European Child & Adolescent Psychiatry

, Volume 22, Issue 7, pp 433–442

Social skills improvement in children with high-functioning autism: a pilot randomized controlled trial

Authors

    • Department of Child and Adolescent Psychiatry, Autism Resources Centre, Clinique Peyre-PlantadeCHU Montpellier
    • Research Team 4556 (Epsylon Laboratory)University Montpellier
  • J. Brisot
    • Department of Child and Adolescent Psychiatry, Autism Resources Centre, Clinique Peyre-PlantadeCHU Montpellier
    • Research Team 4556 (Epsylon Laboratory)University Montpellier
  • V. Henry
    • Department of Child and Adolescent Psychiatry, Autism Resources Centre, Clinique Peyre-PlantadeCHU Montpellier
    • Research Team 4556 (Epsylon Laboratory)University Montpellier
  • C. Michelon
    • Department of Child and Adolescent Psychiatry, Autism Resources Centre, Clinique Peyre-PlantadeCHU Montpellier
    • Research Team 4556 (Epsylon Laboratory)University Montpellier
  • M. Soussana
    • Department of Child and Adolescent Psychiatry, Autism Resources Centre, Clinique Peyre-PlantadeCHU Montpellier
    • Research Team 4556 (Epsylon Laboratory)University Montpellier
  • C. Rattaz
    • Department of Child and Adolescent Psychiatry, Autism Resources Centre, Clinique Peyre-PlantadeCHU Montpellier
    • Research Team 4556 (Epsylon Laboratory)University Montpellier
  • M. C. Picot
    • CIC 1001, Unité Recherche Clinique (DIM)CHRU
Original Contribution

DOI: 10.1007/s00787-013-0388-8

Cite this article as:
Baghdadli, A., Brisot, J., Henry, V. et al. Eur Child Adolesc Psychiatry (2013) 22: 433. doi:10.1007/s00787-013-0388-8

Abstract

High-functioning autism (HFA) is characterized by persistent impairment in social interaction despite the absence of mental retardation. Although an increasing number of group-based programs for the improvement of social skills have been described, randomized controlled trials are needed to evaluate their efficacy. To compare the effect of a Social Skills Training Group-based Program (SST-GP) and a Leisure Activities Group-based Program (LA-GP) on the perception of facial emotions and quality of life (QoL) in young people with HFA. Eligible patients were children and adolescents with HFA. Participants were randomized to the SST or LA group. The primary outcome was defined as an improvement of 2 points in error rates for facial emotion labeling (DANVA2) from baseline. After the 6-month training period, the SST Group made fewer errors in labeling anger on adult faces, whereas error rates in the LA Group remained stable. Progress in the ability to recognize anger in the SST Group was due to better recognition of low intensity stimuli on adult faces. QoL increased in the SST Group in the dimension of school environment, as a marker of the transfer of skills acquired in the treatment setting to their use in the community. The SST-GP had higher efficacy than the LA-GP. Data justify replication using larger samples.

Keywords

High-functioning autismSocial skills trainingFacial emotion recognitionQuality of lifeControlled trials

Introduction

High-functioning autism (HFA) is characterized by persistent impairment in social interaction despite the absence of mental retardation [1]. This impairment generates durable social withdrawal and irritability, which can lead to depression and anxiety disorders and ultimately affect quality of life (QoL) [24]. Some studies suggest that social impairment in HFA is due to difficulties recognizing basic facial expressions [57].

Facial expressions are an important and early source for the identification of the emotional states of others and crucial for understanding their intentions. Newborns have facial preferences [8, 9] and they discriminate between facial emotion (fear, anger, sadness and surprise) from 7 months of age [10, 11]. By 4 years, normally developing children can freely label prototypical displays of happiness, sadness and anger with almost perfect accuracy, and are becoming able to recognize fear [12, 13]. The ability to interpret other people’s emotions improves then with age, achieving adult level around 10 years [1, 14].

