay et al., 1997).

In adults with OCD, randomized clinical trials showing the superiority of ERP compared to psychosocial comparator conditions, such as SMT (Lindsay et al., 1997; Simpson et al., 2008; Whittal, Woody, McLean, Rachman, & Robichaud, 2010), providing strong evidence that exposure and response prevention are the “active” ingredients of ERP, above and beyond the less specific effects of psychotherapy (e.g., psychoeducation, regular visits with a clinician, treatment expectancy), in adults with OCD. However, in pediatric OCD, we were unable to locate a published, randomized trial that compares ERP to a multi-component psychosocial control condition, like SMT. Thus, among youth with OCD, it remains less certain whether exposure and response prevention techniques, relative to non-specific effects of therapy (e.g., psychoeducation, meeting regularly with a clinician) are the primary drivers of change among OCD youth treated with ERP.

The present randomized, controlled trial of ERP versus SMT for adults and adolescents with OCD directly addresses three key questions. First, this trial randomly assigns adults and adolescents to identical versions of ERP or SMT within the same study thus allowing for a direct comparison of the effect of ERP on OCD in different age groups. Second, this paper adds to the limited adult literature (e.g., Lindsay et al., 1997; Simpson et al., 2008) that compares the efficacy of ERP versus an active, credible SMT control condition that involves the same number of therapy sessions and equivalent therapist attention as ERP. Finally, this study provides the first RCT comparing the effects of ERP versus SMT for adolescents with OCD.

The primary hypotheses in the present study are that both adults and adolescents with OCD will experience greater reduction in OCD symptoms after ERP compared to SMT, and that ERP-derived improvements in OCD symptoms would be as robust for affected adolescents as adults.

2. Method

2.1. Trial design

This article reports the results of a 12-week randomized control trial (RCT) comparing the effects of ERP for adolescents and adults with OCD against an active control condition, SMT. This RCT was conducted as part of a parent study that used functional MRI (fMRI) to examine the impact of ERP on neurocircuit mechanisms and cognitive performance for adolescents and adults who experienced early onset of OCD symptoms (<16 years) (Norman et al., 2021). Accounts of fMRI results from this trial have been published elsewhere (De Nadai et al., 2023; Norman et al., 2021; Russman Block et al., 2023) but given that these analyses were designed to test for developmental differences in neural mechanisms of CBT relative to SMT, they did not provide specific symptom data (i.e., OCD, generalized anxiety, depression) comparing ERP versus SMT outcomes for adolescent versus adult participants. The trial was approved by the University of Michigan (IRB-MED) Institutional Review Board. NIH certificates of confidentiality were obtained. All participants provided written informed consent/assent. The trial was pre-registered in clinicaltrials.gov, #NCT02437773.

2.2. Participants

Participants included both adolescents (12–18 years of age) and adults (24–46 years of age) with OCD. These age ranges were chosen to isolate more “plastic” (adolescent) compared to more stable (adult) periods of brain development and because use of the same ERP protocol may be more appropriate with adolescents and adults compared to younger children where more developmental tailoring is likely to be needed (Storch et al., 2018). Given that it was not our intent to examine emerging adulthood, we deliberately excluded potential participants between the ages of 18–24 years. OCD diagnoses were determined using the Structured Clinical Interview for DSM-IV-TR Axis I Disorders - Patient Edition (SCID) for adults (First, Spitzer, Gibbon, & Williams, 1996) or the Structured Clinical Interview for DSM-IV Childhood Diagnoses (KID-SCID) for adolescents (Hein et al., 1998). Eligibility requirements also included moderate or greater OCD symptoms (total score of ≥16 on the adult or child version of the Yale-Brown Obsessive-Compulsive Scale C/Y-BOCS; see below) and OCD symptom onset <16 years given research suggesting that pediatric onset OCD may be distinct from adult-onset illness (Geller, Homayoun, & Johnson, 2021). We required age of onset of OCD symptoms before the age of 16 years. Age at onset was determined using the Schedule for Obsessive-Compulsive and Other Behavioral Syndromes (Hanna, 2007).

Patients with a lifetime diagnosis of bipolar disorder, psychosis, intellectual disability, or hoarding disorder were excluded, as were those with other Axis I disorders less than five years in remission, except social phobia, generalized anxiety, specific phobia, tic disorders, separation anxiety disorder, or OC spectrum disorders (e.g., body dysmorphic disorder, trichotillomania), which are the most common comorbidities in adolescent and adult patients, occurring at the same rates (~30%) in both age groups (Bienvenu et al., 2012; Geller et al., 2001; Hofmeijer-Sevink et al., 2013). Finally, potential participants were excluded if they met criteria for substance dependence in the past two years, substance abuse within the past six months, suicidal intention or behavior in the past six months, or current major depressive disorder or past major depressive disorder in remission for less than two months.

Eligible participants could be taking a stable dose of a selective serotonin (SSRI) or serotonin norepinephrine reuptake inhibitor (SNRI) medication for at least four weeks prior to baseline MRI and initiation of study therapy, and could not be taking antipsychotic medications, anticonvulsants, lithium, or stimulants. Benzodiazepines were allowed for patients who had been previously taking them.

Finally, participants could not have a previously failed course of ERP for OCD. In order to avoid enrolling ERP resistant participants, we interviewed potential participants about their previous experience with ERP for OCD. All potential participants were questioned about whether they had previous ERP, the number of previous ERP sessions, the nature of the ERP exercises, and their response to the previous ERP treatment. If a potential participant experienced a guideline concordant course of ERP for their OCD in the past and did not improve, they were not enrolled. Due to the aims of the parent neuroimaging study, participants were also excluded if they had a history of closed head injury, serious medical or neurological illness, or fMRI contraindications (i.e., vision less than 20/30 after correction, metallic implants, an inability to tolerate the scanner space). All participants (and their a legal guardian, if a minor) were required to demonstrate the ability and willingness to give informed consent and/or assent.

