**Exclusion Criteria Used in Early Behavioral Intervention Studies for Young Children with Autism Spectrum Disorder**

### **Sahr Yazdani 1, Angela Capuano 2, Mohammad Ghaziuddin <sup>2</sup> and Costanza Colombi 2,\***


Received: 16 January 2020; Accepted: 11 February 2020; Published: 13 February 2020

**Abstract:** This literature review evaluated early behavioral intervention studies of Autism Spectrum disorder (ASD) based on their participant exclusion criteria. The studies included were found through searching PsycINFO and PubMed databases, and discussed behavioral interventions for children up to 5 years of age with ASD and utilized a group research design. Studies reviewed were categorized into three groups: Restrictive exclusion criteria, loosely defined exclusion criteria, and exclusion criteria not defined. Results indicated that studies that used restrictive exclusion criteria demonstrated greater differences in terms of outcomes between experimental and control groups in comparison to studies that used loosely defined exclusion criteria and/or did not define any exclusion criteria. We discussed implications for the generalizability of the studies' outcomes in relationship to exclusion criteria.

**Keywords:** autism spectrum disorder; autism; literature review; comorbidity; early intervention; early intensive behavioral intervention; behavioral intervention

### **1. Introduction**

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that involves impairments in social communication, as well as the presence of stereotyped patterns of behaviors and interests [1]. ASD is considered a leading cause of disability in children under 5 years of age [2]. Given that ASD affects approximately 1 in 59 children in the United States [3], it is considered a serious public health concern [4]. This higher prevalence may be partially due to better detection and assessment procedures and an expanded definition of ASD [3,5,6].

While in the past, children with ASD were typically diagnosed around the age of 4 years shortly before entering school, they are now being diagnosed as early as the age of 2 years [7,8] and identified as at-risk for ASD between 12 to 24 mon of age [9]. With the increase in the number of young children being diagnosed, developing early age-appropriate interventions that can support parents and children is an international clinical and research priority [10,11].

Currently, research evidence indicates that high-intensity, long-term behavioral interventions are the most efficacious in supporting development and diminishing ASD symptoms and associated disabilities [12–17]. In a seminal study on behavioral intervention for children with ASD, Lovaas [14] demonstrated that children aged 40 to 46 mon who participated in intensive, long-term applied behavior analysis therapy achieved remarkable improvement in their skills. Specifically, nearly half of the children enrolled in intensive applied behavior analysis (for a minimum of 40 h per week), for at least 2 years showed significant gains in their adaptive and intellectual functioning, with some children becoming nearly indistinguishable from their typically developing peers. At long-term follow-up, the children who made significant gains maintained those gains, with placement in mainstream classrooms. This study led to widespread interest in behavioral interventions as promising treatments for children with ASD, spurring the development of educational treatment programs [18].

Despite the promising results found in the Lovaas [14] study, there was variability in the functioning of the study participants, with 40% of the participants continuing to meet criteria for developmental delays and needing educational supports. Replication of the Lovaas [14] study provided partial support for the treatment gains achieved, but with some disappointing results as the gains made during the replication were not as robust as the original study [18,19]. The variability in the results of the Lovaas [14] study have been related to variability in the severity of the study participants' ASD symptomatology, with participants with Pervasive Developmental Disorder, Not Otherwise Specified (PDD-NOS), a former diagnosis that included fewer symptoms than ASD, showing better outcomes than study participants that met full criteria for ASD [18].

Despite the positive impact of early intervention for preschoolers with ASD (age 12–72 mon), response to the intervention program is variable [18,20]. Outcomes for preschoolers who received early intervention range from loss of diagnosis to lack of improvement in the core ASD symptoms, from dramatic gains in language, cognitive, and adaptive skills to minimal treatment gains [21]. There are at least two possible reasons for the variability in the outcome of early-intervention studies. First, most studies do not describe the sample characteristics in detail. Even less is mentioned about the social and demographic factors that might influence the outcome [22]. Second, is the clinical heterogeneity of autism [23]. Despite the current custom of conceptualizing autism as a spectrum disorder following the publication of fifth edition of The Diagnostic and Statistical Manual of Mental Disorders (DSM-5) [1], it may be the case that subtypes exist within the autistic spectrum [24].

