Advertisement

Journal of Autism and Developmental Disorders

, Volume 47, Issue 12, pp 3872–3882 | Cite as

Long-Term Effects of CBT on Social Impairment in Adolescents with ASD

  • Brenna B. Maddox
  • Yasuo Miyazaki
  • Susan W. White
S.I. : Anxiety in Autism Spectrum Disorders

Abstract

Anxiety interventions involving social skills training and CBT for youth with ASD have shown promise, but few studies have examined the effects on social functioning or the maintenance of treatment gains. This study evaluated change in social skills during a randomized controlled trial of CBT and during the 1-year follow-up for 25 adolescents with ASD and anxiety. We examined the effect of pretreatment social anxiety and loneliness on treatment response. Social impairment improved during treatment and continued to improve through the 3-month follow-up. Although adolescents with higher social anxiety had greater pretreatment social impairment, they showed steeper improvement in social skills during treatment. Loneliness was not a significant predictor of change during treatment. CBT targeting social skills and anxiety can lead to long-term improvements in social functioning.

Keywords

Autism spectrum disorder Anxiety Adolescents CBT Social skills 

Introduction

Anxiety disorders are among the most common psychiatric comorbidities seen in youth with Autism Spectrum Disorder (ASD), affecting approximately 40 % (van Steensel et al. 2011). Anxiety problems may be particularly debilitating in the context of ASD by further exacerbating core ASD symptoms, such as impaired social skills, and by amplifying feelings of loneliness (e.g., Sukhodolsky et al. 2008; White and Roberson-Nay 2009). The treatment of anxiety with cognitive-behavioral therapy (CBT) in typically developing youth has strong empirical support (Seligman and Ollendick 2011), and several recent studies have supported the efficacy of CBT, modified for ASD, in youth with ASD and anxiety (e.g., Reaven et al. 2012; Storch et al. 2013). One CBT modification for youth with ASD is emphasis on targeting social impairment, a core feature of ASD (e.g., White et al. 2013; Wood et al. 2009). Overall, the immediate outcomes of CBT for anxiety in youth with ASD are positive, with CBT demonstrating strong effects for the reduction of anxiety, relative to waitlist and treatment as usual controls (e.g., Storch et al. 2015; Wood et al. 2009).

Although these results are promising, few studies have examined the effects on social functioning or the long-term maintenance of treatment gains. Extant follow-up studies have focused on the long-term outcome of anxiety symptoms (e.g., Selles et al. 2015; White et al. 2015), without examination of the maintenance of social skill improvement. In addition, limited research has investigated individual predictors of treatment response, an important consideration given the significant variability in clinical presentation of youth with ASD (Kanne et al. 2011). Understanding the predictors of treatment maintenance or relapse could shed light on how to best modify a CBT program to improve long-term outcomes. Social anxiety and loneliness are two possible predictors of response to social skills training and CBT.

Social anxiety can amplify the social problems characteristic of ASD (White et al. 2014). For example, heightened arousal and fear of negative evaluation in social situations may lead to inaccurate processing and interpretation of social cues, avoidance of social interactions, and reduced opportunities to practice social skills with others. Research suggests that compared to other anxiety disorders, social anxiety disorder (SAD) may be more impairing to social functioning in children with ASD (Chang et al. 2012). Thus, it is likely that high levels of social anxiety could impede treatment targeting social skills.

One potential consequence of poor or limited social interactions is loneliness (Bauminger et al. 2003). Indeed, children and adolescents with ASD tend to report more loneliness than do their typically developing peers (e.g., Bauminger and Kasari 2000; Lasgaard et al. 2010). Experiencing the complex emotion of loneliness likely requires at least a minimal level of understanding one’s own social standing, in order to be aware of the discrepancy between actual and desired social status. Thus, self-reported loneliness may stem from insight into the consequences of social rejection and isolation and, as such, may reflect awareness of one’s social impairments (Bauminger and Kasari 2000). Awareness of one’s social impairments and loneliness may be powerful motivating factors to initiate or strengthen social relationships, given the finding that many youth with ASD understand that a close friend could protect them from loneliness and other negative social consequences (Bauminger et al. 2003).

Given that social impairment is a core symptom of ASD and currently targeted in several modified CBT programs, careful evaluation of the long-term stability of social skill improvements following CBT is warranted. The purpose of this study was to examine the degree to which intervention affected social impairment 1 year after treatment, and secondarily determine if pretreatment social anxiety and loneliness predicted treatment response or 1-year posttreatment social impairment. Primary outcome data from this treatment program demonstrated a significant improvement in social impairment in the treatment group from pre- to posttreatment (White et al. 2013). We expected that these gains would be largely maintained during the follow-up period. We also expected that greater pretreatment social anxiety would predict poorer treatment response, given that social anxiety can amplify social difficulties, whereas greater pretreatment loneliness would predict greater improvement in social competence, given that self-reported loneliness may signal increased intrinsic motivation to improve social skills and develop friendships.

Method

Participants

Data for the current study were drawn from an Institutional Review Board-approved, randomized waitlist-controlled trial of the Multimodal Anxiety and Social Skills Intervention (MASSI; White et al. 2013), a CBT program designed to concurrently target anxiety and social impairment in adolescents with ASD. The study was conducted at a university-affiliated clinic specializing in ASD. For inclusion in the treatment trial, participants were between 12 and 17 years; met diagnostic criteria for ASD, as determined by a clinical assessment that included the Autism Diagnostic Interview—Revised (Lord et al. 1994) and the Autism Diagnostic Observation Schedule (Lord et al. 2002); had an anxiety disorder diagnosis of SAD, generalized anxiety disorder, specific phobia, or separation anxiety disorder, as determined by the Anxiety Disorders Interview Schedule for Children/Parents (Silverman and Albano 1996); and had a verbal IQ of 70 or above, as determined by the Wechsler Abbreviated Scale of Intelligence (Wechsler 1999).