One-hundred seventy-one participants (74 adolescents) were consented and assessed for eligibility into the parent neuroimaging study (see Fig. 1. Consort diagram). The 58 adolescents (40 female, 69%) and 68 adults (38 female, 56%) examined in the present study are those randomized into treatment. The adolescent group was aged 12.42–17.92 years (M = 15.60; SD = 1.61) and the adult group was aged 24.17–45.42 years (M = 31.58; SD = 5.76). They were recruited through outpatient psychiatry programs at the University of Michigan, referrals from community clinicians, and self-referral through paper flyers and web-based advertising. After an initial telephone screening, participants met with a master’s level clinician to determine study eligibility. Adolescents in the study were assessed either individually or with their parents present, depending on participant preference. If adolescents elected not to have their parents present in the room during their assessment, parents were consulted separately, and their input contributed to diagnostic impressions and scores on symptom-related measures. Adults were offered the opportunity to include a support person (e.g., spouse) during assessments.

Fig. 1.

Consort diagram.

2.3. Randomization

Participants were assigned to either the treatment (ERP) or control group (SMT) using a blocked randomization with a block size of four, stratified by age, sex, and medication status. Randomization was maintained throughout the study with two exceptions: first, in the case of one child-parent dyad who both participated in the study, the randomization block was shifted to ensure that both the adolescent and the parent were engaged in the same study treatment to prevent compromising intervention fidelity through assignment to different conditions; and second, a participant changed their medication status during the initial portion of their study treatment which necessitated a new baseline fMRI scan. After the new baseline scan, for ethical reasons, the participant’s originally assigned study treatment was re-started without re-randomization.

2.4. Treatment conditions

For adults and children randomized to ERP, participants received individualized exposure and response prevention therapy based primarily on our ERP intervention manual used in a range of previously published outcome studies (Håland et al., 2010; Himle et al., 2001, 2006). This manual relies on exposure and response prevention-related elements based on the work of Foa (Foa, Yadin, & Lichner, 2012) and Franklin and colleagues (Franklin, Freeman, & March, 2018). Participants randomized to ERP completed both in-vivo and imaginal exposures based on their fears while refraining from ritualistic behaviors aimed at mitigating distress. The goal of this therapy was to provide the opportunity for participants to learn that disastrous consequences do not follow from prolonged encounters with challenging OCD-related stimuli. ERP was scheduled weekly for a total of 12 sessions completed over a maximum of 16 weeks (allowing for missed sessions based on scheduling conflicts). Each ERP session included a combination of OCD-related psychoeducation and exposure and response prevention exercises. Treatment included: Session 1 - Introduction to OCD and ERP Treatment Overview; Session 2 - Externalizing OCD and Analyzing OCD Symptoms; Session 3 - Principles of ERP and Implement Initial ERP Exercises. Sessions 4–12 included ongoing ERP exercises and the following additional topics: Session 4 - The Causes of OCD and Cognitive Methods and Coping Self-Statements; Session 5 - Family Life with OCD and Attitudes that Help with ERP; Session 6 - OCD and the Family; Session 7 - Family Session Debrief and Cognitive Approaches; Session 8 - Comorbidity with OCD and Strategies for Making ERP more Effective; Session 9 - Social Life with OCD and Extending Beyond Initial Gains from ERP; Session 10 - Living Better with OCD and Becoming your own ERP Therapist/Mock ERP Scenarios; Session 11 - Preparing to End and Becoming your own ERP Therapist/Mock ERP Scenarios; Session 12 - Ending ERP and Maintaining your gains. In-session exposure and response prevention exercises were conducted with the therapist during sessions 3–12 and participants were asked to complete 1 h per day of self-guided exposures at home and to document OCD-related behaviors. A support person, usually a family member, was present during ERP sessions 1, 6 and 12. Given that persons taking benzodiazepines were eligible to participate in the trial, participants randomized to ERP were informed about the importance of not taking benzodiazepines directly before or after exposure sessions. The ERP manual used in this trial is available from the corresponding author.

Participants randomized to SMT were instructed in deep breathing, progressive muscle relaxation, healthy living habits, and problem-solving, but did not engage in exposure exercises. Participants were told that these techniques offer useful strategies for addressing life stressors that may exacerbate OCD. In order to provide fully informed consent to participants, the consent document stated the following: “Stress management training will teach skills to help you relax. Stress management skills reduce stress and may reduce OCD symptoms, although not as much as CBT does.” SMT was selected as a comparison treatment because it had been found to be credible, yet not effective in reducing OCD symptoms (e.g., Lindsay et al., 1997) and could be delivered to match the number of sessions and duration of the ERP treatment used in this study. SMT was scheduled weekly for a total of 12 sessions, which were completed over a maximum of 16 weeks (allowing for missed sessions based on scheduling conflicts). SMT included: Session 1 - Introduction to OCD and SMT Treatment Overview; Session 2 - Physical Components of Anxiety & Stress; Session 3 - Diaphragmatic Breathing; Session 4 - Monitoring Muscle Tension; Session 5 - Progressive Muscle Relaxation; Session 6 - Isometric Relaxation Exercises; Session 7 - Identifying Problems for Problem-Solving; Session 8 - The Process of Problem-Solving; Session 9 - Further Practice in Problem-Solving; Session 10 - The Role of Good Living Habits in Managing Stress & Anxiety; Session 11 - Identifying Most Helpful Treatment Components; Session 12 - Maintaining Gains and Saying Goodbye. A support person, usually a family member, was present during sessions 1, 6 and 12. Like ERP, patients were also asked to complete 1 h per day of self-guided practice of SMT strategies between sessions. Finally, participants who were assigned SMT were offered a no-cost course of ERP following their post-treatment MRI scan.

2.5. Therapists

Both interventions were manualized and treatment was provided by masters-level clinicians receiving weekly supervision from a Ph.D.-level clinician (JAH). Clinicians were specialists in ERP and were given two-day trainings on each of the two intervention manuals. The same clinical team provided both the ERP and SMT interventions in the trial. Supervision included strong attention to preventing drift of SMT treatment components into the ERP condition and ERP elements into the SMT condition.

2.6. Measures

Table 1 provides a list of measures, measure type, and administration schedule.

Table 1.

List of measures and administration schedule.

