Change during cognitive and exposure phases of cognitive-behavioral therapy for autistic youth with anxiety disorders

Abstract

Objective:

To evaluate improvement during the (a) cognitive and (b) exposure therapy phases of cognitive-behavioral therapy (CBT) for autistic youth with anxiety disorders.

Methods:

Participants were 148 autistic youth (ages 7–13; 77% male; 64% White) with clinically significant anxiety who were randomized to standard or adapted CBT. Clinician-rated anxiety severity was recorded at each appointment. Trajectories of change during (a) the cognitive phase and (b) the exposure phase of treatment were analyzed using piecewise multilevel modeling.

Results:

Compared to the psychoeducation and cognitive therapy phases, the exposure phase corresponded with significantly more rapid symptom reduction, b = −.11, 95% CI: −.13 to −.071. This finding was true for standard CBT, which included nine sessions prior to exposure initiation (on average), as well as for CBT personalized for autistic youth, which introduced exposure following the fifth session (on average). In contrast, compared with improvements during initial psychoeducation sessions, the introduction of cognitive skills corresponded with significantly slower symptom reduction, b = .066, 95% CI: .020 to .11. This finding was also true for both adapted and standard CBT.

Conclusions:

Results underscore the central role of exposure in the treatment of anxiety among autistic youth and question the utility of cognitive strategies without subsequent practice of these skills during exposure. Assuming a working relationship, therapists should consider moving through the pre-exposure coping phase of CBT quickly. Future research should evaluate the extent to which exposure can maintain its potency while minimizing coping skill preparation.

1. Introduction

Almost half of autistic youth are estimated to have an anxiety or obsessive-compulsive disorder (OCD; Van Steensel, Bögels, & Perrin, 2011). Fortunately, cognitive-behavioral therapy (CBT) has a strong and growing evidence base for this population (Perihan et al., 2020). Understanding the most central components of CBT would help inform efforts to refine and optimize this therapy, though minimal research has investigated anxiety improvement during different phases of CBT for this population.

CBT for youth with anxiety disorders often occurs in two phases: (a) psychoeducation and coping strategies, and (b) exposure therapy. Exposure has been identified as the most central component for nonautistic youth (Peris et al., 2015; Whiteside et al., 2020). Anxiety has been found to improve more rapidly following the introduction of exposure exercises (Peris et al., 2015), and multiple RCTs have shown that CBT is superior to control conditions only following the second half of CBT after exposure begins, but not following the first half that focuses more on psychoeducation and coping (e.g., Walkup et al., 2008; Wood et al., 2020). It follows that CBT protocols with more exposure sessions have shown stronger efficacy (Whiteside et al., 2020).

There is also preliminary evidence for initiating exposure immediately after psychoeducation for both autistic and nonautistic youth (Whiteside et al., 2020; Storch et al., 2019). One RCT tested this approach for autistic youth with anxiety disorders (n = 32), finding a 79% response rate and significant superiority relative to a treatment-as-usual control (Storch et al., 2019). This literature raises the question of how many pre-exposure coping-focused sessions are necessary, though no research has tested whether the potency of exposure varies based on the quantity of pre-exposure preparation.

Some research has investigated whether different CBT coping strategies contribute differentially to improvement in nonautistic youth. Multiple studies have found that change in anxious self-talk mediate improvement during CBT (Kendall & Treadwell, 2007; Hogendoorn et al., 2014), supporting the role of cognitive strategies. In Peris and colleagues’ (2015) study of rates of change during different phases of CBT, they found an acceleration of symptom improvement during the cognitive therapy phase but not during the relaxation phase. A meta-analysis found no consistent association between the amount of cognitive therapy or relaxation included in CBT and treatment outcome, however (Whiteside et al., 2020).