A total of 30 eligible adolescents were enrolled in the program (Fig. 1). Fifteen participants were randomized to immediately begin the treatment program, and 15 participants were randomized to a 14-week waitlist condition. After the waitlist period, the waitlist group participants began the intervention. Five participants were excluded from the present analyses because they did not complete treatment after the waitlist period. Assessments were completed at five time points: (1) pretreatment, (2) midpoint, (3) immediately posttreatment, (4) 3 months posttreatment (i.e., early follow-up), and (5) 1-year posttreatment (i.e., late follow-up). Participants from the treatment and waitlist groups were combined for the present analyses, in order to evaluate change during and following active intervention with the maximum sample size. Demographic information for the final sample is presented in Table 1.
Fig. 1

CONSORT Flow chart of participants through follow-up from original randomized controlled trial. Note For the current study, all 15 participants allocated to treatment were included, along with the 10 waitlist group participants who completed the full treatment program and completed at least one post-treatment evaluation

Table 1

Participant pretreatment characteristics (n = 25)

 

M

SD

Minimum

Maximum

Age (in years)

14.42

1.55

12

17

VIQ

98.32

15.18

73

126

Social Phobia CSR

4.96

1.40

3

7

Social loneliness

35.80

12.60

15

61

 

n

% Of sample

Male

19

76

Race/ethnicity

 Asian/Asian American

1

4

 African American

1

4

 Caucasian/European American

21

84

 Bi-/multi-racial

2

8

Primary anxiety disorder diagnosis

 Social Phobia

17

68

 Generalized anxiety disorder

6

24

 Specific Phobia

2

8

VIQ verbal IQ from the Wechsler Abbreviated Scale of Intelligence; CSR Clinical Severity Rating from the Anxiety Disorders Interview Schedule for Children/Parents

Treatment Program Overview

The manualized, modular-based treatment program included 12–13 individual CBT sessions, seven group meetings to practice social skills, and parent education and involvement. Individual sessions lasted approximately 60–70 min, with the parents joining the end of each session for approximately 15 min. During this period, the adolescent and therapist summarized the session for the parent and explained the homework assignment. Parents served as coaches for between-session exposure exercises and other homework activities. A case conceptualization was developed after the third individual session, with subsequent treatment modules selected to address the participant-specific social skill problems and anxiety symptoms. The group sessions, led by two co-therapists and attended by the adolescent participants, were 75 min in length and covered broad social skills (e.g., talking to peers, following a conversation, entering a group). Although a full description of the treatment manual and ASD-specific adaptations is beyond the scope of this paper (see White et al. 2010 for a complete description), the content was based on the principles of CBT and applied behavior analysis, with modifications such as the incorporation of visual aids, increased parental involvement, and consideration of the adolescent’s special interests.

The principal investigator and four clinical psychology doctoral students served as study therapists. All student therapists were trained and supervised by the principal investigator, a licensed clinical psychologist. At the conclusion of each session, therapists rated their fidelity to the treatment manual on a checklist reflecting key session elements. All therapy sessions were video-recorded, and 40 % (14 of 35) of group and 14 % (25 of 180) of individual therapy sessions were independently coded for fidelity by trained coders. Additional information about therapist training and fidelity monitoring can be found in previous study publications (White et al. 2010, 2013).

Measures

Anxiety Disorders Interview Schedule for Children/Parents (ADIS-C/P; Silverman and Albano 1996)

The ADIS-C/P is a clinician-administered interview of psychiatric disorders in childhood. The ADIS-C/P has empirical support as a reliable and valid tool for cognitively-able youth on the spectrum, with demonstrated inter-rater reliability (.77–1.00; Ung et al. 2014), sensitivity to change (White et al. 2013; Wood et al. 2009), and convergent and divergent validity in youth with ASD seeking anxiety treatment (Renno and Wood 2013). For the current study, the clinician conducted the ADIS-C/P jointly with the parent and adolescent. The clinician then assigned an overall rating of anxiety severity (Clinical Severity Rating; CSR) for each anxiety disorder in the interview ranging from 0 to 8, with higher scores indicating greater severity. CSRs are based on frequency and intensity of endorsed symptoms, as well as their associated impairment. CSRs of four or higher exceed diagnostic threshold, while a CSR of three is considered sub-threshold. The current study relied on the CSR for the Social Phobia module as a measure of social anxiety. All interviewers in the current study were trained to reliability on the ADIS-C/P and were knowledgeable on specific distinctions between ASD and anxiety disorders. One strength of this semi-structured interview approach is that information from both the parent and adolescent was collected, meaning that multiple reports are reflected in the final CSR.

Loneliness Questionnaire (Bauminger et al. 2003)

The Loneliness Questionnaire is a 30-item self-report measure used to assess loneliness in youth with ASD. The questionnaire distinguishes between emotional loneliness (i.e., feelings of isolation and lack of affective bonding) and social loneliness (i.e., the child’s perceived lack of social involvement with peers). Nine items comprise the emotional loneliness subscale (e.g., “I feel lonely at school”), and 13 questions comprise the social loneliness subscale (e.g., “I don’t have anyone to play with”). Eight questions are “distracter” or “filler” items (e.g., “I love to read”). Participants rated each item from 1 (“not true”) to 5 (“true”) during the pretreatment evaluation. For the current study, the social loneliness subscale was used, with higher scores indicating greater loneliness. In this sample, alpha was .91 at baseline.

Social Responsiveness Scale (SRS; Constantino and Gruber 2005)

The SRS is a well-established parent-report measure of ASD-related social impairment. It contains 65 items, each scored from 1 (“not true”) to 4 (“almost always true”), to assess social awareness, social communication abilities, and other aspects of social functioning. Higher scores on the SRS reflect greater severity of social impairment. The SRS provides a total T-score (standardized by gender and age), with a T-score of 59 or less in the normal range, 60–65 in the mild range, 66–75 in the moderate range, and 76 or greater in the severe range of social impairment. In this sample, internal consistency was high (α = .94).