Measure	Screening	Treatment
		Beginning	Midpoint	End
Assessment of Symptoms and Functioning
Structured Clinical Interview for DSM-IV-TR Axis I Disorders	Clinical Assessment
Schedule for Obsessive-Compulsive and Other Behavioral Syndromes	Clinical Assessment
Yale-Brown Obsessive Compulsive Scale	Clinical Assessment	Blinded Evaluation	Blinded Evaluation	Blinded Evaluation
Clinical Global Impressions Scale - Severity		Blinded Evaluation	Blinded Evaluation	Blinded Evaluation
Clinical Global Impressions Scale - Improvement			Blinded Evaluation	Blinded Evaluation
Generalized Anxiety Disorder-7		Self-Report	Self-Report	Self-Report
Quick Inventory of Depressive Symptomatology		Self-Report	Self-Report	Self-Report
Intervention Fidelity, Acceptability, and Process
Independent Evaluation of Recorded Sessions		Single audio tape rated
Treatment Adherence Rating Scale		Averaged across available treatment sessions
Credibility and Expectancy Questionnaire		Self-Report
Working Alliance Inventory			Self-Report	Self-Report

2.6.1. Primary outcome measure

The primary outcome measure was the total score on Yale-Brown Obsessive Compulsive Scale - Present/Lifetime Version (Y-BOCS; Goodman et al., 1989a) for adults and Child Yale-Brown Obsessive Compulsive Scale - Present/Lifetime Version (CY-BOCS; Scahill et al., 1997) for adolescent participants. These scales are considered the “gold standard” OCD assessment tools (Bloch, Lennington, Szuhay, & Lombroso, 2015; Deacon & Abramowitz, 2005). The C/Y-BOCS total score ranges from 0 to 40 and sums ten items, each with a scale of 0 (None) to 4 (Extreme/No symptom control) (Bloch et al., 2015). Half-points were permitted to allow greater specificity (e.g., 3.5 denoting symptoms between Severe and Extreme). In our experience with the C/YBOCS, we find that adding 4 extra points to the standard 5-point item response (i. e., half-points) adds precision and improves reliability. The literature, particularly Nunnally’s psychometric theory (e.g., Nunnally & Bernstein, 1994), suggests that reliability increases from 3 to 5 to 7 points for a response, with a plateau around 11. While this assertion has been debated, it is generally acknowledged that on scales with few items, having more points for each item is important. The C/Y-BOCS yields both present and lifetime worst severity scores. Study participants met OCD severity criteria if they had a present score of at least 16, indicating moderate or greater symptom severity at time of study enrollment. Total score on the C/Y-BOCS was evaluated as a continuous variable, changing from baseline to the completion of therapy.

The C/Y-BOCS was administered by a master’s level clinician, blinded to treatment condition. To help maintain the blind, these independent evaluators were excluded from clinical discussions related to individual participants’ treatment and were not informed of the randomization block size. Participants were also reminded not to discuss the specifics of their therapy with the independent evaluator. In this study, internal consistency of the C/Y-BOCS was high (baseline Cronbach’s α = 0.83).

2.7. Secondary outcomes

2.7.1. Categorical OCD symptom outcomes

The Global Improvement item of the CGI Clinical Global Impressions Scale (Guy, 1976) was used alongside the C/YBOCS to determine categorical outcomes (e.g., treatment remitter, responder, partial responder, non-responder – see below). The Clinical Global Impressions Scale - Severity (CGI-S) scale is a National Institute of Mental Health instrument that is used to establish initial levels of symptom severity and point-in-time measures of treatment progress. Scores could range from 1 (Normal, not ill at all) to 7 (Among the most extremely ill). The measure for brief assessment within clinical contexts and multiple studies have established correlations between the CGI-S and commonly accepted measures of symptom severity (Busner & Targum, 2007). The Global Improvement item of the CGI (CGI-I) was collected at the midpoint and end of treatment. Improvement was anchored to the beginning of treatment to ensure that any improvement captured was due to the treatment itself and not unrelated change between the screening evaluation and the start of treatment.

Participants in both groups were classified as treatment completers if they attended at least eight sessions of therapy. Treatment response and remission classifications were evaluated using the categorical benchmarks recommended by Mataix-Cols and colleagues (Mataix-Cols et al., 2016), although this present protocol did not verify that all scores lasted for at least one week. Following Mataix-Cols et al., treatment responders achieved a ≥35% reduction in C/Y-BOCS severity from start to end of treatment and a CGI improvement (CGI-I) score of ≤2 (“much improved”; lower scores indicate increased improvement). Participants who achieved a ≥35% C/Y-BOCS reduction but a CGI-I score >2 were and those who achieved a ≥25% C/Y-BOCS reduction and a CGI-I ≤ 3 (“minimally improved”) were classified as partial responders. All others were classified as non-responders. To be classified as a remitter, participants required a C/Y-BOCS severity ≤12 at the end of treatment and a CGI severity (CGI-S) score ≤2 (“borderline mentally ill”; lower scores indicate lower severity).

2.8. Exploratory outcomes

2.8.1. Generalized anxiety

To assess the impact of treatment condition on generalized anxiety, and to control for impact of baseline generalized anxiety symptoms on OCD outcomes, participants completed the Generalized Anxiety Disorder-7 (GAD-7) scale (Spitzer, Kroenke, Williams, & Löwe, 2006). This scale is a 7-item self-report questionnaire assessing generalized anxiety symptoms, which has been established to have good reliability and construct, factorial, and criterion validity (Spitzer et al., 2006). Participants provide a rating of anxiety symptoms on a 4-point scale ranging from 0 (Not at all) to 3 (Nearly every day). Items are summed to provide a total score ranging from 0 (no anxiety) to 21 (severe anxiety) (Spitzer et al., 2006). In this study, internal consistency of the GAD-7 was high (baseline Cronbach’s α = 0.88).

2.8.2. Depression

To assess the impact of treatment condition on depressive symptoms and to control for the impact of baseline depressive symptoms on OCD outcomes, participants completed the Quick Inventory of Depressive Symptomatology (QIDS-SR) (Rush et al., 2003). This scale is a 16-item self-report questionnaire examining depressive symptoms. Participants rate symptoms over the past seven days using a 4-point scale, ranging from 0 to 3. Item anchoring text is tailored to the individual symptom with 0 indicating a lack of symptom severity and 3 indicating severe symptom severity, producing a total score ranging from 0 (no depressive symptoms) to 27 (severe depressive symptoms). The validity and reliability of the QIDS-SR has been established with both major depressive disorder and bipolar disorder (Bernstein et al., 2009). In this study, internal consistency of the QIDS-SR was acceptable (baseline Cronbach’s α = 0.79).