It is important to evaluate components of CBT for anxiety that are most helpful for autistic youth to make this treatment as effective and efficient as possible. To date, however, no studies have evaluated the relative importance of the two most central elements of CBT for autistic youth with anxiety disorders: exposure and cognitive therapy. This population may similarly benefit from cognitive therapy, though these skills may be more challenging for autistic youth due to the frequency of cognitive rigidity problems in this population (Kerns et al., 2016). Presentations of anxiety in autistic youth may be less responsive to cognitive techniques as well. For example, unusual, illogical fears that many autistic youth experience (Kerns et al., 2014) may not respond to a highly cognitive (i.e., logical) approach. Further, autistic youth have difficulties with social communication, which is likely integral to more “talk-heavy” psychotherapy approaches. For all these reasons, the relative importance of different phases of CBT may differ for autistic relative to nonautistic youth, though dismantling research has not been conducted with this population.

To this end, this study evaluated trajectories of change during different CBT phases for autistic youth with anxiety disorders. Symptoms were expected to improve more rapidly following the initiation of cognitive strategies and again when beginning exposure therapy based on work in nonautistic youth (Peris et al., 2015). This extended Peris and colleagues’ (2015) analysis by evaluating CBT phases in autistic youth with anxiety disorders and comparing protocols with different numbers of pre-exposure preparatory sessions. Specifically, this study compared standard CBT (exposure introduced at session 10) and CBT adapted for autistic youth (exposure introduced at session 6). It was hypothesized that symptom reduction would accelerate following the introduction of exposure across both CBT protocols.

2. Methods

2.1. Procedures

This study was a secondary analysis from a three-site RCT testing CBT for autistic youth (ages 7–13) and clinically significant anxiety. Institutional review boards at all sites approved the study. Youth were randomized in a 4.5:4.5:1 ratio to (1) standard CBT (2) CBT adapted for autistic youth or (3) treatment-as-usual. Therapists included advanced graduate students in clinical psychology and postdoctoral fellows who received expert training and supervision in both protocols. Please see Wood et al. (2020) for more procedural details.

2.2.1. Standard CBT

Standard CBT followed the Coping Cat program (Kendall & Hedke, 2006). The program includes 16 weekly 60-minute sessions. During the first half of Coping Cat, youth develop a coping plan through psychoeducation, self-awareness, relaxation, cognitive skills, and self-reinforcement. Cognitive skills include recognizing anxious self-talk, cognitive restructuring, and problem-solving. Next, youth participate in six exposure therapy sessions, which emphasizes practice of the coping plan. Parent-focused sessions are included at sessions 4 and 9.

2.2.2. Adapted CBT

The Behavioral Interventions for Anxiety in Children with Autism (BIACA) program was also tested (see Wood et al., 2020), which also includes 16 weekly sessions. Differences from Coping Cat include: 1) sessions are 90 minutes and are split evenly between children and parents, 2) exposure begins earlier after the child has demonstrated understanding of at least one coping skill, 3) modules are available for disruptive behavior, 4) modules are available for social skills, 5) special interests are incorporated to enhance engagement, and 6) a reward system is used. Sessions of standard and adapted CBT are summarized in the Supplement.

2.2. Participants

Only participants in CBT conditions were included in this secondary analysis (n = 148). Inclusion criteria were: ages 7–13, an autism spectrum disorder diagnosis, an IQ of 70 or higher, clinically significant anxiety or OCD, stability on psychiatric medication, and anxiety/OCD being considered primary. OCD was considered a qualifying diagnosis due to its similar cognitive-behavioral treatment and mechanisms as anxiety disorders (Abramowitz, 2018), as well as the frequent co-occurrence and overlap between anxiety/OCD and autistic features (Kerns et al., 2014). Exclusion criteria included participation in concurrent psychotherapy, past six-month suicidality, nonresponse to CBT in the past, and lifetime bipolar or psychotic disorder. See Wood et al. (2020) for more information.