In addition to the possible subtypes of autism, several medical and behavioral conditions are known to co-exist with it. It is estimated that approximately 75% of individuals with ASD present with associated medical conditions, genetic syndromes, or mental health disorders [3,25]. On the other hand, in some studies, due to the attempt to recruit homogeneous samples of individuals with "pure" ASD, children with associated conditions such as epilepsy, severe intellectual disabilities, or genetic abnormalities, are not included [12]. Many studies also used small clinical samples or lacked details about the ASD characteristics that lead to diagnosis [22].

Thus, by excluding persons with ASD who have associated medical and behavioral disorders, who constitute the majority of the general ASD population [26], these stringent exclusion criteria significantly reduce the generalizability of results and reduce their utility in the real world. Without knowing the characteristics of the children who benefit from the intervention, it is difficult to make treatment recommendations in clinical practice. This review aimed to examine the exclusion criteria used in the early-intervention studies of ASD, in order to ascertain how these criteria are related to the efficacy of behavioral interventions for young children with ASD.

### **2. Materials and Methods**

Our review included 26 papers written between 2002 and 2018 that highlighted studies with three varying levels of exclusion criteria used in early behavioral interventions for children with ASD. PubMed and PsycINFO were the databases used to identify articles included in this review. Search terms used included various combinations of the following terms: "Early intervention", "Autism", "Autism Spectrum Disorder", "children with autism", "children with ASD", "clinical trial", and "group design". A filter limiting the results to publication years of 2002 to 2018 was applied. Other studies were found from the reference list of the articles that met these inclusion criteria. The search was conducted through December 2018.

The titles and abstracts of these studies were reviewed by the first, second, and fourth authors for appropriateness to include in the literature review, particularly for the inclusion of a behavioral intervention and the age of study participants. Inclusion criteria for this review were studies that (1) used participants between the ages of 2 and 5 years with Autism Spectrum Disorder, (2) investigated a behavioral intervention, (3) used a group design, and (4) were published within the last 15 years. Group studies were the focus of this review so that comparisons could be drawn among studies. Early behavioral intervention was another focus of this review, which is why studies that only used young participants and behavioral intervention were included. Given the increasing prevalence of ASD and corresponding treatments, the focus was also on studies recently published. Studies using both DSM-IV-TR [27] and DSM-5 [1] criteria were included, as there were few studies using DSM-5 criteria. Studies that employed single-case design and nonbehavioral interventions, such as dietary and pharmacological interventions, were excluded. No language filters were applied, but only one study was excluded for being in a language other than English.

### **3. Results**

There were 26 studies found based on the search methods and inclusion criteria specified above, published between the years of 2002 and 2018. For this review, the term "restrictive exclusion criteria" categorizes studies that excluded children with comorbidities and/or associated family mental health conditions. The term "loosely defined exclusion criteria" defined studies that included children with comorbidities but excluded certain individuals on the basis of other factors, such as distance of the family from the treatment center, non-English-speaking participants, or severe sensory or motor deficits. The term "exclusion criteria not defined" highlighted studies that did not significantly excluded any children. A summary of all studies can be found in Table A1.

### *3.1. Restrictive Exclusion Criteria*

Of the studies, 57% (*n* = 15/26) used comparably restrictive exclusion criteria to select their participants. Studies with this type of restrictive criteria mainly excluded participants with medical conditions other than ASD, such as genetic syndromes, epilepsy, and intellectual impairments.

Perera, Jeewandara, Seneviratne, and Guruge [28] investigated an early-intervention program for children aged 18 to 40 mon in Sri Lanka. Study participants were children who had just received an initial diagnosis of Autism, were 18 to 40 mon in age, and had never received behavioral or developmental intervention previously. Participants were excluded if they had a diagnosis of PDD-NOS or Asperger's Disorder, had severe cognitive impairments, experienced co-occurring sensory or motor disorders, genetic disorders, or if they had participated in developmental intervention prior to joining the study. Experimental group participants received home-based therapy in which their mothers were taught to use developmental and behavioral interventions to use with their children. Participants in the comparison group had received a diagnosis of autism over the age of 40 mon and did not receive any autism-specific developmental intervention. This study did not use random assignment. Results indicated that the children in the experimental group showed more improvement on measures of autism severity and social interaction, despite some improvement in the children in the comparison group.

