Exposure-Focused CBT Outperforms Relaxation-Based Control in an RCT of Treatment for Child and Adolescent Anxiety

Abstract

Objective:

The relative contribution of individual cognitive behavioral therapy (CBT) components to treatment outcomes for child anxiety disorders (CADs) is unclear. Recent meta-analyses suggest that exposure may be the primary active ingredient in CBT for CADs, and that relaxation may be relatively less effective. This brief report tests the hypothesis that exposure-focused CBT (EF-CBT) would outperform a relaxation-based active therapy control (Relaxation Mentorship Training; RMT) for the treatment of CADs.

Method:

Participants were 102 youth with CADs (mean age=11.91, 26 males; 76.4% White, 14.7% Multiracial, 3.9% Black, 3.9% Asian, 0.9% other/do not wish to identify) as part of an ongoing neuroimaging randomized controlled trial. Participants were randomly assigned (ratio 2:1) to receive 12 sessions of EF-CBT (n=70) or RMT (n=32). Clinical improvement was measured at Week 12 (Clinical Global Impression – Improvement scale; CGI-I); treatment response was defined as receiving a rating of ‘very much’ or ‘much improved’ on the CGI-I. Anxiety severity was measured at Weeks 1, 6, 9, 12 (Pediatric Anxiety Rating Scale; PARS). Outcome measures were completed by an independent evaluator unaware of condition.

Results:

EF-CBT exhibited 2.98 times higher odds of treatment completion than RMT; 13 treatment non-completers were included in analyses. Estimated treatment response rates were higher for EF-CBT (57.3%) than for RMT (19.2%). Longitudinal analyses indicated that EF-CBT was associated with faster and more pronounced anxiety reductions than RMT on the PARS (Hedges’ g=.77).

Conclusions:

Results suggest that EF-CBT without relaxation is effective for CADs, and more effective than a relaxation-based intervention.

Anxiety disorders are the most prevalent psychiatric presentation of childhood (Merikangas et al., 2010). Cognitive behavioral therapy (CBT) is a well-established treatment for Child Anxiety Disorders (CADs), and typically comprises a variety of components including: psychoeducation, relaxation training, cognitive restructuring, and exposure (e.g. Walkup et al., 2008). CBT is supported as an improvement over treatment-as-usual by a wealth of research, including the landmark Child/Adolescent Anxiety Multimodal Study (CAMS, e.g. Walkup et al., 2008). However, few studies have compared CBT to active control, with mixed results (James et al., 2020; Silk et al., 2018), and relative contributions of the elements of CBT remain poorly understood (Whiteside et al., 2020). Though CBT has primarily been tested as a package, community-based clinicians tend to favor relaxation-based strategies, while specialty clinics prioritize exposure (Stewart et al., 2016; Whiteside et al., 2016). This differential implementation is potentially problematic, as meta-analyses suggest that relaxation strategies may not provide substantial benefit, and that exposure may be the active ingredient in treating CADs (Ale et al., 2015; Whiteside et al., 2020).

While no large-scale dismantling studies have been completed, one pilot study compared parent-guided exposure to non-exposure components of CBT for CADs and found differential effect sizes favoring exposure for many variables, including anxiety severity and overall improvement (Whiteside et al., 2015). Similarly, shifts in treatment trajectories following the introduction of different CBT components show evidence of improvement after introduction of cognitive restructuring and exposure but not relaxation (Peris et al., 2015). Meta-analytic work has revealed that including relaxation in CBT for CADs is associated with either equivalent (Ale et al., 2015) or worse treatment effect sizes (Whiteside et al., 2020), the relative contribution of cognitive restructuring is unclear (Whiteside et al., 2020), and the inclusion of more in-session exposures is associated with larger treatment effect sizes (Whiteside et al., 2020). Taken together, the evidence appears to favor reducing reliance on relaxation strategies and emphasizing exposures.