2.3. Measure

2.3.1. Clinical Global Impressions-Severity (Guy, 1976).

The Clinical Global Impressions Scale-Severity (CGI-S), a single clinician-rated question scored on a 7-point Likert scale, evaluated the severity of the participant’s anxiety at each session. The CGI-S ranges from 0 (No illness) to 6 (Extremely Severe symptoms – completely nonfunctional). Therapists made ratings based on interval review between sessions, a discussion of therapy homework, and observations of in-session child and parent behavior.

2.4. Analysis Plan

Piecewise multilevel modeling evaluated rates of change during different CBT phases (Singer & Willett, 2003), following Peris et al. (2015). Sample sizes of 50 or more with at least 5 repeated measurements have been recommended for multilevel models (Maas and Hox, 2005). In this study, 148 participants were included with an average of 13 CGI-S ratings per participant.

The “unconditional means model” included the fixed effect of the intercept and random effects of the intercept and residual. The first nested model included the fixed and random effects of session number. Subsequent models included dummy-coded site variables, CBT condition, and a condition X time interaction. “Post-cognitive” and “post-exposure” variables were included in their own nested models to analyze symptom acceleration/deceleration during each phase. These variables were individualized for each child, with “0” coded for each session until the introduction of each skill, at which point the variable increased by one for each subsequent session. Exposure sessions were included in the post-cognitive variable (i.e., they included all sessions after the first cognitive sessions), though including both post-cognitive and post-exposure sessions in the final model allowed for an estimation of the rate of change between the first cognitive and the first exposure session, as well as the rate of change during the exposure phase. Condition X post-cognitive and condition X post-exposure interactions also evaluated whether effects varied by treatment type. Age was also investigated as a moderator.

Cognitive therapy began at session 3 for 97% of participants in adapted CBT (n=70/72) and at session 6 for all participants in standard CBT (n=66/66). Exposure began at session 6 for 98% of participants in adapted CBT (n=71/72) and at session 10 in standard CBT (n=65/66). These modal estimates were used for the ten participants with missing session summary data.

One-hundred twenty-six youth (85%) completed all 16 sessions of treatment, and sixteen (11%) completed at least one session but did not complete the full protocol. Intent-to-treat samples were used. Regarding missing data, 10.6% of CGI-S ratings were missing for attended sessions. A full information maximum likelihood approach was used for missing data.

3. Results

3.1. Demographic characteristics

See Table 1 for demographic information and Wood et al. (2020) for a full CONSORT diagram.

Table 1.

Demographics

	Participants N/Total N (%)
Characteristic	Adapted CBT	Standard CBT
Female gender	21/75(28)	13/72(18)
Latinx ethnicity	12/63(19)	15/54 (28)
Race
African American/African	7/75 (9)	2/71 (3)
Asian/Pacific Islander	6/75 (8)	3/71 (4)
White	46/75 (61)	48/71(68)
Native American or Alaskan	2/75 (3)	1/71 (1)
Multiracial	2/75 (3)	2/71 (3)
Total household income <$40,000	15/72 (21)	13/71(18)
Father’s education
≤High school diploma	14/72 (19)	12/69(17)
≥4-y College degree	40/72 (56)	40/69 (60)
Mother’s education
≤High school diploma	6/74 (8)	5/70 (7)
≥4-y College degree	47/74 (64)	50/70 (71)
Parents currently married	58/75 (77)	52/71 (73)
Autism Diagnostic Observation Schedule-2 total score, M (SD)	12.92 (3.88)	13.01 (4.09)
Estimated Full Scale Intelligence Quotient	102.79 (14.56)	101.64 (15.67)

Note: CBT=cognitive behavioral therapy

3.2. Impact of cognitive and exposure therapy on anxiety severity

Please see Supplemental Table 1 for summary of all model parameters. The final model is summarized in Table 2. Adding time (i.e., session number) in the first nested model improved fit and showed significant fixed and random effects, with decreasing CGI-S scores associated with progression in treatment. Adding two dummy-coded site variables, treatment condition, and the condition X time interaction, also significantly improved model fit. The fixed effects of time and Site 1 were significant in this model, with CGI-S scores being lower at Site 1.