Data Analyses

Initially, paired samples t tests were conducted to examine change in social impairment from pretreatment to posttreatment and follow-up. Then, hierarchical linear modeling (HLM) was employed to assess (1) the stability in social skill improvement during treatment and follow-up, and (2) the influence of pretreatment social anxiety and loneliness on trajectories of social impairment. Time variables (created from the duration of time passed from entry to each subsequent measurement occasion) were entered as the level-1 predictors to represent the within-individual change trajectory over time. Social anxiety and loneliness (i.e., ADIS-C/P Social Phobia CSR and Loneliness Questionnaire score) were entered as the level-2 variables to determine whether they could explain the variation in growth trajectories.

It should be noted that, unlike the repeated measures analysis of variance (ANOVA), which is also used for analyzing longitudinal data, HLM allows missing observations at level 1 (Howell 2013). Thus, no participants were eliminated, despite missing observations at certain measurement occasions. HLM makes an efficient estimation by utilizing all the available data, and the estimates are valid (i.e., asymptotically unbiased) when missingness occurs at random (MAR; Little and Schenker 1995). MAR means that missingness may be nonrandom, but the association between the probability of missingness and the missing value is explainable by the observed data. In this sense, the MAR assumption is far less restrictive than the assumption that data are missing completely at random (MCAR; Little and Rubin 1987). Further, even when MAR is not satisfied, the robustness of the results is maximized when all available data are used in the analysis (Schafer 1997).

Analyses were conducted using the HLM v.7 software package (Raudenbush et al. 2011) with the full maximum likelihood estimation method. Visual examination of the SRS data revealed three separate time periods of interest, given that SRS scores, on average, declined during the treatment period, continued to decline through the 3-month follow-up period, and then increased between the 3-month and 1-year assessments (Table 2). Therefore, a piecewise linear growth model (Raudenbush and Bryk 2002, pp. 178–181) was utilized to estimate three separate slopes for the treatment, early follow-up (i.e., through 3-month evaluation), and late follow-up (i.e., through 1-year evaluation) periods (see Table 3) by introducing three Time variables, referred to as Time 1, Time 2, and Time 3 variables, respectively. The Time 1 variable reflected the time period during the treatment (approximately 15 weeks). The average number of days between pretreatment and midpoint (64.85 days) was divided by the average number of days between pretreatment and posttreatment (104.80 days), which was used as the unit length of time. Similarly, coding of the early follow-up (i.e., Time 2 variable) and late follow-up (i.e., Time 3 variable) periods were calculated by dividing the average number of days between posttreatment and the 3-month follow-up (104.37 days) and the average number of days between the 3-month follow-up and the 1-year follow-up (270.62 days) by the average number of days between pretreatment and posttreatment, respectively.
Table 2

Social Responsiveness Scale (SRS) descriptive statistics

Time point

n (% Missing)

M

SD

Minimum

Maximum

SRS

Pretreatment

25 (0 %)

86.12

13.14

66

115

Midpoint

13 (48 %)

79.85

10.49

62

97

Posttreatment

15 (40 %)

74.33

12.63

52

97

3-month follow-up

19 (24 %)

68.84

12.79

45

94

1-year follow-up

15 (40 %)

77.60

14.88

55

98

All participants had SRS data from at least two time points. Four participants (16 %) had missing data for both the 3-month and 1-year follow-ups

Table 3

Coding of time variables and expected value of the outcome variable

Time Point

Pretreatment

Midpoint

Posttreatment

3-Month follow-up

1-Year follow-up

Time 1

0

.619

1

1

1

Time 2

0

0

0

0.996

0.996

Time 3

0

0

0

0

2.582

Expected value

π 0i

π 0i  + 0.619π 1i

π 0i  + π 1i

π 0i  + π 1i  + 0.996π 2i

π 0i  + π 1i  + 0.996π 2i  + 2.582π 3i

π 0i , expected SRS value for participant i at the beginning of treatment; π 1i , expected change in SRS score for participant i at the end of the treatment; π 2i , expected change in SRS score for participant i for the unit change in time (about 3 months) during the early follow-up period; π 3i , expected change in SRS score for participant i for the unit change in time (about 3 months) during the late follow-up period

The unconditional growth model with no level-2 predictors was used to examine the level of variability in the intercept and slopes, before adding the level-2 predictors. Specifically, at level 1, the unconditional piecewise linear growth model was formulated as follows:

Level 1 model.
$$ SRS_{ti} = \pi_{0i} + \pi_{1i}^{*}\left( {Time\,1_{ti} } \right) \, + \pi_{2i} ^ {*}\left( {Time\,2_{ti} } \right) \, + \pi_{3i}^ {*}\left( {Time\,3_{ti} } \right) \, + e_{ti} $$
(1)
where SRS ti is the SRS score for participant i at time t, t = 1,…, 5; i = 1,…, 25; Time 1 ti , Time 2 ti , and Time 3 ti are the coded time variables as defined in Table 3, to represent the piecewise regression; because of the coding scheme we introduced, π 0i is the expected SRS score of participant i at the beginning of the treatment; π 1i is the expected rate of change per unit length of time (i.e., about 3 months) for participant i during the treatment period; π 2i is the expected rate of change for participant i during the early follow-up period; π 3 is the expected rate of change for participant i during the late follow-up period; and e ti is the random within-subject error, with e ti ’s assumed to be mutually independent and identically distributed (i.i.d.) as a normal distribution with mean of zero and variance σ2, that is \( e_{ti} \mathop \sim\limits^{i.i.d.} N(0,\sigma^{2} ) \).