In response to this work, and the need to further distinguish the benefits of relaxation from exposure, this report directly compares a CBT package emphasizing exposure and omitting relaxation (Exposure-focused CBT; EF-CBT) to a relaxation-based active control (Relaxation Mentorship Training; RMT) in a randomized controlled trial (RCT) for CADs. The umbrella, neuroimaging RCT is ongoing (Premo et al., 2020); however, due to COVID-19, the remaining participants are receiving therapy remotely. Thus, given the large sample size relative to other trials examining CADs (e.g. Ale et al., 2015) and due to potential confounding factors related to virtual therapy provision, an interim analysis of clinical outcomes comparing RMT to EF-CBT is presented here. We hypothesized that EF-CBT, as compared to RMT, would be associated with better outcomes as defined by higher rates of treatment response as well as greater reductions in anxiety severity.

Methods

Sample

The current study was conducted in an academic medical center in the Midwest. Participants were 102 youth with CADs (ages 7–17 years; 26 males) recruited through clinic referrals, community flyers, and online advertisements. Consistent with prior work (Walkup et al., 2008), clinical anxiety was defined as meeting criteria for one or more anxiety disorders, including generalized anxiety disorder, social anxiety disorder, and separation anxiety disorder. Participants with other specified anxiety disorder were also included, supporting recent emphasis on a more dimensional approach to conceptualizing anxiety among youth (Insel et al., 2010). Diagnoses were made based on a structured clinical interview administered by a masters-level clinician.

Study inclusion criteria required anxiety to be the primary source of interference and distress, although comorbidities, such as attention-deficit/hyperactivity, obsessive–compulsive, and oppositional–defiant disorders were allowed to increase generalizability (Walkup et al., 2008). Patients with current major depression diagnoses were excluded; however, other depressive disorders (e.g. persistent depressive disorder and other/unspecified depressive disorders) were accepted if less interfering than anxiety. Children with intellectual disability, comorbid autism spectrum, substance use, psychotic disorders and/or presenting an acute risk to self or others were excluded. Across the sample, the most commonly occurring non-anxious comorbidities were depressive disorders (14%) and ADHD (11.2%) followed by enuresis (5.9%) and OCD (5.9%). There were no significant differences in rates of comorbidities between intervention groups (Table 1). No psychoactive medications were allowed, except for stable doses of stimulants in youth with comorbid ADHD (n=3). Included participants completed their course of therapy prior to the initiation of alternate study procedures due to COVID-19 in the Spring of 2020. Individuals assigned to RMT were offered EF-CBT after completing their primary treatment. Participants and guardians provided informed assent/consent in compliance with the policies of the Michigan Medicine Institutional Review Board (HUM00118950).

Table 1.

Sample Baseline Characteristics and p-Values for Tests of Covariate Balance Across Conditions

	EF-CBT (n = 70)	RMT (n = 32)	P value
Age – mean	11.86 +/− 3.1	12.03+/− 3.1	0.80
Sessions Attended – mean	10.97 +/− 2.3	9.25 +/− 3.4	< 0.01
Female sex – no. (%)	51 (72.9)	25 (78.1)	0.57
Race – no. (%)
White	56 (80.0)	22 (68.8)	0.21
Black	0 (0)	4 (12.5)	< 0.01
Asian	3 (4.3)	1 (3.1)	0.78
Multiracial	11 (15.7)	4 (12.5)	0.67
Undisclosed	0 (0)	1 (3.1)	--
Low Socioeconomic Status¥ – no. (%)	12 (17.1)	8 (25)	0.35
Primary Anxiety Diagnosis – no. (%)
Generalized Anxiety Disorder	40 (57.1)	14 (43.8)	0.21
Social Anxiety Disorder	18 (25.7)	9 (28.1)	0.80
Separation Anxiety Disorder	8 (11.4)	8 (25.0)	0.08
Other/Unspecified Anxiety Disorder	4 (5.7)	1 (3.1)	0.57
Non-Anxious Comorbid Diagnoses – no. (%)
Additional anxiety diagnosis	50 (71.4)	19 (59.4)	0.23
Depressive Disorders†	10 (14.3)	4 (12.5)	0.81
Obsessive Compulsive Disorder	4 (5.7)	2 (6.3)	0.92
Traumatic Stress Disordersꜝ	0 (0)	1 (3.1)	--
Enuresis	5 (7.1)	1 (3.1)	0.42
Binge Eating Disorder	0 (0)	1 (3.1)	--
Attention-Deficit/Hyperactivity Disorder	6 (8.6)	6 (18.8)	0.14
Oppositional Defiant Disorder	1 (1.4)	0 (0)	--
Tic Disorders	1 (1.4)	0 (0)	--