Table 2.

Model predicting Clinical Global Impressions-Severity

Fixed Effects B [95% CI Lower, Upper]
Intercept	4.30*** [3.89, 4.71]
Time	−.086*** [−.13, −.046]
Site 1 a	−.65*** [−.92, −.38]
Site 2 a	−.11 [−.39, .17]
Condition	−.032 [−.25, .19]
Time X Condition	.0059 [−.018, .029]
Post-Cognitive	.066** [.020, .11]
Post-Exposure	−.11*** [−.13, −.071]
Random Effects B [95% CI Lower, Upper]
Residual	.26*** [.24, .28]
Intercept	.31*** [.23, .41]
Time	.0022*** [.0014, .0033]
Post-Cognitive	-- b
Post-Exposure	.0024* [.00089, .0063]
Fit Statistics
−2LL	3352.22
AIC	3377.22
BIC	3449.10

^*p<.05

^**p<.01

^***p<.001

^aSite variables were dummy-coded against the other two sites

^bThe random effect of this variable was also included in this model but could not be estimated due to shared variance with the random effect of the linear time and post-exposure variables

The model including the “post-cognitive’ variable significantly improved the fit of the model, with this variable showing a significant negative effect, suggesting that symptoms decreased at a more rapid rate following the introduction of cognitive therapy.

Adding post-exposure improved model fit. When controlling for post-exposure, the post-cognitive variable can be interpreted as change between the first cognitive session and the first exposure session (see Figure 1). This model showed a significant positive relationship between the onset of cognitive therapy and CGI-S, indicating a significant deceleration in symptom improvement, and a significant negative relationship between the onset of exposure and CGI-S, suggesting that symptoms improved more rapidly during the exposure portion. Thus, the significant negative effect of “post-cognitive” observed in the previous model was carried by the exposure phase. During the exposure phase of CBT, symptoms reduced 0.11 CGI-S points per session; during the cognitive phase, they reduced by an 0.02 CGI-S points per session.

Figure 1.

Clinical severity across different phases of therapy

Note: CBT=cognitive behavioral therapy; CGI-S= Clinical Global Impressions Scale-Severity; Cog=cognitive therapy; Exp=exposure therapy

Moderations investigated whether the potency of cognitive or exposure therapy depending on age or treatment condition. Adding these interactions did not improve model fit, and accordingly, none of the included variables were significantly associated with CGI-S.

4. Discussion

Do specific components of CBT contribute differentially to progress during therapy for autistic youth with anxiety disorders? Results suggest that the exposure therapy portion of CBT led to more rapid symptom improvement compared to earlier phases of treatment. This provides further evidence for the central role of exposure in CBT for youth with anxiety disorders or OCD and extends this pattern to autistic youth. In contrast, the cognitive portion led to a deceleration in symptom improvement compared to introductory sessions. This finding held for standard CBT and for CBT adapted and personalized for autistic youth, which included different amounts of pre-exposure preparatory sessions (9 vs. 5 sessions, on average), as well as different session lengths (90 vs. 60 minutes) and degrees of parent involvement. Thus, exposure was found to be effective regardless of the number of preparatory sessions, the duration of therapy sessions, or the degree of parental participation (though it is important to note parents were involved in both therapy protocols but to a higher degree in adapted CBT).

Unlike findings among nonautistic youth (Peris et al., 2015), in which cognitive skills led to an acceleration of improvement, beginning cognitive therapy was associated with a significant deceleration in improvement for autistic youth. The accelerated rates of improvement that occurred during exposure sessions may suggest that practicing these coping skills in vivo is required for them to be useful for autistic youth, as they were encouraged to revisit and implement their coping plan during exposure exercises. Cognitive techniques were developmentally tailored for school-age youth in the CBT protocols tested in this trial, though still may have required relatively advanced verbal and cognitive flexibility that would be difficult to internalize without explicitly implementing them during exposure.