Equation 1 characterizes each participant’s trajectory for the period of the treatment study by four parameters, π 0i , π 1i , π 2i , and π 3i . At level 2, these individual parameters become the outcome variables. For the unconditional growth model, a simple level-2 model that does not include any predictors was formulated as follows:

Level 2 model.
$$ \begin{aligned} \pi_{0i} = \beta_{00} + r_{0i} , \hfill \\ \pi_{1i} = \beta_{10} + r_{1i} , \hfill \\ \pi_{2i} = \beta_{20} + r_{2i} , \hfill \\ \pi_{3i} = \beta_{30} + r_{3i} , \hfill \\ \end{aligned} $$
(2)
where β 00 is the mean SRS score at the beginning of the intervention (i.e., initial status), β 10 is the mean rate of change during the intervention period (i.e., immediate treatment effect), β 20 is the mean rate of change during the early follow-up period measured, and β 30 is the mean rate of change during the late follow-up period; r 0i is the random effect of person i on SRS score at the pretreatment time; and r 1i , r 2i , and r 3i are the random effects of person i on the slopes of the intervention period, early follow-up period, and late follow-up period, respectively. The random effects \( r_{i}^{\prime } \) = (r 0i , r 1i , r 2i , r 3i ) are assumed to be mutually independent and identically distributed as multivariate normal with zero means and variance–covariance matrix T, where T is a 4 by 4 symmetric matrix. That is, \( r_{i} \mathop \sim\limits^{i.i.d.} N(0,T) \), where
$$ T = \left( {\begin{array}{*{20}c} {\tau_{00} } & {\tau_{01} } & {\tau_{02} } & {\tau_{03} } \\ {\tau_{10} } & {\tau_{11} } & {\tau_{12} } & {\tau_{13} } \\ {\tau_{20} } & {\tau_{21} } & {\tau_{22} } & {\tau_{23} } \\ {\tau_{30} } & {\tau_{31} } & {\tau_{32} } & {\tau_{33} } \\ \end{array} } \right) $$
(3)

After fitting the above unconditional growth model, two predictors (i.e., social anxiety and loneliness levels at pretreatment), grand mean centered, were included in the level-2 model. Since all the level-2 random effects were found to be statistically significant (Table 3), both predictors were included with all the intercept and slopes dependent variables at level 2, which was referred to as the full or saturated model. Then, backward elimination was used to determine the final model, with non-significant level-2 predictors sequentially deleted (starting with the predictor with the largest p value for the slopes). Based on a series of deviance tests, the final model, presented below, was determined.

Results

Descriptive statistics for the outcome variable (i.e., SRS total score) are presented in Table 2, which shows that social impairment decreased during the treatment and early follow-up periods and increased somewhat between the 3-month follow-up and 1-year follow-up time points. Compared to pretreatment, paired t tests indicated that SRS scores were significantly lower at midpoint, t(12) = 3.84, p = .002; posttreatment, t(14) = 3.60, p = .003; 3-month follow-up, t(18) = 7.60, p < .001; and 1-year follow-up, t(14) = 3.64, p = .003. The effect sizes, indexed by Cohen’s d (1988), for each score difference were all quite large: 1.055, .930, 1.744, and .939, respectively. All participants had data for the level-2 variables (ADIS-C/P Social Phobia CSR: M = 4.96, SD = 1.40; Loneliness Questionnaire: M = 35.80, SD = 12.60), and the level-2 variables were not significantly correlated with each other (r = −.02, p = .95).

To test the assumption that data were missing at random, we compared participant demographic characteristics, level-2 variable scores, and pretreatment SRS scores for the 15 participants with 1-year follow-up data to the 10 participants who did not provide 1-year follow-up data. The two groups did not differ on any of these variables (all ps > .18).

In the unconditional model, there was a significant amount of variability in the intercept, π 0i [estimated variance = 150.46, χ 2(7) = 89.70, p < .001]; Time 1 slope, π 1i [estimated variance = 181.95, χ 2(7) = 43.53, p < .001]; Time 2 slope π 2i [estimated variance = 148.51, χ 2(7) = 51.91, p < .001]; and Time 3 slope π 3i [estimated variance = 15.04, χ 2(7) = 32.61, p < .001] (see the first block in Table 4). This variability indicated a significant amount of individual differences in each growth parameter, which prompted us to examine whether social anxiety and loneliness could be significant predictors for each element of growth parameters in the full model.
Table 4

Hierarchical linear model results

Variable

Coefficient

Standard error

t

d.f.