Findings suggest that by 10 years, children with HFA are poorer than controls at labeling basic expressions [15]. Mazefsky and Oswald [16] demonstrate that individuals with HFA have difficulties when the stimuli are shown only briefly. Some authors suggest that individuals with HFA have impaired facial expression recognition because they cannot correctly categorize basic negative facial emotions, which could be related to their difficulties with interactive synchrony [1, 15, 17, 18]. Those data suggest that improved facial expression labeling should lead to enhanced emotion recognition and consequently improve social skills and QoL.

There are crucial questions regarding strategies that could improve social skills in people with HFA, and an increasing number of social skills training programs have been described [1, 19]. These involve teaching specific skills (both verbal and non verbal behaviors; e.g., making eye contact during social exchange) through behavioral and social techniques. Reviews of the efficacy of social skills training programs in HFA highlight that the heterogeneity of the studies (in terms of programs described, design, recruited population and outcome variables) limits their comparison [1].

The lack of controlled studies using specific and standardized measures of outcome also limits conclusions regarding efficacy. Ozonoff and Miller [20] compared individuals treated by social skills training programs with a control group without intervention, but their outcome measures were insufficiently standardized. Barnhill et al. [2] and Bauminger [21] used standardized measures of emotion labeling, but they had no control group. The study by Solomon et al. [22] is particularly interesting as it used a randomized design, objective and standardized measure of social cognition. However, the scope of their results is limited because the control group was a waiting list. Consequently, despite growing interest in social skills training programs, none of them have been validated [1] and further research is needed to find out how effective these programs really are, especially in comparison with non specific programs.

Methods

Participants and procedure

The participants with HFA were 8–12 years old. All patients fulfilling inclusion criteria were recruited from the Autism Resources Centre (University Hospital of Montpellier, France) from August 2007 to April 2008. Diagnoses were based on a multidisciplinary clinical assessment based on the International Classification of Diseases (ICD-10, [23]) and the Autism Diagnostic Interview-Revised (ADI-R, [24, 25]). The inclusion criteria were (a) diagnosis of infantile autism according to ICD-10 criteria (F 84.0), (b) an ADI-R score in the infantile autism range and an ADOS score [26] in the ASD range and (c) a verbal intelligence quotient (VIQ) in the normal range as measured by standard psychometric tests (WISC III, WISC IV [27] or K-ABC [28]). Due to transportation problems, we did not include children when there was more than 1 h of travel time between their home and our centre. The exclusion criteria were: participation in another SST program and parental or child refusal. After the participants (and their parents) signed a written informed consent form, and after baseline (V0) assessments, they were randomly assigned to one of the two intervention conditions (SST-GP or LA-GP).

Interventions

Both interventions were group-based programs. The sessions took place outside of school hours, in a quiet room in our clinic, each week at the same time. The two programs were conducted by the same two therapists who had training and expertise in working with children with HFA and in group therapy interventions. Each group met weekly for 1 h and 30 min for a total of 20 sessions (6 months) in accordance with the literature [21, 22, 29]. Details about each group are given in appendix 1.

The experimental treatment was a social skills training group-based program (SST-GP), inspired by Solomon et al.’s [22] program and recommendations made by Konstantareas [30], Krasny et al. [31] and Paul [32]. It proposed explicit training in social skills using techniques such as video modeling, social scenarios, problem-solving exercises and role-play. The program was gradual, beginning with essential social skills (such as non verbal communication of emotions; e.g., prosody, looks, gestures, etc.), and then moving up to more complex behaviors (including facial expressions of emotion, conversation, social problem solving, stress management and theory of mind). All sessions were filmed and supervised by a psychiatrist with experience in autism and the behavioral approach. Children were given a “logbook” in which they described the content of each session. The logbooks could be used to write down the tasks to be done at home between sessions. In order to replicate the program, we wrote a manual in which we detailed goals, techniques, stages and tools [33].

The control treatment was a structured leisure activities group-based program (LA-GP), in which social skills were not explicitly taught, and in which play was used as an approach: introduction to the five senses, music training, self-expression activities, visual-spatial activities and board games. This occupational approach is common practice in France, and there was no specific training schedule, assignment of homework or gradual progression.

Fidelity

At each session, therapists rated their fidelity to the manual [33] on a checklist reflecting key elements for that session. All group therapy interventions were video-recorded, reviewed and independently coded for fidelity by trained professionals.