Note. Table presents demographic and baseline diagnostic presentations

^¥Low SES was defined as a score of 3 or greater on the Hollingshead Scale – Ranking variable (range 1–5)

^†Persistent Depressive Disorder, Unspecified Depression, Adjustment Disorder with Depressed Mood

^ꜝPTSD, Unspecified Trauma Related Disorde

Intervention Details

Both interventions involved 12 weekly sessions (approximately 45–60 minutes; for details see Table 2). Treatment completion was defined as completing >7 sessions of the assigned therapy. Across conditions, guardian/parent involvement was encouraged in sessions 1, 6, and 12; guardians were invited to join additional sessions as needed. Participants received a workbook to assist with skill acquisition and homework practice. If participants appeared to deteriorate during the course of the study, the study team met to determine best practice recommendations (e.g. withdrawing and providing referrals for subjects for whom non-anxiety symptoms such as depression or suicidality emerged as primary during the course of treatment).

Table 2.

Session Level Details by Intervention

Session Number	EF-CBT: Major Topic	RMT: Major Topic
1	*Building Rapport and Treatment Orientation	*Building Rapport and Treatment Orientation
2	Somatic Feelings and Externalization of Anxiety	Getting to Know You
3	Cognitive Restructuring and Problem Solving	Diaphragmatic Breathing
4	Introduction to Exposure and Initial Practice	Interesting Things
5	Exposure Review and Practice	Monitoring Muscle Tension
6	*Family Session (with Exposure Practice)	*Family Meeting
7	Exposure Review and Practice	Autobiographical Diaries
8	Exposure Review and Practice	Progressive Muscle Relaxation
9	Exposure Review and Practice	Coloring for Relaxation
10	Exposure Review and Practice	Coloring for Relaxation
11	Exposure Review and Practice	Learning about Healthy Lifestyles
12	*Final Practice and Celebrating Success	*Wrapping up

Note. Table presents topics covered in each session for both interventions

^*indicates that a family member is invited to be present

Exposure-focused Cognitive Behavioral Therapy (EF-CBT).

The EF-CBT intervention adopted components of established CBT manuals, including the Coping Cat manual and workbook (Kendall & Hedtke, 2006a, 2006b). However, EF-CBT was designed to move quickly to exposure, omitting relaxation components typically delivered in the early course of CBT and reducing emphasis on cognitive restructuring. Specifically, the EF-CBT manual delivered psychoeducation in two initial sessions, cognitive restructuring in a third session (with opportunities to practice in subsequent sessions, if indicated), and initiated exposure by the fourth session. Remaining sessions (5–12) focused on completing exposures of increasing difficulty while also incorporating inhibitory learning techniques (Craske et al., 2014).

Relaxation Mentorship Training (RMT).

The RMT intervention was adapted from prior work detailing relaxation (Kendall & Hedtke, 2006b, 2006a) and other strategies for youth with CADs (Lindsay et al., 1997; Reynolds et al., 2000). The relaxation components closely resembled those taught in standard child CBT packages, including progressive muscle relaxation and diaphragmatic breathing (e.g. Kendall & Hedtke, 2006a, 2006b). To match the 12-week EF-CBT intervention, RMT also included non-anxiety-specific education and support to prevent boredom and increase plausibility. These components included elements of credible, non-specific attention control conditions used in prior psychotherapy research, specifically autobiographical writing (Lindsay et al., 1997; Reynolds et al., 2000). Additional mentorship and engaging activities were incorporated to maintain engagement, rapport, and credibility. Sessions 3, 5, and 8 were devoted to the delivery of relaxation techniques, with these techniques assigned for homework practice as well.