Brian, Smith, Zwaigenbaum, and Bryson [29] conducted a cross-site, randomized, controlled trial investigating the efficacy of a parent-mediated intervention, social ABCs, for toddlers aged 16 to 30 mon with suspected or confirmed ASD. Exact numbers of male and female participants were not given. Inclusion criteria included children who met criteria for ASD or displayed behaviors consistent with ASD, did not spend more than half their time in childcare, were products of full-term delivery, and had a birthweight above 2500 g. Exclusion criteria included the occurrence of any co-occurring genetic, neurological, or severe sensory or motor conditions. Results indicated that children in the treatment group showed more gains in functional vocal responsiveness to parent prompts and child vocal initiation as compared to the control group.

Rogers et al. [30] conducted a randomized controlled trial with 98 children (76 boys) aged 12 to 24 mon. The study strove to investigate the efficacy of the Early Start Denver Model (ESDM), which fosters parental involvement within a child-centered interactive context and may be compared to conventional community therapies. Inclusion criteria specified that the children met risk criteria for ASD in a clinical assessment, were ambulatory, had a development quotient of 35 or higher, and

primarily spoke English at home. The exclusion criteria included children who had parents that self-reported mental illness or substance abuse, children who had significant medical conditions such as cerebral palsy, a gestational age of less than 35 weeks, and/or genetic disorders related to developmental disabilities, or individuals who had current or prior enrollment in an intensive 1:1 autism intervention curriculum for more than 10 h per week. The main outcomes of this study were that individuals who had received parental training with the ESDM technique established more productive working alliances with their therapists as compared to the community group. However, the effects seen in intensive-treatment studies were not observed. They demonstrated that younger age and greater intervention positively affected the developmental rates for children with autism.

Carter et al. [31] conducted a study with 62 children (51 boys) aged 15 to 25 mon. The study aimed to investigate the efficacy of Hanen's More Than Words (HMTW), a parent-implemented intervention, as compared to a control group. The inclusion criteria required the children to meet the diagnostic criteria of ASD and to be recruited from ASD specialty clinics. Children with a genetic disorder, those who did not obtain a predetermined "at-risk" score on the Screening Tool for Children with Autism (STAT), or those who did meet the symptom criteria for an ASD diagnosis based on clinical evaluations were excluded. The main outcomes of this study were that the HMTW group showed differential effects on child communication. However, parents of children who possessed higher object interest may require additional support to implement proper strategies.

Dawson et al. [12] evaluated the efficacy of the ESDM with a sample size of 48 children aged 18 to 30 mon. Exact numbers of male and female participants were not given, but the ratio of males to females was 3.5 to 1. The inclusion criteria for this randomized controlled trial stipulated that the children meet criteria for ASD on the Toddler Autism Diagnostic Interview and Autism Diagnostic Observation Schedule (ADOS), receive a clinical diagnosis for ASD based on DSM-IV criteria, reside within half an hour of the testing location, and demonstrate a willingness to participate in a two-year or greater intervention program. Children who had a neurodevelopmental disorder of known etiology, significant sensory or motor impairments, major physical problems such as chronic or serious health conditions, seizures at the time of entry, use of psychoactive medication, a history of serious head injury or neurological disease, alcohol or drug exposure during the prenatal period, or developmental quotient below 35 were excluded. The main outcomes of this study were that the children who received ESDM training demonstrated significant improvements in IQ scores and adaptive behavior and were more likely to have a change in diagnosis to pervasive developmental disorder. Moreover, the comparison group manifested greater delays in adaptive behaviors and demonstrated minimal improvement in baseline scores.

Kasari, Gulsrud, Wong, Kwon, and Locke [32] aimed to identify if a joint attention intervention would result in greater engagement between caregivers and toddlers with autism. The randomized controlled trial investigated 38 children (29 boys), aged 21 to 36 mon. Inclusion criteria stated that children must have met criteria for autism following DSM-IV criteria by an independent clinician; children with additional syndromes were excluded. The main outcomes were that both caregivers and toddlers in the experimental group made significant improvements in areas of joint engagement, including responsiveness to joint attention and diversity of functional play acts, as compared to the control group.