Aims of the study

To compare the effects of a Social Skills Training Group-based Program (SST-GP) and a Leisure Activities Group-based Program (LA-GP) on the perception of facial emotions and QoL in young people with HFA.

Outcome measures

The primary outcome measure were DANVA2 scores (Diagnostic Analysis of Non Verbal Accuracy 2 (DANVA2)—Adult Facial Expressions (AF) and Child Facial Expressions (CF) [34]. These tests are used to assess affective decoding, i.e., the ability to recognize emotions in facial expressions. In our study, we only used the short form of the DANVA2 Adult facial expressions subtest (AF) and Child facial expressions subtest (CF). Children were presented with 24 pictures of adult facial expressions (AF), and 24 pictures of child facial expressions (CF), depicting four affective states (happy, sad, angry and scared) at either high or low intensity (Fig. 1). The DANVA2 was individually administered, following the procedure described in the DANVA2 manual. Stimuli were exposed for only 2 s, followed by a blank screen. Then another picture was shown for 2 s, by clicking on the button. The instruction manual [34] reports that all four subtests have been used with participants of both genders, and of various ages, intellectual abilities and psychological adjustment levels. The DANVA2 is easy to administer and studies have demonstrated internal consistency (current sample α = .61) and reliability over time (test–re-test reliabilities between 70 and 80 over 6–8-week periods) in populations of normally developing children, adolescents and adults. The DANVA 2AF has not been validated in populations of HFA but all participants had language skills that allowed them to understand the therapist’s instructions.
https://static-content.springer.com/image/art%3A10.1007%2Fs00787-013-0388-8/MediaObjects/787_2013_388_Fig1_HTML.jpg
Fig. 1

Description of the emotion of Anger DANVA 2AF

The secondary outcome measure was the Kidscreen-27 score (Health Related Quality Of Life questionnaire—HRQOL; Kidscreen-27 Parental Form) [35]. The parental form of the scale was chosen for different reasons: individuals with ASD tend to have limited self-awareness [36], none of the commonly used self-report measures have been validated specifically in individuals with ASD [37] and parents may be assumed to be the best source of information about their child’s well-being [38]. In this study, parents were given a standard HRQOL that has been developed across 13 European countries [39]. This 27-item questionnaire is filled in by parents to assess their perception of their child’s QoL. It targets children and adolescents aged between 8 and 18 years. Kidscreen 27 consists of five dimensions: physical well-being, psychological well-being, autonomy and parents, social support and peers, and school environment. The Kidscreen items assess either the frequency of behavior/feelings or, in a few cases, the intensity of an attitude. Both possible item formats use a 5-point response scale (not at all, a little, mildly, a lot, extremely), and the recall period is 1 week. The total score is obtained by adding the scores for each question. The average score for the general population in each domain is 50 and the standard deviation is 10. The reliability and validity of Kidscreen 27 has been demonstrated across Europe [39] and it was developed in consultation with parents and children.

Study design (ethics, sample size, randomization and implementation)

This randomized controlled trial was approved by the ethical committee of Southern France, the Advisory Committee on Data Processing in Health Research (CCTIRS) and the National Committee on Data Processing and Liberties (CNIL).

A 2-point difference between the two groups in the mean number of errors at DANVA 2AF/CF (primary outcome) was considered significant (power of 80 % and alpha risk of 5 %). Sample size was estimated at 38 patients (19 per group), considering a 10 % loss of follow-up. Of 20 children assessed for eligibility, 14 were included, which correspond to a power of 40 % (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00787-013-0388-8/MediaObjects/787_2013_388_Fig2_HTML.gif
Fig. 2

Flow diagram for individual randomized, controlled trials of non-pharmacologic treatment

Participants were enrolled for this study after signing informed consent forms. Each participant was randomly assigned to one of the two interventions (SST-GP or LA-GP) through a list of patients established by a computer number generator (DATA RANUNI function of SAS software). This list made by an independent statistical expert was given to the therapists who had defined a calendar of sessions for each program. At an informative meeting, the therapists distributed to participants and their families a kit containing the program schedule and the documents needed for participation in the program. Until the end of the study, none of the participants (children and parents) were told which group they had been assigned to. Assessments were made blindly (30 days before the first session and 30 days after the last session) by two experienced psychologists who were unaware of the interventions and were not part of the intervention team. Data was analyzed on a strict intention-to-treat basis.