Therapist Training and Fidelity

Clinical supervision was provided by a licensed psychologist and/or a licensed social worker with established CBT expertise. Study therapists were masters- or doctoral-level clinicians who received standardized training in both treatment manuals, and were assigned participants across both conditions. After therapists were formally trained, the clinical supervisor listened to available audio recordings from the first course of treatment for both EF-CBT and RMT conditions. To identify any cross-contamination of exposure therapy in the RMT condition and relaxation in the EF-CBT condition, the supervisor audited portions of sessions for additional study participants.

Measures

Treatment Responder Status.

The Clinical Global Impression–Global Improvement (CGI-I; Guy, 1976) is a widely used, brief clinician-rated instrument for the assessment of patient improvement since the start of treatment (week 1). The CGI-I was completed at weeks 6, 9, and 12 (post-treatment) by an independent evaluator. The scale ranges from 1 to 7, with lower scores reflecting greater improvement since beginning treatment, and higher scores reflecting worsening symptoms with treatment. The CGI-I was the categorical primary outcome measure, with a score of 1 or 2 at post-treatment, indicating ‘very much improved’ or ‘much improved’, considered to reflect significant improvement after the course of treatment.

Anxiety Severity.

The Pediatric Anxiety Rating Scale (PARS) is a well-established, semi-structured clinical interview used to ascertain the severity of pediatric anxiety disorder(s) along a continuous range (Group, 2002). The PARS was administered at weeks 1, 6, 9, and 12 by an independent evaluator unaware of study condition. Children and guardian(s) were interviewed together or separately, with most families choosing to be interviewed together. The PARS includes a checklist to ascertain anxiety symptom type(s) (e.g., separation, social, generalized, etc.), and then collapses across types to generate an overall severity score (Group, 2002). Consistent with CAMS (Walkup et al., 2008), PARS is our continuous primary outcome measure. PARS scores range from 0 to 30, with a higher score reflecting higher anxiety severity.

Planned Analyses

Demographics.

Analyses were conducted to examine the balance of covariates and other demographic categories of interest across conditions. Specifically, equal variance t-tests were conducted for continuous covariates, and Pearson’s χ² were conducted for categorical covariates (Table 1).

Treatment Responder Status.

Logistic-regression models compared treatment groups (EF-CBT and RMT) on the proportion of participants achieving treatment responder status (CGI-I ≤ 2). Age, gender, and number of sessions attended were covariates. We considered female gender and RMT assignment to be the reference group; age was centered at the group median (12 years), and sessions attended was centered at 12 (weeks).

Missing Data.

Fourteen participants did not have CGI-I values at week 12 (CBT: n=8, 11.4%; RMT: n=6, 18.8%). Week 12 CGI scores were multiply imputed (10 fold imputation) using the “mice” package in R (R Core Team, 2018) to avoid dropping incomplete cases entirely. This approach is robust in that any potential systematic differences in treatment response rates among non-completers would bias results toward the null.

Anxiety Severity.

Linear longitudinal mixed models were fit (using the lme4 package in R 3.5.0; Bates et al., 2015; R Core Team, 2018) to test effects of treatment group, time and treatment group × time interaction on anxiety severity trajectories (PARS). The resulting model included random intercepts and coefficients (on time) for each patient and considered age, gender, and sessions attended as covariates. As above, female gender and RMT assignment were the reference conditions, and age and total sessions attended were centered at 12 years and 12 weeks, respectively. In addition, to promote interpretability, time on treatment was centered at week 1. With this covariate organization, the global model intercept can be interpreted as the expected week 1 PARS score for a 12-year-old female patient who was assigned to the RMT group and eventually completed 12 weeks of therapy. We compared EF-CBT and RMT groups on average scores on the PARS using a series of model-based comparisons at weeks 1, 6, 9, and 12.

Missing Data.

The primary continuous model was fit using only observed data (i.e., no imputation), given that mixed effects models are designed to handle data that are “missing at random” (Fitzmaurice et al., 2008). As above, results of this analysis may be biased towards the null. For comparison with CAMS (e.g. Walkup et al., 2008), a secondary analysis was performed in which the last PARS observation was carried forward for individuals with missing data on the PARS. This approach assumes that study non-completers would not have improved further over the remainder of the 12-week treatment period had they continued to receive treatment.