The initial portion of therapy including psychoeducation (as well as relaxation in standard CBT) was also associated with significant reductions in symptom severity. These findings suggest that the opening period of CBT is also beneficial; providing families with a more concrete framework for addressing anxiety or OCD and building a supportive therapeutic relationship during early sessions also likely contributed to improvement (Karver et al., 2018). The deceleration effect during the cognitive phase and subsequent acceleration of improvement during exposure was apparent regardless of whether relaxation was used in initial sessions, suggesting that inclusion of relaxation did not change the effectiveness of later therapy components.

5.1. Clinical implications

Findings suggest that therapists should emphasize exposure when treating anxiety and OCD in autistic youth and move through the pre-exposure coping phase efficiently. The lack of improvement observed during the cognitive phase of CBT may suggest that autistic youth are better served if they “learn by doing.” For nonautistic youth, non-exposure CBT techniques are used more often (Reid et al., 2018). It is likely this pattern persists in autistic youth, and this study underscores the potential detriment of this trend. Exposure-based CBT may also be a more scalable option, as it may be more straightforward to learn this streamlined approach. This study also highlights the need personalize CBT to the individual; for example, it may be that coping skills should only be taught to youth who demonstrate engagement with this portion.

5.2. Limitations

Several limitations should be noted. First, therapist expectancies may have biased anxiety severity ratings; multi-informant assessments as well as multi-item rating scales conducted by independent evaluators would have yielded more reliable results, though unfortunately when the study was started it was determined that it would not be feasible to have extensive weekly assessments (this week-to-week analysis was not planned a priori). Another limitation is that this study cannot provide conclusions about exposure or cognitive therapy as standalone approaches, as they occurred in a sequenced order in both conditions. Thus, this trial does not provide conclusive evidence about the superiority of cognitive or exposure therapies. The sample consisted of primarily White, male children with well-educated parents, limiting our ability to generalize. There were also more exposure than cognitive sessions, and thus estimates of slopes during the exposure phase could be made more reliably.

5.3. Future directions

Dismantling studies have the potential to build on current findings by directly comparing the different components of CBT, though these studies have not been conducted to date. Developing more streamlined treatment packages that emphasize the central active ingredients in CBT have the potential to improve outcomes, make treatment more efficient, and increase the uptake of evidence-based practices in routine care. Future prospectively designed studies should investigate the mechanisms of exposure in this population as well as moderators of treatment efficacy for exposure or cognitive techniques (e.g., social communication deficits).

Public health significance statement:

Clinically significant anxiety is highly prevalent and causes substantial functional impairment in autistic youth. This study identified an acceleration of improvement during the exposure therapy phase of cognitive behavioral therapy and minimal change during the cognitive portion. This finding was the case in a protocol that introduced exposure after the fifth session and a protocol that introduced exposure after the ninth session, suggesting that introducing exposure earlier in treatment has the potential to enhance the efficiency and efficacy of treatment.

Appendix. Data Transparency statement:

The data reported in this manuscript have been previously published. The primary outcomes from the randomized trial (published) presented mid- and posttreatment time X condition comparisons on primary and secondary outcome measures, including the Pediatric Anxiety Rating Scale, the Child Behavior Checklist Anxious and Depressed and Internalizing scales, the Social Responsiveness Scale-2, and the Child Anxiety Impact Scale. Another manuscript evaluated family accommodation of anxiety in autistic youth (published), while another investigated hoarding behaviors in the sample (published). There are also currently three manuscripts under review that used these data; one evaluated moderators of response to adapted vs. standard CBT, one investigated therapeutic alliance during CBT, and one that evaluated suicidal ideation and externalizing behaviors. None of these manuscripts evaluated session-by-session rates of change during different phases of treatment, as was presented in this study.