Statistical methods

For each training group, we first carried out an intragroup analysis of change (mean or median gain) between the baseline (V0) and the follow-up point (V1), for the DANVA2 and the Kidscreen-27 scores (using Student’s t test and Mann–Whitney test). Second, differential scores between V0 and V1 were compared between the two groups to assess the impact of the intervention group. Concerning DANVA 2, we analyzed the change in the total score and subscores for each emotion and intensity. p Values lower or equal to .05 are considered significant. SAS statistical software version 9.2 was used for this analysis.

Results

Of 20 children who were screened for eligibility, 14 satisfied all eligibility criteria and underwent randomization, 7 to SST-GP and 7 to LA-GP. Intent-to-treat and treatment-completer sample sizes were 14 and 13, respectively (see Fig. 2). Recruitment began in 08/2007 and ended in 09/2007; pre-treatment and post-treatment assessments were completed, respectively in 10/2007 and 04/2008. One child dropped out after 8 sessions because his family moved. The study was well accepted by the children and their parents as their participation rate in the program was 100 %. Their adherence therapy was deemed excellent, based on their positive opinions expressed during a meeting organized at the end of the program and on the observation made by the therapists of the active cooperation of teenagers during each session.

Baseline characteristics of the sample

Baseline demographic and clinical findings are reported in Table 1. All participants were boys with infantile autism (according to ICD10 criteria) without mental retardation (Verbal intellectual Quotient, VIQ over 70). Children attended a regular (N = 12) or special (N = 1) class (5 days a week on average), and five of them benefited from individual tutoring (averaging 2 days a week). Eleven children received specialized outpatient servicing on a regular basis (speech or educational therapy), and there was no significant inter-group difference in the number of hours of therapy. In each group, one child had methylphenidate prescribed at least 3 months before our study.
Table 1

Demographic and clinical characteristics for each group at Baseline (V0)

 

SST-GP (N = 7)

LA-GP (N = 7)

Mean (SD)

Mean (SD)

Age (year)

10.7 (±1.8)

11.5 (±1.2)

VIQ

92.7 (±24.2)

97.6 (±20.3)

QoL-Kidscreen 27 domains

Physical well-being

37.3 (±6.5)

41.0 (±8.9)

Psychological

45.2 (±7.8)

45.1 (±14.3)

Autonomy and parents

40.8 (±9.4)

45.5 (±11.2)

Social support and peers

34.9 (±11.5)

30.9 (±7.6)

School environment

42.2 (±10.3)

49.5 (±11.5)

 

MEDIAN (IntQa)

MEDIAN (IntQa)

DANVA 2 AF

7.0 (5;9)

7.0 (5;8)

DANVA 2 CF

6.0 (3;9)

5.0 (4;7)

 

%

%

SESb

 Low

28.6

14.2

 Middle

71.4

42.9

 High

0.0

42.9

aIntQ = InterQuartile (Q25–Q75)

bParents’ socio-economical status (SES), scored as high (business owners, professionals, executives), middle (farmers, supervisors, skilled craftsmen), or low (farm workers, laborers, and service employees)

At baseline, the two groups did not differ significantly for VIQ, or for any demographic or outcome variables (emotion recognition and quality of life) (Table 1). Outcome variables for the whole sample were compared with normative data [34]. First, our entire sample scored lower on the DANVA2 than the general population of similar age. They made significantly more errors in labeling emotion on adult faces (HFA: M = 7.7, SD = 3.5; vs. normal M = 4.9, SD = 2.7; p < 0.05) and child faces (HFA: M = 5.77, SD = 2.3; vs. normal M = 3.4, SD = 2.3; p < 0.05). Second, the mean subscores of the sample on the Kidscreen-27 were below the cut-off point for an average QoL (mean = 50, SD = 10).