Results

Sample

The interim analyses reflect 200 subjects who were assessed for eligibility during an initial assessment that included the KSADS and PARS (Fig. 1). Of these, 79 did not meet criteria, 14 declined further participation, eight participated during COVID-era, and one was lost to follow up, leaving 102 who were eligible for these analyses. Of these, 13 (EF-CBT=6; RMT=7) were non-completers (defined as attending <8 sessions). EF-CBT (91.4% completion rate) exhibited 2.98 times higher odds of treatment completion than the RMT group (78.1% completion rate; χ²(1) = 3.50, p = 0.06; Figure 1 displays reasons for treatment withdrawal). Participants assigned to EF-CBT attended an average of 11.0 sessions, whereas RMT participants attended an average of only 9.3 sessions (t(44.2)=2.55, p<.05).

Figure 1.

CONSORT diagram

Analyses

Participants were randomly assigned (ratio of 2:1) to receive either EF-CBT or RMT. Randomization was stratified by age and gender. Ultimately, 70 participants were assigned to EF-CBT, and 32 participants to RMT. Across EF-CBT and RMT groups, most subjects were diagnosed with two or more anxiety disorders (n=71; 70% of EF-CBT, 68.9% of RMT; Table 1); non-anxious secondary disorders were evenly distributed across groups (n=43; 41.4% of EF-CBT, 43.8% of RMT; Table 1). At week 1, subjects in both treatment arms had comparable levels of anxiety, as assessed by the PARS (Table 3). A difference across conditions emerged on race/ethnicity; participants who identified as Black (n=4) were more likely to be assigned to the RMT condition than to the CBT condition (p<0.01; Table 1). We compared age, gender, and number of treatment sessions attended across the EF-CBT and RMT groups.

Table 3.

Treatment Outcomes by Condition

	EF-CBT	RMT
CGI-Improvement Scale Independent Evaluator (95% CI)
% treatment responders†	57.3% (42.4–71.0)	19.2% (6.2–46.2)
Pediatric Anxiety Rating Scaleꜝ
Week 1 (95% CI)	19.2 [18.2, 20.2]	19.3 [17.8, 20.9]
Week 6*	16.3 [15.4, 17.3]	18.1 [16.6, 19.5]
Week 9**	14.6 [13.6, 15.6]	17.3 [15.7, 18.9]
Week 12***	12.9 [11.7, 14.0]	16.5 [14.7, 18.3]

Note. This table presents proportion of participants who were designated as treatment responders on CGI-I, as well as means for anxiety severity on PARS across the duration of the study for each condition.

^†Treatment responders defined as individuals receiving a 1 (very much improved) or 2 (much improved) on the CGI-I. Estimated proportions as determined by a logistic regression model reported

^ꜝEstimated means as determined by a linear mixed-effect model

^*significant difference between groups p<.05.

^**significant difference between groups p<.01.

^***significant difference between groups p<.001.

Treatment Responder Status

A logistic regression model compared treatment response across the two treatment groups, controlling for age, sex, and completed weeks of therapy. Treatment response was more likely in the EF-CBT than RMT condition (B=1.73±0.66, p=.01), with the estimated probability of treatment response 57.3% (95% CI: [42.7, 71.0]) for EF-CBT compared to 19.2% for RMT (95% CI: [6.2, 46.2]; Table 3). Adjusting for sessions attended, a patient in the EF-CBT group was 5.64 times more likely to respond to treatment than a patient in the RMT group (95% CI: [1.53, 20.75]). The number needed to treat (NNT) estimate was 2.63 (95% CI: [1.78, 8.49]), indicating that about 2.63 people need to be treated with the EF-CBT condition over the RMT condition to mitigate excess treatment non-response (see Table 4 for all model parameters).

Table 4.

Logistic Regression Parameters

Model	β	SE(β)	z	p
(Intercept)	−1.44	0.65	−2.22	0.03
EF-CBT	1.73	0.66	2.64	0.01
Age (years)	−0.08	0.07	−1.13	0.26
Gender (male)	−0.58	0.55	−1.06	0.29
Sessions Completed	0.21	0.20	1.04	0.31

Note. Logistic regression of CGI-I independent evaluator report on treatment condition.