2.9. Intervention fidelity, acceptability, and process

2.9.1. Independent evaluation of recorded sessions

To quantify how closely therapy administration adhered to study protocols, an audio recording from one of three sessions (sessions 1, 3, or 7) from each participant was randomly selected for review by an ERP specialist with at least a master’s degree. Raters were trained against a gold standard of twelve tapes, six for ERP and six for SMT. Gold standard ratings were set by the corresponding author, who also provided clinical supervision throughout the course of therapy. Fidelity ratings were based on a modified version of the CBT Scale (Young & Beck, 1980) created specifically for the present study. Each session component was rated for competence on a scale from 1 to 5.1 indicated that the component was “not at all competently” administered and 5 meant it was “fully competently” administered. Competence scores were averaged within a session, yielding a single score of 1–5 to describe treatment fidelity.

2.9.2. Credibility and expectancy

Perceived credibility and initial expectations for improvement for ERP and SMT were assessed using The Credibility and Expectancy Questionnaire (CEQ) (Borkovec & Nau, 1972). The CEQ is a 6-item self-report questionnaire examining participant thoughts and feelings about treatment, administered at the beginning of therapy. Using the scoring described in Thompson-Hollands (Thompson-Hollands, Bentley, Gallagher, Boswell, & Barlow, 2014), items #1–3 were averaged to create a composite score of treatment credibility, with 1 indicating low credibility and 9 indicating high credibility on a 9-point scale. Item #4 was used to calculate expectancy, with participants responding in 10% increments on an 11-point scale, indicating from 0% to 100%, indicating how much they expected their symptoms to improve.

2.9.3. Working alliance inventory

The nature of the relationship between the study clinician and the participant was measured using The Working Alliance Inventory (WAI; Hatcher & Gillaspy, 2006). The WAI is a 12-item self-report inventory administered at the midpoint and end of treatment. It assesses the therapeutic alliance across three summed subscales, including alignment on therapy tasks, alignment on therapy goals, and the strength of the therapeutic bond. Each item is rated on a scale from 1 = Never to 7 = Always, yielding subscale ranges from 4 to 28, with higher scores indicating a stronger alliance (Hatcher & Gillaspy, 2006).

3. Statistical analysis

3.1. Multivariate analysis

The analyses were conducting using an intent-to-treat approach. Multilevel analysis for longitudinal data, accounting for the repeated measures design of the study (Raudenbush & Bryk, 2002), was used to examine the relationship of treatment assignment to outcomes. In the first step of the analysis, unconditional multilevel models were constructed to estimate variation in C/YBOCS, GAD-7, and QIDS scores that could be explained by the clustering of the observations across the 3 timepoints at which each dependent variable was measured (Rabe-Hesketh & Skrondal, 2008). Models were then estimated containing the relevant independent variables (i.e., treatment assignment, time, age, sex, OCD medication status). There was a minimal amount of missing data. Therefore, we employed multiple imputation (StataCorp, 2019) procedures using the recommended 20 imputations (Graham, Olchowski, & Gilreath, 2007). Notationally, our models were of the following form:

$${\text{y}}_{\text{it}}={\text{β}}_0+{\text{β}}_1\text{treatment}+{\text{β}}_2\text{week}6+{\text{β}}_3\text{week}12+$$

$${\text{β}}_4\text{week}6\ast \text{treatment}+{\text{β}}_5\text{week}12\ast \text{treatment}+$$

$${\text{β}}_6\text{age}+{\text{β}}_7\text{sex}+{\text{β}}_8\text{OCD medication usage}+{\text{u}}_{0\text{i}}+{\text{e}}_{\text{it}}$$

Here ${\text{y}}_{\text{it}}$ represented our various outcomes of interest for subject $i$ at time $t$. ${\text{β}}_1$ was a main effect for treatment condition (ERP, SMT), and ${\text{β}}_2$ and ${\text{β}}_3$ were the coefficients for the two later timepoints of the study. ${\text{β}}_4$ and ${\text{β}}_5$ were coefficients for the interaction of treatment and time and indicated the degree to which participants in the treatment condition improved more rapidly than participants in the control condition. ${\text{β}}_6$ and ${\text{β}}_7$ were regression coefficients for age and sex respectively. ${\text{β}}_8$ was the regression coefficient for OCD medication usage. ${\text{e}}_{\text{it}}$ indicated the “usual” error term for subject $\text{i}$ at time $\text{t}$, while ${\text{u}}_{0\text{i}}$ was a random intercept for individual $i$.

3.2. Bivariate analysis

After completion of the study, we examined the relationship of two key treatment outcomes (responder vs. non-responder and remitter vs. non-remitter) to treatment assignment. We examined the relationship of each outcome with treatment assignment by performing a cross-tabulation and conducting a ${\chi }^2$ square test.

3.3. Power analysis

The required sample size of a research project depends on several quantities, including the proposed level of statistical significance, the desired statistical power to detect an effect if one is present, and the estimated size of the effects one plans to be able to detect. Below we conduct some power calculations. We employ some common assumptions. We assume that α, the desired level of statistical significance will be 0.05. We assume that $\text{β}$, our power to detect an effect if it is present, will be 0.80. In our study, we found that the treatment had a Cohen’s d effect size of 0.72. Under these assumptions, our required sample size is 32 participants in both the treatment and the control group leading to a proposed total sample size of 64 participants, suggesting that we have more than enough statistical power in this study.

4. Results

4.1. Participants

Enrollment occurred between March of 2015 and August of 2019 with final follow-up assessments in January 2020. Randomization was successful in producing equivalent groups. The participants randomly assigned to ERP (n = 63) and SMT (n = 63) did not differ significantly on comparisons of baseline demographic or clinical characteristics or on continuous outcome variables measured at baseline with the exception of higher baseline total C/YBOCS in the SMT group (see Table 2).

Table 2.

Baseline characteristics.