Treatment outcome

Treatment-completer analyses

For the primary outcome measure (Emotion recognition), we compared DANVA 2 (AF and CF) total scores between SST and LA Groups. Tables 2 and 3 show median error rates and inter-quartile (Q25–Q75) of each group at baseline (V0) and after 6 months of treatment (V1), and median gain between V0 and V1. There was no difference between the two groups for total error rates on the DANVA2 (adult faces, Table 2, and child faces, Table 3) at V1 and there was no significant change in the total score in both groups between V0 and V1.
Table 2

At V0 and V1 medians (and inter-quartile) for raw scores on the four emotions and intensity (DANVA adult face) for children of the SST and LA groups

Error rate

Groups

V0

V1

Median gain (V1–V0)

Effect sizea

p value

Adult face total

SST

7.5 (6;9)

6.5 (4;7)

−1.5 (−2;−1)

−0.3

0.64

LA

7.0 (5;8)

6.0 (4;10)

1.0 (−3;3)

  

Emotion and intensity

Happy

SST

0.5 (0;1)

0.0 (0;0)

0.0 (−1;0)

−0.5

0.53

LA

0.0 (0;1)

0.0 (0;1)

0.0 (−1;0)

  

Sad

SST

1.5 (0;3)

3.0 (1;3)

1.5 (0;3)

1.1

0.07

LA

1.0 (0;3)

1.0 (0;2)

−1.0 (−2;1)

  

Angry

SST

3.0 (2;4)

3.0 (1;3)

−1.0 (−2;−1)

−0.8

0.05

LA

3.0 (2;4)

3.0 (2;4)

0.0 (0;0)

  

 Low intensity

SST

2.0 (2;3)

2.0 (1;2)

−1.0 (−1;0)

−1.2

0.05

LA

2.0 (2;3)

2.0 (2;2)

0.0 (0;0)

  

 High intensity

SST

0.5 (0;1)

1.0 (0;1)

0.0 (0;0)

−0.14

0.80

LA

1.0 (0;1)

1.0 (0;2)

0.0 (0;1)

  

Fear

SST

2.5 (2;4)

1.0 (1;2)

−1.5 (−3;0)

−0.9

0.15

LA

1.0 (0;3)

2.0 (1;3)

0.0 (−1;2)

  

Medians are based on all available data for treatment completers; for SST n = 6; for LA n = 7

Statistically significant comparisons are given in bold

aEffect sizes are negative when the change in SST group is greater (more beneficial) than the change in LA group (Cohen’s d: 0.2 = small effect; 0.5 = medium; 0.8 = large effect)

Table 3

At V0 and V1 Medians (and inter-quartile) for raw scores on the four emotions (DANVA Child face) for children of the SST and LA groups

Error rate

Groups

V0

V1

Median gain (V1–V0)

Effect sizea

p value

Child face total

SST

6.5 (3;9)

6.0 (4;7)

0.0 (−1;2)

−0.3

0.62

LA

5.0 (4;7)

6.0 (3;9)

1.0 (−4;2)

  

Emotion

Happy

SST

0.0 (0;1)

0.0 (0;1)

0.0 (0;0)

−0.17

0.94

LA

1.0 (0;1)

1.0 (0;1)

0.0 (0;1)

  

Sad

SST

1.0 (0;2)

1.0 (1;3)

0.0 (−1;1)

0.17

0.77

LA

1.0 (0;2)

1.0 (0;1)

0.0 (−1;1)

  

Angry

SST

3.5 (2;5)

3.0 (1;4)

−0.5 (−1;0)

−0.9

0.14

LA

2.0 (2;3)

4.0 (1;5)

1.0 (−1;2)

  

Fearful

SST

1.0 (1;1)

1.0 (1;2)

0.0 (0;0)

0.25

0.56

LA

1.0 (1;2)

1.0 (0;2)

0.0 (−2;2)

  

Medians are based on all available data for treatment completers; for SST n = 6; for LA n = 7

aEffect sizes are negative when the change in SST group is greater (more beneficial) than the change in LA group (Cohen’s d: 0.2 = small effect; 0.5 = medium; 0.8 = large effect)

We then compared DANVA 2 (AF and CF) subscores between groups for emotions (happiness, sadness, anger, fear) and intensity of visual stimuli (high or low). At V1, children in the SST group made less errors in labeling anger on DANVA2 Adult faces than at V0, whereas errors rates in LA Group remained stable (Table 2). The post-treatment median gain was significantly different between the two groups (ΔSST GP = −1.0, IntQ = (−2;−1); ΔLA GP = 0.0, IntQ = (0;0); p = 0.05; Cohen’s d = −0.8, a large effect; [40]).