p

Unconditional model

Intercept, β 00

86.13

2.60

33.14

24

<.001

 Time 1 slope, β 10

−12.95

3.45

−3.75

24

<.001

 Time 2 slope, β 20

−4.60

3.41

−1.35

24

.190

 Time 3 slope, β 30

3.33

1.16

2.88

24

.008

Full model

Intercept, β 00

86.13

2.05

41.96

22

<.001

 Social anxiety, β 01

5.71

1.50

3.81

22

<.001

 Loneliness, β 02

−0.09

0.17

−0.55

22

.586

Time 1 slope, β 10

−13.12

3.34

−3.93

22

<.001

 Social anxiety, β 11

−4.78

2.58

−1.85

22

.077

 Loneliness, β 12

0.11

0.27

0.39

22

.698

Time 2 Slope, β 20

−4.79

3.29

−1.46

22

.159

 Social anxiety, β 21

3.66

2.58

1.42

22

.169

 Loneliness, β 22

−0.18

0.27

−0.66

22

.518

Time 3 Slope, β 30

4.09

0.83

4.77

22

<.001

 Social anxiety, β 31

0.13

0.70

0.18

22

.857

 Loneliness, β 32

0.26

0.07

3.84

22

<.001

Final model

Intercept, β 00

86.13

2.05

41.96

22

<.001

 Social anxiety, β 01

5.62

1.39

4.03

22

<.001

 Loneliness, β 02

−0.09

0.13

−0.71

22

0.487

Time 1 Slope, β 10

−13.19

3.37

−3.91

23

<.001

 Social anxiety, β 11

−5.19

2.54

−2.05

23

0.052

Time 2 Slope, β 20

−4.69

3.36

−1.39

23

0.177

 Social anxiety, β 21

4.15

2.48

1.68

23

0.107

Time 3 Slope, β 30

4.08

0.82

4.84

23

<.001

 Loneliness, β 32

0.24

0.05

4.42

23

<.001

As mentioned, the full model included social anxiety and loneliness as level-2 predictors, both grand mean centered. In the final model, the intercept, Time 1 slope, and Time 3 slope were significant (see the second block of Table 4). SRS scores decreased (i.e., improved) significantly over the course of treatment, but then increased (i.e., worsened) between the 3-month and 1-month follow-up assessments. It should be noted that a multiparameter test (Raudenbush and Bryk 2002), which is also known as a Wald test, revealed that the increase in SRS score did not cancel out the improvement from treatment [\( \chi_{obs.}^{2} \) (1) = 12.824, p < .001 for the test of H0: β 10 + .996 β 20 + 2.582 β 30 = 0]. In other words, participants did not return to pre-treatment levels of social impairment.

Additionally, social anxiety was only a significant predictor of the intercept, although it approached statistical significance at the .05 level with a two-tailed test when predicting the Time 1 and Time 2 slopes. Adolescents with higher social anxiety had higher pretreatment SRS scores (i.e., more social impairment) (\( \hat{\beta }_{01} \) = 5.62, p < .001). The effect of social anxiety on the Time 1 slope was negative (\( \hat{\beta }_{11} \) = −5.19, p = .052), suggesting that adolescents with higher social anxiety had more improvement in social skills during treatment. The effect of social anxiety on the Time 2 slope, though non-significant, was positive (\( \hat{\beta }_{21} \) = 4.15, p = .107), suggesting that adolescents with higher social anxiety experienced greater worsening in social skills during the early follow-up period (i.e., from posttreatment to 3-month follow-up).1 Self-reported loneliness was only a significant predictor of the Time 3 slope, in the positive direction (\( \hat{\beta }_{32} \) = .24, p < .001). On average, adolescents who reported greater pretreatment loneliness had poorer maintenance of social skill improvements during the late follow-up period (i.e., between the 3-month and 1-year follow-up evaluations). The effects of social anxiety and self-reported loneliness on the trajectories of social impairment are graphically depicted in Figs. 2 and 3, respectively.
Fig. 2

Effect of social anxiety on social impairment over time. Note Pretreatment Social Phobia Clinical Severity Rating (CSR) M = 4.96, SD = 1.40. SRS Score = Social Responsiveness Scale total T-score, with scores ≤59 in the normal range, 60–65 in the mild range, 66–75 in the moderate range, and ≥76 in the severe range of social impairment

Fig. 3

Effect of loneliness on social impairment over time. Note Pretreatment Loneliness Questionnaire M = 35.80, SD = 12.60. SRS Score = Social Responsiveness Scale total T-score, with scores ≤59 in the normal range, 60–65 in the mild range, 66–75 in the moderate range, and ≥76 in the severe range of social impairment

Discussion

Previous CBT studies with youth who have ASD and anxiety reported on the long-term outcome of anxiety symptoms (e.g., Selles et al. 2015; White et al. 2015), without investigating the maintenance of social skill improvement. The current findings demonstrate that modified CBT for adolescents with ASD and co-occurring anxiety can produce sustained improvement in parent-rated social impairment, a core feature of ASD. Social impairment improved significantly over the course of active treatment, and was mostly maintained through the 3-month follow-up assessment, which could suggest that participants and their parents actively practiced and generalized the skills learned from treatment. Although there was some worsening of social impairment between the 3-month and 1-year follow-up evaluations, it remained significantly improved at 1-year posttreatment, relative to pretreatment. A similar pattern of long-term reduction in anxiety symptoms was recently reported in this sample (White et al. 2015). These promising findings specific to social skills improvement add to the growing literature supporting CBT for youth with ASD and anxiety (e.g., Storch et al. 2015). Our results are also similar to a long-term follow-up study (i.e., 1–5 years after treatment) of the Program for the Education and Enrichment of Relational Skills (PEERS; Laugeson and Frankel 2010), a parent-assisted social skills group intervention targeting friendship skills, which found that significant posttreatment improvements on the SRS total score maintained over time for 27 adolescents with ASD (Mandelberg et al. 2014).

Results did not support our hypotheses about predictors of treatment response. The relationships between pretreatment social anxiety and loneliness and the effects of treatment over time were found to be complex. Adolescents with higher social anxiety had greater pretreatment social impairment, consistent with a bi-directional relationship between social deficits and social anxiety (White et al. 2014) and previous research suggesting that social anxiety is particularly detrimental to social functioning in youth with ASD (Chang et al. 2012). However, adolescents with higher pretreatment social anxiety tended to show a steeper rate of improvement in social skills during treatment. Perhaps certain components or correlates of social anxiety, such as social motivation, insight into personal social difficulties, or social awareness (Kuusikko et al. 2008; White et al. 2014; Williamson et al. 2008) can bolster treatment response by increasing adolescents’ engagement in the intervention. It is also possible that as these youth received treatment targeting both anxiety and social skills, their social anxiety decreased, which led to less avoidance of social activities and more opportunities to practice social skills (White et al. 2010). Yet another possibility is that those youth with better insight into their thoughts and feelings related to peer rejection and social difficulty were more likely to both receive higher severity ratings from the interview (owing to insight and verbalization of such) and reap greater benefit from the intervention.