ERP (N = 63)			SMT (N = 63)
	Adolescents	Adults	Total	Adolescents	Adults	Total
	(N = 30)	(N = 33)	(N = 63)	(N = 28)	(N = 35)	(N = 63)
	M (SD)	M (SD)	Raw Score	M (SD)	M (SD)	Raw Score
YBOCS - Baseline	26.28 (5.54)	23.58 (5.17)	24.87	27.80 (5.05)	27.23 (4.58)	27.48	*
YBOCS - Week 6			18.54			24.30	*
YBOCS - Week 12			12.49			22.46	*
QIDS - Baseline	7.54 (5.20)	6.62 (3.77)	7.05	7.37 (3.92)	7.76 (5.29)	7.58
QIDS - Week 6			5.98			6.71
QIDS - Week 12			5.11			6.73
GAD-7 - Baseline	9.79 (5.46)	9.24 (5.40)	9.50	9.57 (5.51)	9.76 (5.60)	9.68
GAD-7 - Week 6			7.30			8.26
GAD-7 Week 12			6.09			7.98
Age (yr)	15.77 (1.53)	31.59 (5.78)		15.43 (1.69)	31.56 (5.83)
	N (%)	N (%)		N (%)	N (%)
Female	21 (70)	18 (55)		19 (68)	20 (57)
Minority Racea	6 (20)	2 (6)		1 (4)	8 (23)
Hispanic/Latino	4 (13)	3 (9)		1 (4)	1 (3)
OCD Medsb	17 (57)	13 (39)		15 (54)	14 (40)
Benzodiazepines	3 (10)	3 (9)		0 (0)	1 (3)
Comorbiditiesc
Anxiety	17 (57)	12 (36)		17 (61)	19 (54)
Body-Focused	6 (20)	4 (12)		5 (18)	7 (20)
Depressived	3 (10)	1 (3)		3 (11)	1 (3)
Eating	0 (0)	0 (0)		1 (4)	0 (0)
Tic	3 (10)	6 (18)		1 (4)	2 (6)

^*p < .05.

For each of the following, multiple selections were allowed.

^aBlack/African American (N = 6); Asian (N = 5); Other Pacific Islander (N = 1); Latino (N = 5) (Some participants identified as both a Minority Race and White).

^bCitalopram (N = 2); Clomipramine (N = 4); Escitalopram (N = 10); Fluoxetine (N = 14); Fluvoxamine (N = 6); Paroxetine (N = 3); Sertraline (N = 20); Venlafaxine (N = 1).

^cAnxiety: Anxiety Disorder NOS (N = 4), Generalized Anxiety Disorder (N = 44), Panic Disorder with Agoraphobia (N = 2), Panic Disorder Without Agoraphobia (N = 5), Separation Anxiety Disorder (N = 1), Social Phobia (N = 29), Specific Phobia (N = 8) Body-Focused: Body Dysmorphic Disorder (N = 1), Presence of grooming disorders, yes or no (N = 21). Formal grooming disorders were not assessed.

^dDepression: Depressive Disorder NOS (N = 7), Dysthymic Disorder (N = 1) Eating: Eating Disorder NOS (N = 1) Tic: Presence of tics, yes or no (N = 12). Formal tic disorders were not assessed.

At baseline, 46.8% of participants were taking psychotropic SSRI and SNRI medications for OCD (i.e., Citalopram, Clomipramine, Escitalopram, Fluoxetine, Fluvoxamine, Paroxetine, Sertraline, or Venlafaxine) (see Table 2).

4.2. Primary outcomes by condition over time: multilevel models

Outcome variables across time and condition are reported in Table 3.

Table 3.

Primary outcomes by condition over time: Multilevel models.

	YBOCS	YBOCS Youth	YBOCS Adults	QIDS	QIDS Youth	QIDS Adults	GAD	GAD Youth	GAD Adults
Treatment
ERP	− 2.642*	− 1.671	− 3.707**	− 0.313	0.195	− 0.883	− 0.023	0.105	− 0.883
Time
Week 6	− 3.306***	− 2.964*	− 3.579***	− 0.812	− 0.773	− 0.833	− 1.447**	− 1.500*	− 0.833
Week 12	− 4.866***	− 5.505***	− 4.354***	− 0.794	− 1.271*	− 0.402	− 1.746**	− 2.653***	− 0.402
Treatment x Time
ERP x Week 6	− 3.120**	− 3.594*	− 2.727	− 0.386	0.183	− 0.913	− 1.013	− 1.077	− 0.913
ERP x Week 12	− 7.709***	− 7.614***	− 7.726***	− 1.203	− 0.548	− 1.752	− 1.808*	− 1.380	− 1.752
Characteristics
Age	− 0.046	0.554	0.104	− 0.020	0.270	− 0.067	0.065	0.826*	− 0.067
Sex	2.348*	3.047	1.925	0.359	− 0.479	1.354	− 0.240	− 0.751	1.354
Medications	0.473	0.879	− 0.188	1.323	2.112*	0.223	1.693	2.293	0.223
Intercept	27.491***	17.806*	23.209***	7.178***	2.139	9.001**	7.363***	− 4.161	9.001**
Standard Deviation of Random Effects
Intercept	4.573***	4.982***	3.861***	3.734***	3.737***	3.550***	4.479***	4.360***	3.550***
Error	4.424***	4.680***	4.153***	2.281***	2.259***	2.257***	2.791***	2.697***	2.257***

^*if p < .05,

^**p < .01,

^***p < .001.

4.2.1. C/YBOCS

The ERP group began treatment with a lower C/YBOCS score than did the SMT comparison group, but these differences are controlled for by the multi-level model (beta = − 2.642, see Table 3). Both groups showed a reduction in C/YBOCS over time but improvements in the ERP group exceeded those observed in the SMT group (beta = − 3.120, beta = − 7.709, see Table 3 and Fig. 2). Interactions of treatment by time showed that there was an even greater.

Fig. 2.

OCD symptoms over time and treatment for combined youth and adult sample.

Reduction in C/YBOCS for the ERP group than for the SMT group. Age group was not significantly associated with C/YBOCS scores over the course of the study (see Table 4). Differential effects of age group on C/YBOCS scores were tested by treatment condition (treatment x time period x age group interaction) but did not differ statistically (See Table 4).