Further analyses showed that the progress in the ability to recognize anger in the SST Group was due to better recognition of low intensity stimuli (ΔSST GP = −1.0, IntQ = (−1;0); ΔLA GP = 0.0, IntQ = (0;0); p = 0.05; Cohen’s d = −1.2, a large effect) (Table 2). No significant differences between the two groups were found for the other emotions.

Overall mean error rates for the DANVA2 Child faces revealed that in both groups emotion recognition remained stable for happiness, sadness and fear whereas differences emerged between the two groups for anger, without significance (Table 3).

At the end of follow-up, we compared our sample with normative data. The results to the scale of DANVA Adult showed an improvement in our sample; they made the same number of errors as normally developing children They significantly made the same errors in labeling emotion on adult faces (HFA: M = 6.8, SD = 3.4; vs. normal M = 4.9, SD = 2.7; p > 0.05) and more errors for child faces (HFA: M = 6.1, SD = 2.6; vs. normal M = 3.4, SD = 2.3; p < 0.05).

For the secondary outcome measure (QoL), there were significant group differences in the Kidscreen-27 subscores between V0 and V1, in the dimension of school environment, which increased by 2.1 points in the SST group, while falling by 8.4 points in the LA Group (ΔSST GP = 2.1, SD = 6.2; ΔLA GP = −8.4, SD = 8.6; p = 0.01; Cohen’s d = 1.4). In the SST-GP, there was also a trend towards an improvement in QoL in the autonomy and parents domain (ΔSST GP = 6.9, SD = 6.9; ΔLA GP = −3.9; SD = 13, p = 0.1; Cohen’s d = 1) (Table 4).
Table 4

At V0 and V1 Means (and standard deviation) for 5 subscores on the Kidscreen 27 for children of the SST and LA groups

Measure

Groups

V0

V1

Mean gain (V1–V0)

Effect sizea

p value

Physical well-being

SST

35.8 (±5.6)

36.3 (±4.9)

0.4 (±4.8)

0.5

0.41

LA

41.0 (±8.9)

39.1 (±6.6)

−1.9 (±5.1)

  

Psychological well-being

SST

44.5 (± 8.3)

41.7 (± 8.1)

−2.9 (±6.2)

0.2

0.78

LA

45.1 (±14.3)

39.8 (± 11.9)

−5.3 (±19.9)

  

Autonomy and parents

SST

39.1 (±9.0)

46.0 (± 9.3)

6.9 (±6.9)

1.0

0.10

LA

45.5 (±11.2)

41.3 (±6.2)

−3.9 (±13)

  

Social support and peers

SST

31.9 (±9.3)

30.9 (±13.2)

−1.0 (±11.9)

0.7

0.25

LA

30.9 (±7.6)

21.4 (±10.7)

−9.5 (±12.9)

  

School environment

SST

39.3 (±7.6)

41.4 (±12.5)

2.1 (±6.2)

1.4

0.01

LA

49.5 (±11.5)

41.1 (±5.9)

−8.4 (±8.6)

  

Means are based on all available data for treatment completers; for SST n = 6; for LA n = 7

Statistically significant comparisons are given in bold

aEffect sizes are positive when the change in SST group is greater (more beneficial) than the change in LA group (Cohen’s d: 0.2 = small effect; 0.5 = medium; 0.8 = large effect)

Intent-to-treat analyses

Intent-to-treat (ITT) analyses were also conducted (N = 14). For the one child who dropped out, baseline scores were carried forward to post-treatment. All statistically significant group differences from the treatment-completer analyses remained significant in the ITT analyses.

Discussion

Despite the widespread clinical use of social skills training programs for children and adolescents with high-functioning autism (HFA), their efficacy has yet to be documented. The purpose of this study was to evaluate the effectiveness of a Social Skills Training group-based program (SST-GP) for improving facial recognition skills (measured using the DANVA2 AF, CF) and QoL (Kidscreen 27) of children and adolescents with HFA.