Anxiety Severity

A mixed-effects model examined differences between treatment conditions on anxiety severity over time. PARS scores decreased significantly for both the EF-CBT and RMT groups, but improvement was faster and more pronounced for the EF-CBT group (decrease of 0.578 points/week; SE=0.047, 95% CI: [0.485, 0.672], p<0.001) than for RMT (decrease of 0.256 points/week; SE=0.073, 95% CI: [0.109, 0.401], p<0.001; Figure 2). Model-based comparisons of PARS scores showed significant differences between conditions, with EF-CBT outperforming RMT at weeks 6, 9 and 12 (Table 3). Hedges’ g was computed to compare estimated means at week 12, resulting in a score of 0.77 (95% CI: [.34, 1.19]), indicating a medium-large effect of EF-CBT over the RMT condition (See Table 5 for all model parameters). Additionally, there was an effect of age such that after 12 weeks of therapy, older youth had modestly higher PARS on average than their younger counterparts. Although the majority (86.4%) of patients improved, holding all else constant a 4.14 year age gap (95% CI: [0.37, 7.91]) predicted a 1-point difference in final PARS. In secondary analyses considering the last observation carried forward (LOCF), findings remained the same (Supplementary Materials - S1).

Figure 2.

Longitudinal Graph for Model Estimate Treatment Means

Note. Estimated mean PARS scores across 12 weeks of treatment.

Table 5.

Longitudinal Model Fixed Effect Parameters

	β	SE(β)	z	p
(Intercept)	19.33	0.76	25.34	< 0.001
Week	−0.26	0.07	−3.49	< 0.001
Treatment (EF-CBT)	−0.11	0.82	−0.14	0.89
Age (years)	0.24	0.11	2.19	0.03
Gender (male)	−0.83	0.79	−1.05	0.29
Sessions Completed	−0.18	0.14	−1.32	0.19
Week × EF-CBT	−0.32	0.09	−3.73	< 0.001

Note. Fixed effects summaries from the Longitudinal PARS model. Intercepts can be interpreted as the expected week 1 PARS scores for a 12-year-old female patient who was assigned to the RMT group and eventually completed 12 weeks of therapy.

Discussion

This report presents interim findings from an ongoing RCT (Premo et al., 2020) in which youth with CADs are randomized to EF-CBT or RMT. We hypothesized that EF-CBT would outperform RMT, yielding higher rates of treatment response on the CGI-I and greater pre- to post-treatment reduction in anxiety severity on the PARS. Our results provided strong support for this hypothesis. While adjusting for sessions attended, youth were 5.64 times more likely to be identified as a treatment responder on the CGI-I if they were assigned to the EF-CBT condition. Moreover, we found that EF-CBT was associated not only with greater improvements in anxiety symptom severity (PARS) post-treatment, but also with faster response to treatment compared to RMT. Indeed, EF-CBT outperformed RMT at week 6, and this gap grew through week 12, as individuals receiving EF-CBT continued to improve while rates of improvement for RMT slowed.

The results from analyses of both categorical (CGI-I) and dimensional (PARS) response provide clear evidence that EF-CBT, which eliminated traditional relaxation training, is more effective at reducing anxiety severity than a treatment focused on relaxation training. Participants receiving EF-CBT improved more overall and improved more quickly throughout treatment delivery. These findings support work suggesting the importance of exposure therapy in treatment of CADs, provide further evidence that exposure therapy is palatable and effective for children even without relaxation, and underscore the importance of improving implementation of exposure over relaxation among clinicians treating youth with anxiety (Ale et al., 2015; Becker-Haimes et al., 2017; Whiteside et al., 2020).

Covariates

We identified several covariates a priori: age, gender, and the number of sessions attended. In the CAMS study, older participants were less likely to enter remission (Ginsburg et al., 2011), and male sex predicted remission as well as lower baseline anxiety scores at follow-up (Ginsburg et al., 2014). However, within our sample, age and gender did not significantly improve within-sample prediction of binary treatment response. We did find an effect of age on PARS scores such that, across treatments, younger participants had slightly better post-treatment outcomes.