However, the bolstering effect of pretreatment social anxiety on improvement of social functioning did not last once the treatment was terminated. During the early follow-up period (i.e., between endpoint and 3-month follow-up), adolescents with higher pretreatment social anxiety tended to show a worsening in social skills, compared to those with less pretreatment social anxiety. This pattern is consistent with research suggesting that social anxiety contributes to social disability in youth with ASD and may prevent the successful implementation of well-learned skills (e.g., Chang et al. 2012; Myles et al. 2001). It is possible that during the active treatment phase, the more socially anxious participants were more likely to engage in the treatment for several reasons, such as fears of the therapist being disappointed or upset with them (related to fears of negative evaluation by others, a core feature of SAD; American Psychiatric Association 2013), or simply doing the exposures involved in the CBT. Once the active treatment phase ended, perhaps these youth engaged in social activities less often, so they experienced a decline in parent-observed social skills. This finding may highlight the importance of emphasizing relapse prevention material during treatment and the utility of incorporating booster sessions during the posttreatment period. For youth without ASD who struggle with anxiety, CBT programs with booster sessions are more effective than CBT programs without booster sessions (Gearing et al. 2013). In addition, more intensive coaching of parents may be needed on how to promote social engagement, in order to sustain treatment benefits (e.g., Laugeson et al. 2009; Mandelberg et al. 2014).

Self-reported loneliness, which may reflect a greater degree of insight into an individual’s personal difficulties related to ASD-related social impairment, was not a significant predictor of change in social skills during the treatment period. However, contrary to study hypotheses, adolescents who reported greater pretreatment loneliness had poorer maintenance of social skill improvements between the 3-month and 1-year follow-up evaluations. The reasons for this finding are unclear and require replication with longitudinal research and additional measures of loneliness. Speculatively, it is possible that these youth felt less supported (Lasgaard et al. 2010) or more hopeless with regards to developing more friendships, given that stable loneliness in adolescence is associated with increased depressive symptoms (Ladd and Ettekal 2013).

Finally, it should be noted that despite the relatively small sample size available in the present study, the t tests (α = .05, two-tailed) for detecting the treatment effect (from beginning of the intervention to the end of treatment), the continuing effect (through the 3-month follow-up), and the sustaining effect (through the 1-year follow-up) exhibited excellent power with .917, >.999, and .939, respectively. These extremely high power estimates resulted from the very large effect sizes (Cohen’s d) of .930, 1.744, and .939. These large effect sizes suggest that MASSI is not only an effective, but also a long-lasting, intervention program for improving social impairment in adolescents with ASD who do not have intellectual disability. Another strength of this study is the use of multi-method data. Indeed, the three variables of interest relied on clinical interview, parent-report, and youth self-report.

Limitations

This study’s primary limitations are the small sample size and missing data at certain time points, given that only 25 participants completed the treatment phase, and even fewer completed all of the assessments. Even though the overall treatment effects exhibited excellent power with this small sample size, the preliminary results for the effects of social anxiety and loneliness, some of which demonstrated slightly larger than the .05 or .10 significance level, should be interpreted with caution. To be able to draw firm conclusions about the long-term gains associated with MASSI, or CBT programs for ASD more generally, replication of these findings with larger samples is needed (e.g., n = 51 for testing β 11 and n = 76 for β 21 with power of .8). In addition, the sample was cognitively unimpaired, meaning these results may not generalize to youth with ASD who have intellectual impairment. Another limitation is that social functioning treatment outcome was assessed using a single parent-report measure (SRS). Although the SRS is a well-established unidimensional measure of social impairment, several items assess repetitive, restricted behaviors and interests, which could have influenced our findings related to core social functioning. Future studies should include additional informants and methods of social impairment assessment.

Future Directions

Future studies should include larger samples and consider additional factors that may affect variability in treatment response. As evidence builds for the interconnectedness among anxiety, loneliness, and social impairment, clinicians should strongly consider treatment programs with a dual focus on improving anxiety and social impairment for adolescents with ASD. In addition, future research could explore which components or correlates of pretreatment social anxiety (e.g., social motivation, awareness, insight, information processing) bolster treatment response.

Conclusions

Despite these limitations, our results suggest that modified CBT that emphasizes social skill development and anxiety reduction can have a lasting impact on ASD-related social impairment. Long-term outcome is an important consideration when determining the true effectiveness of a treatment program. Although further research and replication is needed, these findings have considerable clinical and potential policy/funding import, given the strong extant research base supporting the efficacy of CBT, and the fairly minimal cost and time investment associated with CBT implementation. Future research must explore within-person factors that moderate both immediate treatment response and sustained response after treatment cessation.

Footnotes

  1. 1.

    The powers for the two t tests with p values slightly greater than either .05 or .10 were significantly lower than .80, a value that is typically recommended in behavioral sciences (.50 for testing H0: β 11 = 0 and .36 for H0: β 21 = 0, respectively). The insufficient power was mainly the consequence of the small sample size for the present study.

Notes

Acknowledgments

This project was funded by the NIMH (K01MH079945; PI: S. White). Portions of these findings were presented at the 2015 Society for Research in Child Development Biennial Meeting.

Author Contributions

BM conceived of the study, participated in its design, performed the statistical analyses, interpreted the data, and drafted the manuscript; YM consulted on the study design, performed the statistical analyses, and consulted on interpretation of results; SW participated in the design of the study, oversaw the collection of the data, assisted with interpretation of the data, and revised the manuscript critically. All authors read and approved the final manuscript.

Compliance with Ethical Standards

Conflict of interest

The authors have no conflicts of interest.