Table 4.

Primary outcomes by treatment, time, and age group: Multilevel models.

Treatment	YBOCS	$p>|\text{t}|$
ERP	− 3.715*	0.015
Time
Week 6	− 3.579**	0.001
Week 12	− 4.354***	0.000
Treatment x Time
ERP x Week 6	− 2.727	0.087
ERP x Week 12	− 7.726***	0.000
Group Age
Youth	0.797	0.620
Treatment x Group Age
ERP x Youth	2.238	0.321
Time x Group Age
Week 6 x Youth	0.614	0.702
Week 12 x Youth	− 1.151	0.487
Treatment x Time x Group Age
ERP x Week 6 x Youth	− 0.867	0.712
ERP x Week 12 x Youth	0.112	0.963
Characteristics
Sex	2.464*	0.012
Medications	0.312	0.746
Intercept	26.048***	0.000
Standard Deviation of Random Effects
Intercept	4.500***
Error	4.405***

^*if p < .05,

^**p < .01,

^***p < .001.

Men in both the CBT and SMT conditions had significantly higher C/YBOCS scores than women over the course of the study (β = 2.35, p < .05).

Finally, there were no main effects of OCD medication status on treatment outcomes (p > .05). Nor were there any statistically significant interactions of OCD medication with treatment assignment (p > .05).

The pre to post treatment within group score effect size for the combined adolescent and adult sample randomized to ERP was large (Cohen’s d = 1.46) with C/YBOCS scores decreasing from mean of 24.87 (SD = 5.48) (severe symptoms) at pre-treatment to 12.49 (SD = 8.00) (mild symptoms) at post-treatment. The between group (ERP vs. SMT) Cohen’s d effect size was − 0.72.

Finally, these results were mostly paralleled in analyses which separated the sample by age group. For adolescents, the same pattern of results held: there were reductions in C/YBOCS for the SMT group; and also more pronounced reductions in C/YBOCS for the ERP group for the six-week and twelve-week time points (see Fig. 2). For adults, results were slightly different: there were reductions in C/YBOCS for the SMT group; but there were more pronounced reductions in the ERP group only at the twelve-week time point (see Fig. 2).

4.3. Secondary 0utcomes

4.3.1. Secondary outcomes by condition over time: bivariate models

Bivariate findings indicated relationships between treatment condition and C/YBOCS outcomes when categorizing study participants as “responders”, “partial responders” or “non-responders” (${\chi }^2=46.370$, p < .001). In the ERP group, 78.95% met the standard as responders and 7.02% met post-treatment criteria as partial responders. In the SMT group, only 15.79% met criteria as responders and 15.79% of the participants were partial responders post-treatment.

Additionally, we calculated whether participants met criteria as “remitters.” Treatment condition was also related to this outcome (${\chi }^2=16.143$, p < .001). In the ERP group, 38.60% of those participating in ERP met post-treatment criteria as a remitter whereas only 7.02% of those participating in SMT met this benchmark.

We additionally calculated these numbers separately by age group. Among adults, 83.33% of the ERP group were responders, whereas 6.67% of the ERP group were partial responders. In the SMT group, 9.68% were responders, and 12.90% were partial responders. These outcomes differed across treatment assignment (${\chi }^2=34.279$, p < .001). Among adolescents in the ERP group, 74.07% were responders, whereas 7.41% were partial responders. In the SMT group, 23.08% of adolescents were responders, whereas 19.23% were partial responders (${\chi }^2=13.810$, p = .001). These categorical outcomes did not significantly differ by age group (${\chi }^2=0.64$, p > .05).

With regard to remitter status, among the adults, 43.33% of ERP participants met the criteria to be classified as a remitter, whereas 3.23% of SMT participants met this criterion (${\chi }^2=13.869$, p < .001). Among adolescents, 33.33% of ERP participants met criteria for remitter status, whereas 11.54% of SMT participants met remitter status criteria. Remission status outcomes did not significantly differ by age group (${\chi }^2=0.61$, p > .05).

4.4. Exploratory outcomes –anxiety and depression

4.4.1. Depression - QIDS

The ICC for QIDS was 0.713, indicating that 71.3% of variation in the QIDS outcome was explainable by between person differences. QIDS results showed a very different pattern compared to OCD-specific outcomes. There were no significant reductions in QIDS for either the SMT group or the ERP group, except for the finding that there were significant reductions for adolescents on the QIDS at 12 weeks. These reductions for adolescents at 12 weeks were statistically equivalent across SMT and ERP. Additionally, age and sex were not related to differences in QIDS (see Supplemental Figure A). For both adolescents and adult participants assigned to either ERP or SMT, mean QIDS scores at baseline, 6 weeks and 12 weeks were in the mild range.

4.4.2. General anxiety - GAD-7

The ICC for GAD-7 was 0.673 showing that 67.3% of the variation in GAD was attributable to between person differences. There was a different pattern for GAD compared to the other outcomes (i.e., C/YBOCS, QIDS). Individuals in both the SMT and ERP groups showed reductions in GAD-7 scores, however, there were no significant differences between the groups. Somewhat similar patterns were observed when the analysis was restricted only to adolescents where individuals in both the SMT and ERP groups showed reductions in GAD-7 scores regardless of whether participants received SMT or ERP. In the adult group, there were no statistically significant results on the GAD-7, indicating that neither SMT nor ERP were associated with changes in GAD-7 scores (see Supplemental Figure B). For both adolescents and adult participants assigned to either ERP or SMT, mean GAD-7 scores at baseline, 6 weeks and 12 weeks were in the mild range.

4.5. Intervention fidelity, acceptability, and process

Treatment fidelity for randomly selected ERP (55 of 677) and SMT (53 of 656) sessions was rated on a 1–5 point Likert scale (≥4 indicating high fidelity) by two independent ERP experts. Mean fidelity ratings were 4.54 (SD = 0.40) for ERP and 4.39 (SD = 0.47) for SMT, which was consistent with the threshold for high fidelity treatment that we established prior to study initiation.