The third covariate, number of sessions attended, was included to ensure that any differences identified between treatment modalities were not due to differences in treatment completion rates. This was especially important given that participants assigned to EF-CBT attended more sessions on average than did those assigned to RMT (11.0 and 9.3 respectively), and were more likely to be treatment completers. However, we found that EF-CBT was more effective than RMT even when adjusting for sessions attended. Moreover, significant reduction in anxiety was noted earlier (i.e. with fewer sessions) for youth treated with EF-CBT than RMT.

Comparison to CBT for CADs

The present study examined a more exposure-focused course of CBT than prior studies (Ale et al., 2015; Walkup et al., 2008; Whiteside et al., 2020). Despite these differences, our study demonstrated similar findings as those found in CAMS across numerous dimensions including: pre-treatment anxiety symptom severity (current study: EF-CBT group=19.2 (95% CI: [18.2, 20.2]); CAMS: CBT group=18.9 (95% CI: [18.2–19.6])), treatment response (current study: EF-CBT group=57.3%, RMT group=19.2% (NNT=2.63, 95% CI: [1.78, 8.49]); CAMS study: CBT group=59.7%, placebo group=23.7% (NNT=2.8, 95% CI: [2.7, 3.0])), and anxiety severity reduction (current study: CBT group (Hedges’ g=.77); CAMS: CBT group (Hedges’ g=.31)). Thus, EF-CBT delivered in the current study was associated with significant treatment gains despite the omission of relaxation training and increased focus on exposures. Moreover, response rates to RMT delivered without the other elements of CBT yielded response rates no better than those exhibited among patient receiving placebo in prior work.

Limitations and Future Directions

The current study presents interim clinical analyses from an ongoing neuroimaging RCT. The analyzed sample is somewhat homogenous: participants are predominantly Caucasian, from non-low-socioeconomic-status backgrounds, and not receiving psychopharmacological treatment (with the exception of stimulants). This potentially limits the generalizability of the findings.

Additionally, the sample size while large, was not sufficient to examine a number of treatment predictors. Upon study completion, we aim to examine whether race/ethnicity, principal anxiety diagnosis, comorbid diagnoses, or treatment delivery mechanism (in-person vs. teletherapy) serve as moderators of treatment outcome. The current manuscript also does not examine guardian- or child-report of symptoms across conditions. Such measures are being administered; however, they are not examined or reported here. Reports from these informants may confirm current findings, or may demonstrate weaker or no differentiation between conditions. Conclusions based on these findings should thus be considered with caution. We plan to examine informant reports, along with treatment predictors upon study completion.

Another limitation is that RMT likely does not mimic protocols typically delivered in the community. The RMT protocol engaged only three sessions of pure relaxation training, which is fewer than the nine sessions of exposure offered in the EF-CBT condition. This decision was made to best mirror the relaxation activities offered in CBT packages for youth (e.g. Coping Cat) and to maintain youth engagement. However, even with this protocol, participants assigned to RMT completed significantly fewer sessions on average than those assigned to EF-CBT. Future work could incorporate more relaxation sessions to assess palatability, and introduce other common strategies like problem solving, to better replicate treatment as usual. This study is also not a true dismantling study. Further research is needed to identify the unique contribution of each CBT component to treatment gains. Finally, although our findings provide strong support for EF-CBT, the treatment effect closely mirrors that found in CAMS. While this supports the finding that relaxation is not necessary for treating youth anxiety, more research is needed to determine whether EF-CBT might outperform a traditional CBT package for youth anxiety.

Conclusions

This RCT is the first to directly compare exposure-focused CBT, which eliminated traditional relaxation training, to a relaxation-based comparison intervention. Our findings demonstrate that EF-CBT is associated with higher treatment response rates and lower anxiety severity as compared to a relaxation based intervention, and that it offers similar treatment outcomes to the landmark CAMS trial (Walkup et al., 2008). This supports the growing literature indicating that CBT for CADs is effective without the use of relaxation strategies, and in fact outperforms a relaxation-based control. These findings have implications for the future refinement of interventions for CADs, the most prevalent psychiatric presentation of childhood.