References

  1. American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Washington, DC: Author.CrossRefGoogle Scholar
  2. Bauminger, N., & Kasari, C. (2000). Loneliness and friendship in high-functioning children with autism. Child Development, 71, 447–456. doi: 10.1111/1467-8624.00156.CrossRefPubMedGoogle Scholar
  3. Bauminger, N., Shulman, C., & Agam, G. (2003). Peer interaction and loneliness in high-functioning children with autism. Journal of Autism and Developmental Disorders, 33, 489–507. doi: 10.1023/A:1025827427901.CrossRefPubMedGoogle Scholar
  4. Chang, Y.-C., Quan, J., & Wood, J. J. (2012). Effects of anxiety disorder severity on social functioning in children with autism spectrum disorders. Journal of Developmental and Physical Disabilities, 24, 235–345. doi: 10.1007/s10882-012-9268-2.CrossRefGoogle Scholar
  5. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.Google Scholar
  6. Constantino, J. N., & Gruber, C. P. (2005). Social Responsiveness Scale (SRS). Los Angeles, CA: Western Psychological Services.Google Scholar
  7. Gearing, R. E., Schwalbe, C. S., Lee, R., & Hoagwood, K. E. (2013). The effectiveness of booster sessions in CBT treatment for child and adolescent mood and anxiety disorders. Depression and Anxiety, 30, 800–808. doi: 10.1002/da.22118.CrossRefPubMedGoogle Scholar
  8. Howell, D. C. (2013). Statistical methods for psychology (8th ed.). Belmont, CA: Wadsworth.Google Scholar
  9. Kanne, S. M., Gerber, A. J., Quirmbach, L. M., Sparrow, S. S., Cicchetti, D. V., & Saulnier, C. A. (2011). The role of adaptive behavior in autism spectrum disorders: Implications for functional outcome. Journal of Autism and Developmental Disorders, 41, 1007–1018. doi: 10.1007/s10803-010-1126-4.CrossRefPubMedGoogle Scholar
  10. Kuusikko, S., Pollock-Wurman, R., Jussila, K., Carter, A. S., Mattila, M.-L., Ebeling, H., & Moilanen, I. (2008). Social anxiety in high-functioning children and adolescents with autism and Asperger syndrome. Journal of Autism and Developmental Disorders, 38, 1697–1709. doi: 10.1007/s10803-008-0555-9.CrossRefPubMedGoogle Scholar
  11. Ladd, G. W., & Ettekal, I. (2013). Peer-related loneliness across early to late adolescence: Normative trends, intra-individual trajectories, and links with depressive symptoms. Journal of Adolescence, 36, 1269–1282. doi: 10.1016/j.adolescence.2013.05.004.CrossRefPubMedGoogle Scholar
  12. Lasgaard, M., Nielsen, A., Eriksen, M. E., & Goossens, L. (2010). Loneliness and social support in adolescent boys with autism spectrum disorders. Journal of Autism and Developmental Disorders, 40, 218–226. doi: 10.1007/s10803-009-0851-z.CrossRefPubMedGoogle Scholar
  13. Laugeson, E. A., & Frankel, F. (2010). Social skills for teenagers with developmental and autism spectrum disorders: The PEERS treatment manual. New York, NY: Routledge.Google Scholar
  14. Laugeson, E. A., Frankel, F., Mogil, C., & Dillon, A. R. (2009). Parent-assisted social skills training to improve friendships in teens with autism spectrum disorders. Journal of Autism and Developmental Disorders, 39, 596–606. doi: 10.1007/s10803-008-0664-5.CrossRefPubMedGoogle Scholar
  15. Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing data. New York, NY: Wiley.Google Scholar
  16. Little, R. J. A., & Schenker, N. (1995). Missing data. In G. Arminger, C. C. Clogg, & M. E. Sobel (Eds.), Handbook of statistical modeling for the social sciences (pp. 39–75). New York, NY: Plenum Press.CrossRefGoogle Scholar
  17. Lord, C., Rutter, M., DiLavore, P. C., & Risi, S. (2002). Autism diagnostic observation schedule. Los Angeles, CA: Western Psychological Services.Google Scholar
  18. Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism diagnostic interview-revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders, 24, 659–685. doi: 10.1007/BF02172145.CrossRefPubMedGoogle Scholar
  19. Mandelberg, J., Laugeson, E. A., Cunningham, T. D., Ellingsen, R., Bates, S., & Frankel, F. (2014). Long-term treatment outcomes for parent-assisted social skills training for adolescents with autism spectrum disorders: The UCLA PEERS program. Journal of Mental Health Research in Intellectual Disabilities, 7, 45–73. doi: 10.1080/5864.2012.730600.CrossRefGoogle Scholar
  20. Myles, B. S., Barnhill, G. P., Hagiwara, T., Griswold, D. E., & Simpson, R. L. (2001). A synthesis of studies on the intellectual, academic, social/emotional and sensory characteristics of children and youth with Asperger syndrome. Education and Training in Mental Retardation and Developmental Disabilities, 36, 304–311.Google Scholar
  21. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage.Google Scholar
  22. Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., Congdon, R. T., & du Toit, M. (2011). HLM for Windows (Version 7.01) [Computer software]. Skokie, IL: Scientific Software International, Inc.Google Scholar
  23. Reaven, J., Blakeley-Smith, A., Culhane-Shelburne, K., & Hepburn, S. (2012). Group cognitive behavior therapy for children with high-functioning autism spectrum disorders and anxiety: A randomized trial. Journal of Child Psychology and Psychiatry, 53, 410–419. doi: 10.1111/j.1469-7610.2011.02486.x.CrossRefPubMedPubMedCentralGoogle Scholar
  24. Renno, P., & Wood, J. J. (2013). Discriminant and convergent validity of the anxiety construct in children with autism spectrum disorders. Journal of Autism and Developmental Disorders, 43, 2135–2146. doi: 10.1007/s10803-013-1767-1.CrossRefPubMedPubMedCentralGoogle Scholar