Patient-rated treatment credibility ratings for ERP (M = 7.15, SD = 1.50) and SMT (M = 6.04, SD = 1.71) ranged between somewhat (5) to very credible (9), but favored ERP over SMT (p < .01), indicating that ERP was judged to be a more credible treatment approach than SMT. Patient-rated expectancy of improvement scores (0–100%) were also significantly higher for ERP (M = 62.00, SD = 20.98) compared to SMT (M = 46.1, SD = 23.07) (p < .01) indicating that individuals assigned to ERP were more confident that the therapy would be helpful for their OCD.

Overall, working alliance (WAI) scores for both groups were high, with item scores averaging approximately 6 (very often) points out of a possible 7 (always), indicating that the therapeutic relationship in both ERP and SMT was strong (see Table 5). These overall positive.

Table 5.

Working alliance inventory scores by condition.

Item	SMT	ERP	Difference	p
WAI Tasks	22.74	24.78	2.04	0.0051
WAI Bonds	24.47	25.04	0.56	0.369
WAI Goals	24.54	25.44	0.90	0.1051
WAI Total	71.75	75.19	3.44	0.0445

Scores notwithstanding, participants in the ERP group reported significantly higher scores on the WAI Tasks and WAI total scores at the treatment midpoint and endpoint compared to SMT.

To further investigate the impact of treatment expectations and working alliance on the comparison between ERP and SMT, we conducted an additional post-hoc multi-level modeling analysis, controlling for differences in expectancy of improvement and total WAI scores, and found that ERP remained a statistically significant predictor of C/YBOCS improvements over SMT (see Supplemental Table A).

Finally, for both study conditions, session attendance was high. The mean number of sessions attended for ERP was 10.74 (SD = 2.56) and 10.41 (SD = 2.54) for SMT (p = .4690). Only 6 of 63 individuals dropped out of ERP and 5 of 63 individuals dropped out of SMT before meeting the study threshold as a treatment completer (at least 8 of 12 scheduled sessions attended).

4.6. Adverse events

Finally, one possibly related adverse event occurred (symptom worsening during treatment forcing withdrawal from study) and one mild, related, adverse event occurred (participant experienced peripheral nerve stimulation during the MRI and withdrew from the study prior to randomization).

5. Discussion

This investigation provided a rigorous test of exposure and response prevention-based, cognitive behavioral therapy (ERP) versus stress management training (SMT) for adults and adolescents with OCD. Our multivariate analyses indicated that our 12-session ERP intervention was superior to a 12-session, multi-component SMT in improving OCD symptoms when considering the combined sample of adolescents and adults. Further, when comparing ERP and SMT outcomes among adolescents and adults separately, statistically and clinically significant improvements in OCD symptom measures for adolescents and adult participants randomized to ERP over SMT were also observed. There were no significant differences in OCD symptom response between adults and adolescents who received ERP by end of treatment. However, adolescents responded more quickly to ERP than adults, with superiority of ERP to SMT apparent mid-way through treatment (week 6) in adolescents, but not until the end of treatment (week 12) in adults. The lack of significant pre-to post differences in ERP outcomes for adolescents versus adults fits with a recent meta-analysis finding no significant difference in outcome for youth and adults given ERP for OCD (Reid et al., 2021) yet differs from a previous meta-analysis where superior outcomes were found for youth given ERP compared to adults (Olatunji et al., 2013). The finding in the present trial, that adolescents and adults experienced similar OCD improvement levels pre-to post ERP, may be more definitive compared to the meta-analysis-derived findings given that the adolescents and adults were treated with the same ERP protocol.

Bivariate analyses involving the combined sample of adolescents and adult participants also revealed clear superiority of ERP over SMT with the large majority of persons receiving ERP meeting study criteria as treatment responders (78.97%) or partial responders (7.02%) whereas only a small percentage of persons assigned to SMT met study criteria as responders (15.9%) or partial responders (15.9%). The same pattern held true for remitter status with over one third of those randomized to ERP (38.6%) meeting this standard compared to only 7.02% of those assigned to the SMT group.

As for the fidelity of the active (ERP) and comparator (SMT) interventions delivered in this study, ratings of therapist adherence to protocol exceeded standard quality thresholds for both conditions. Retention was high in both the ERP and SMT conditions with nearly all participants meeting study criteria as a treatment “completer” by attending at least 8 of the 12 scheduled sessions. Further, working alliance scores were high in both the SMT and ERP groups but the WAI task and WAI total scores were slightly higher for those who received ERP. This suggests that those receiving ERP rated the therapeutic tasks (e.g., exposure and response prevention) more positively compared to the therapeutic activities (e.g., relaxation, breathing training) involved in SMT. Further, ratings of intervention credibility and expectancy of improvement at the beginning of treatment were higher for those assigned to ERP over those assigned to SMT. These working alliance and expectancy differences indicate that some of the outcome advantages for ERP over SMT may be due to non-specific factors. However, it is important to note that post-hoc multi-level modeling, controlling for differences in treatment expectancies and working alliance between ERP and SMT, revealed that ERP remained a statistically significant predictor of C/YBOCS improvements over SMT, even after controlling for these factors. Noteworthy, higher working alliance and expectancy ratings among patients assigned to ERP suggests that this treatment is preferred by patients, relative to a non-specific relaxation-based therapy, despite the challenge of facing fear-triggering exposures while resisting compulsions.

Comparing outcomes for those assigned to ERP in the present study to previously published studies, effect sizes for ERP in the present study align with those reported in a recently published meta-analytic study comparing ERP to psychological placebo conditions (e.g., supportive therapy, relaxation therapy, anxiety management) for adults and children with OCD (Reid et al., 2021). Further, the superiority of ERP to SMT in this study aligns with results from previously published studies comparing ERP to a multi-component SMT control condition for adults with OCD (Lindsay et al., 1997; Simpson et al., 2008). Finally, the total number of participants (N = 126) participating in this RCT makes it among the largest studies of its type available in the literature to date (Reid et al., 2021).