This randomized controlled trial examined whether using a specific and structured group-based approach that focused on social skills training improved outcomes in children and adolescents with HFA. To our knowledge, this study is one of the first to use a control group composed of children benefiting from some other type of intervention (a Leisure Activities Group-based Program) rather than a control group of children on a waiting list. Moreover, we used standardized and validated outcome measures. Data to determine the participants’ response to treatment were also recorded blindly by independent investigators. In addition, the impact of the two interventions on the participants’ quality of life was examined to discern whether the gains were generalized to daily living. Finally, to permit replication, our interventions were structured and fully documented in a manual [33].

In this pilot study, we found that treatment in the SST Group did not lead to a statistically significant improvement in facial recognition skills in children with ASD. Although non-significant, there was a greater reduction in the error rate for the SST group versus the LA group in this study, which corresponds to a standardized effect size between −0.3 and is considered to be a small treatment effect.

Before treatment, the two groups were similar in terms of age, socio-economic status, VIQ and outcome measures (DANVA2 and QoL-Kidscreen 27 scores). As previously shown, they scored lower than the norm in DANVA 2 subscores [34] and in the 4 domains of Kidscreen 27 [39]. Our study confirms the low level of QoL among adolescents with HFA even when compared with adolescents with chronic diseases such as diabetes [5].

Our results suggest a distinct impact of the SST-GP, as compared to LA-GP, for some subscores of the DANVA and Kidscreen 27. First, after the 6-month SST-GP, participants showed better perception of low intensity anger for DANVA2AF. These findings are concordant with those of Solomon et al. [22] and Wood et al. [41]. In our results, improvement in the SST group was observed for low intensity stimuli that are particularly problematic in autism [42]. One interpretation of this result could be that our program is effective at improving the categorization of emotions in their most subtle expression, which could be important for the outcome. Better recognition of anger by the SST-GP could also be a consequence of the progressive organization of our two modules. Teaching unpleasant emotion recognition and social problem solving ("Appendix") has certainly put the focus on the best way to anticipate expressions on the faces of other people to avoid certain everyday problems. This could result in lower social failures and partly explain the improved QoL. As several studies, including those using “eye tracking”, show a misinterpretation of emotions in autism, it is essential to develop therapeutic methods to improve the visual processing of social information [4244].

Second, only in the SST-GP, mean raw QoL scores (Kidscreen-27) in the school environment dimension increased significantly, suggesting a positive impact of SST on social skills required at school. This result is particularly interesting as the few existing studies of the QoL of children with HFA report more scholastic and social QoL problems than among typically developing children [45]. Thus, as stressed by Renty and Roeyers [46], efforts have to be made to improve inclusion of young individuals with autism because their QoL is associated with support characteristics and not only determined by autism severity.

The results of this study must be interpreted in the light of its limitations. First of all, the small sample size, limited by the need to keep group size small and manageable [1], limits its scope. Despite this, our results show significant effects of SST-GP relative to control condition. Second, the tool we used measures facial emotion labeling but does not assess its consequences for behavioral adaptation or everyday emotional feelings. It would be interesting to study this impact in future research using a measure of social adaptive skills. Third, we did not gather data about children and their families’ compliance with intervention and homework. Finally, we did not measure the maintenance over months of the positive effects of the SST-GP.

Conclusion

The results suggest that the SST Group-based Program might be useful for young people with HFA for improving their social skills and general adaptive functioning as part of a package of interventions. In addition, this research helps to promote controlled studies of the effectiveness of interventions proposed for individuals with HFA, and confirms the feasibility of controlled trials in this field. Our next objective will be to replicate the present study with a larger sample, using a multicentric protocol and a phase for testing the long-term maintenance of positive behavioral change.

Acknowledgments

This work was partially supported by a grant awarded by the first author of this paper from Pfizer Foundation. The authors are very grateful to the participating families and they also thank S. Nowicki for his valuable comments. Special thanks are extended to C. Boy and P. Antoine for their help with the bibliographical research, to Valerie Macioce for her careful reading of the draft and to C. Souche, F Séganti, and M Liratni for the collection of the data.

Conflict of interest

None.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013