  25. Schafer, J. L. (1997). Analysis of incomplete multivariate data. London: Chapman & Hall.CrossRefGoogle Scholar
  26. Seligman, L. D., & Ollendick, T. H. (2011). Cognitive-behavioral therapy for anxiety disorders in youth. Child and Adolescent Psychiatric Clinics of North America, 20, 217–238. doi: 10.1016/j.chc.2011.01.003.CrossRefPubMedPubMedCentralGoogle Scholar
  27. Selles, R. R., Arnold, E. B., Phares, V., Lewin, A. B., Murphy, T. K., & Storch, E. A. (2015). Cognitive-behavioral therapy for anxiety in youth with an autism spectrum disorder: A follow-up study. Autism, 19, 613–621. doi: 10.1177/1362361314537912.CrossRefPubMedGoogle Scholar
  28. Silverman, W. K., & Albano, A. M. (1996). The anxiety disorders interview schedule for DSM-IV: Child and parent versions. San Antonio, TX: Graywind.Google Scholar
  29. Storch, E. A., Arnold, E. B., Lewin, A. B., Nadeau, J. M., Jones, A. M., De Nadai, A. S., & Murphy, T. K. (2013). The effect of cognitive-behavioral therapy versus treatment as usual for anxiety in children with autism spectrum disorders: A randomized, controlled trial. Journal of the American Academy of Child and Adolescent Psychiatry, 52, 132–142. doi: 10.1016/j.jaac.2012.11.007.CrossRefPubMedGoogle Scholar
  30. Storch, E. A., Lewin, A. B., Collier, A. B., Arnold, E., De Nadai, A. S., Dane, B. F., & Murphy, T. K. (2015). A randomized controlled trial of cognitive-behavioral therapy versus treatment as usual for adolescents with autism spectrum disorders and comorbid anxiety. Depression and Anxiety, 32, 174–181. doi: 10.1002/da.22332.CrossRefPubMedGoogle Scholar
  31. Sukhodolsky, D. G., Scahill, L., Gadow, K. D., Arnold, L. E., Aman, M. G., McDougle, C. J., & Vitiello, B. (2008). Parent-rated anxiety symptoms in children with pervasive developmental disorders: Frequency and association with core autism symptoms and cognitive functioning. Journal of Abnormal Child Psychology, 36, 117–128. doi: 10.1007/s10802-007-9165-9.CrossRefPubMedGoogle Scholar
  32. Ung, D., Arnold, E. B., De Nadai, A. S., Lewin, A. B., Phares, V., Murphy, T. K., & Storch, E. A. (2014). Inter-rater reliability of the anxiety disorders interview schedule for DSM-IV in high-functioning youth with autism spectrum disorder. Journal of Developmental and Physical Disabilities, 26, 53–65. doi: 10.1007/s10882-013-9343-3.CrossRefPubMedGoogle Scholar
  33. van Steensel, F. J. A., Bögels, S. M., & Perrin, S. (2011). Anxiety disorders in children and adolescents with autistic spectrum disorders: A meta-analysis. Clinical Child and Family Psychology Review, 14, 302–317. doi: 10.1007/s10567-011-0097-0.CrossRefPubMedPubMedCentralGoogle Scholar
  34. Wechsler, D. (1999). Manual for the Wechsler abbreviated scale of intelligence. San Antonio, TX: Psychological Corporation.Google Scholar
  35. White, S. W., Albano, A. M., Johnson, C. R., Kasari, C., Ollendick, T., Klin, A., & Scahill, L. (2010). Development of a cognitive-behavioral intervention program to treat anxiety and social deficits in teens with high-functioning autism. Clinical Child and Family Psychology Review, 13, 77–90. doi: 10.1007/s10567-009-0062-3.CrossRefPubMedPubMedCentralGoogle Scholar
  36. White, S. W., Ollendick, T., Albano, A. M., Oswald, D., Johnson, C., Southam-Gerow, M. A., & Scahill, L. (2013). Randomized controlled trial: Multimodal anxiety and social skill intervention for adolescents with autism spectrum disorder. Journal of Autism and Developmental Disorders, 43, 382–394. doi: 10.1007/s10803-012-1577-x.CrossRefPubMedPubMedCentralGoogle Scholar
  37. White, S. W., & Roberson-Nay, R. (2009). Anxiety, social deficits, and loneliness in youth with autism spectrum disorders. Journal of Autism and Developmental Disorders, 39, 1006–1013. doi: 10.1007/s10803-009-0713-8.CrossRefPubMedGoogle Scholar
  38. White, S. W., Schry, A. R., & Kreiser, N. L. (2014). Social worries and difficulties: Autism and/or social anxiety disorder? In T. E. Davis III, S. W. White, & T. H. Ollendick (Eds.), Handbook of autism and anxiety. New York, NY: Springer.Google Scholar
  39. White, S. W., Schry, A. R., Miyazaki, Y., Ollendick, T. H., & Scahill, L. (2015). Effects of verbal ability and severity of autism on anxiety in adolescents with ASD: One-year follow-up after cognitive behavioral therapy. Journal of Clinical Child and Adolescent Psychology, 44, 839–845. doi: 10.1080/15374416.2014.893515.CrossRefPubMedGoogle Scholar
  40. Williamson, S., Craig, J., & Slinger, R. (2008). Exploring the relationship between measures of self-esteem and psychological adjustment among adolescents with Asperger syndrome. Autism, 12, 391–402. doi: 10.1177/1362361308091652.CrossRefPubMedGoogle Scholar
  41. Wood, J. J., Drahota, A., Sze, K., Har, K., Chiu, A., & Langer, D. A. (2009). Cognitive behavioral therapy for anxiety in children with autism spectrum disorders: A randomized, controlled trial. Journal of Child Psychology and Psychiatry, 50, 224–234. doi: 10.1111/j.1469-7610.2008.01948.x.CrossRefPubMedPubMedCentralGoogle Scholar

Copyright information

© Springer Science+Business Media New York 2016

Authors and Affiliations

  • Brenna B. Maddox
    • 1
    • 2
  • Yasuo Miyazaki
    • 3
  • Susan W. White
    • 2
  1. 1.Center for Autism ResearchChildren’s Hospital of PhiladelphiaPhiladelphiaUSA
  2. 2.Department of PsychologyVirginia TechBlacksburgUSA
  3. 3.School of EducationVirginia TechBlacksburgUSA

Personalised recommendations