The significance of the current study’s finding of a clear superiority of ERP over SMT is underscored by the reality that, despite ERP being the gold standard therapy for OCD, patients seeking treatment for OCD are often offered components of SMT instead (Voort, Svecova, Jacobson, & Whiteside, 2010). A likely factor driving the use of SMT components is reluctance on the part of some clinicians and patients to engage in exposure-based interventions out of concern that they may cause harm in most or in some circumstances (Olatunji, Wolitzky-Taylor, Willems, Lohr, & Armstrong, 2009). This reluctance occurs despite evidence that exposure-based approaches are both safe and effective for the majority of patients who engage in them (Law & Boisseau, 2019). The current trial adds to the evidence that ERP is a safe treatment given that no ERP-related adverse events were reported by participants. Also, it is interesting to note that although working alliances were strong in both the ERP and SMT groups in the present study, some aspects of the working alliance were actually stronger for those assigned to ERP compared to SMT, suggesting that ERP treatment is not arduous or off-putting to deliver or receive.

Another significant aspect of the current study is that it is the only available data directly comparing response rates between ERP and SMT where adolescent and adult participants were exposed to the same ERP protocol within the same study. Responses to ERP were substantial and similar across both groups except for a somewhat more rapid response to ERP among adolescent participants. The single protocol used in the present study borrowed elements common in adult ERP (e.g., significant psychoeducation about the nature and causes of OCD) and approaches present in most ERP protocols for adolescents with OCD (e.g., externalizing OCD, coping self-statements, sessions that include family members) to build a treatment that was acceptable to both adolescents and adults. Our equivalent OCD-specific outcomes for those randomized to ERP across age groups, high levels of treatment retention, positive intervention fidelity rating and strong ratings of intervention credibility, are all indicators that we have designed a version of ERP for OCD that is useful across a broad age-span. One advantage of this versatile intervention is that it has the potential to reduce training burdens for those who plan to treat both adolescents and adults with OCD. A second potential advantage associated with this treatment is that it allows for adolescents and adults to benefit from techniques that might not have been emphasized if they were given a protocol specifically designed for one group (e.g., externalizing OCD as an enemy to battle presented to adults with OCD).

Another interesting finding relates to improvements in generalized anxiety as measured by the GAD-7. Even though SMT included interventions typically used for individuals with generalized anxiety (Cuijpers et al., 2014), adolescent participants assigned to ERP experienced reductions in generalized anxiety that were similar to those observed among those assigned to SMT. In the adult sample, generalized anxiety symptoms did not change over the course of treatment, irrespective of whether a participant was assigned to ERP or SMT. These findings suggest that when co-occurring generalized anxiety symptoms are present among individuals with OCD, perhaps the best approach is to provide ERP alone as a starting point and then consider adding additional stress management approaches if substantial generalized anxiety symptoms remain after completing ERP.

Finally, like the above findings related to co-occurring generalized anxiety symptoms among adolescent participants, depression scores for adolescents declined similarly throughout treatment, irrespective of assignment to ERP or SMT. Depression scores did not significantly improve over the course of treatment for adults assigned to ERP or SMT. The implications of these findings are somewhat limited given that potential participants with co-occurring major depressive disorder were excluded from the sample due to the neuroscience-focused aims of this study.

5.1. Limitations

One limitation of this study is that it includes only participants who were willing to join a study that included MRI scans before and after treatment. It is possible those who fear and avoid MRI scans would have also been less likely to engage in exposure exercises associated with ERP for OCD if they were included in the trial.

A second limitation concerns the inclusion criteria necessitated by the neuroscience-related aims in this study including excluding those with current major depression, certain other lifetime co-occurring disorders and potential participants taking certain medications. These exclusions somewhat limit the generalizability of our findings to the broader population of persons with OCD. This limitation may be particularly relevant with respect to our adult participants given that higher levels of psychiatric comorbidity and greater use of medication are typically observed among adults with OCD compared to youth with OCD (Farrell, Barrett, & Piacentini, 2006).

A third limitation involves whether the treatment experience between adolescents and adults may have differed in certain ways (e.g., more family involvement in ERP exercises, more vigorous verbal reinforcement for ERP adherence, greater use of externalizing OCD as an enemy to battle for adolescent participants). Although we did not observe evidence of age-related tailoring in our audio-recorded fidelity monitoring activities, it is possible that this occurred for some participants. Future research would benefit from documenting potential naturalisic ERP delivery tailoring in comparative efficacy trials involving ERP for youth and adults with OCD.

A fourth limitation is the somewhat lower baseline C/YBOCS score at baseline for the SMT group, but it is important to note that the multilevel modeling approach controlled for this baseline difference. Fifth, expectancy/credibility ratings and therapeutic task-related working alliance ratings were higher for participants randomized to ERP compared to those randomized to SMT which likely relates to knowledge about the use of ERP techniques for OCD disseminated from various sources and possibly from the language related to the effectiveness of CBT versus stress management in the consent document. Although both ERP and SMT were rated as credible and likely to be helpful and working alliance scores in both conditions were high, higher credibility and expectancy and stronger working alliances for ERP may have played some role in the observed advantages of ERP over SMT. However, in post-hoc multi-level modeling controlling for differences in treatment expectancies and working alliance between ERP and SMT, ERP remained superior to SMT from pre to post treatment.

Finally, the study does not provide follow up outcome ratings because participants were immediately offered access to ERP or SMT when they finished their assigned study intervention. Future research investigating long term outcome differences between ERP and SMT is needed.

5.2. Implications and conclusions

Overall, these results indicate that cognitive behavioral therapy focused on exposure and response prevention is superior to stress management in the treatment of OCD among adults and adolescents. Additionally, this study finds that ERP response is similar between adults and adolescents exposed to the same ERP protocol that blends features typically present in adult and child versions of ERP for OCD. These results are especially valuable given that both ERP and SMT were delivered according to protocol, with high participant retention in both groups. These findings are particularly significant given that stress management interventions are often offered to OCD patients in the community, yet this trial indicates that they are credible but not effective in reducing OCD symptoms. Publication now of an updated, unusually large RCT – using a single ERP protocol to treat both adolescents and adults in a structured, time-limited (12 session) therapy and showing clear superiority of ERP – could be helpful in public health efforts to get empirically-proven treatments to those who need them. This may be especially important for OCD, which is a particularly severe and disabling disorder (Eisen et al., 2006), and one that is only rarely treated properly (Schwartz, Schlegl, Kuelz, & Voderholzer, 2013).

Data availability

The data that has been used is confidential.