Zachor and Itzchak [33] compared the efficacy of applied behavior analysis (ABA) and the integration of several intervention approaches for children with varying levels of autism severity. The quasi-experiment investigated a sample size of 78 (71 boys), aged 15 to 35 mon. Participating children had to meet a clinical diagnosis of autism based on DSM-IV criteria and the cut-off points on the ADI-R (Autism Diagnostic Interview-Revised); those with additional major medical diagnoses or incomplete post-intervention assessments were excluded. While there were no significant between-group differences in terms of improved cognitive abilities or adaptive skills, Zachor and Itzchak demonstrated that in the group with less severe baseline ASD symptoms, the children who had

received the eclectic intervention approach had better outcomes in communication and socialization adaptive skills.

Itzchak and Zachor [34] also sought to characterize the stability and changes of autism diagnosis in correlation with pretreatment predictors and post-intervention outcomes. The open-design study investigated a sample size of 68 (62 boys), aged 18 to 35 mon. Inclusion criteria required that the child met established DSM-IV criteria for autism. Exclusion criteria were comorbidities, including genetic syndromes and seizure disorders. The main outcomes of this experiment suggest that individuals who had a changed diagnostic classification to ASD or Off Spectrum had better receptive language scores, as well as significant improvements in cognitive outcomes, adaptive outcomes, and reduction of stereotyped behaviors, as compared to individuals within the unchanged classification group.

Kasari, Paparella, Freeman, and Jahromi [35] investigated the effects of joint attention (JA) and symbolic play (SP) behavioral interventions in accordance with prediction to language outcomes. The study analyzed a sample size of 46 boys, aged 36 to 48 mon. Inclusion criteria required that the children had been diagnosed with autism on the ADI-R and ADOS scale, had to be of 5 years of age or younger, and had to be accessible for follow-ups. Exclusion criteria included seizure disorder and additional medical diagnoses, such as genetic syndromes. The main outcomes of this experiment included greater JA and SP skills and ability to execute these skills during play, within the respective groups as compared to the control group.

Ben-Itzchak and Zachor [36] sought to understand the correlation between cognitive, socialization, and communication pre-intervention variables to outcome in children with autism post-intervention. The study investigated a sample size of 25 (23 boys), aged 20 to 32 mon. Inclusion criteria included children diagnosed using the ADI-R and ADOS protocols. Exclusion criteria included children who demonstrated comorbidities, including genetic syndromes and seizure disorders. The main outcomes of this experiment were that the children demonstrated significant improvements in imitation, receptive and expressive language, nonverbal communication, play skills, and stereotyped behaviors.

Remington et al. [37] investigated the effects of early intensive behavioral intervention for children with autism. The quasi-experiment analyzed a sample size of 44, aged 30 to 42 mon. Exact numbers of male and female participants were not given. Inclusion criteria included that the children had to be diagnosed with autism based on the ADI-R, had a previous diagnosis of autism by a clinician independent of the research program, or had a suspected diagnosis of autism, to be between 30 and 42 mon of age at the time of induction, and had to live in their family home. The exclusion criteria included that the child had to be free of any other chronic or serious medical conditions that might interfere with the ability to deliver consistent intervention or might adversely affect development. The main outcomes included significant improvements in IQ scores, daily living skills, motor skills, and language abilities subsequent to the interventional therapies. Moreover, children who participated in the early behavioral intervention therapy were more likely to attend mainstream schools, as compared to children within the control group.

Zachor, Ben-Itzchak, Rabinovich, and Lahat [38] compared the Eclectic-Development (ED) and ABA intervention approaches in children with autism. The quasi-experiment analyzed a sample size of 39 (37 boys), aged 22 to 34 mon. Inclusion criteria included that the children were diagnosed with autism using the ADI, met established criteria for Autism/PDD-NOS according to DSM-IV criteria. Exclusion criteria included children who had medical abnormalities such as seizures or hearing deficiencies. The main outcomes of this experiment demonstrated that ABA intervention approaches provided children with greater improvements in language communication and social interaction, as well as allowed for greater changes in diagnostic classifications, as compared to ED intervention approaches.

Cohen, Amerine-Dickens, and Smith [39] sought to investigate the effects of early intensive behavioral treatment (EIBT) for children with autism. The quasi-experiment utilized a sample of 42 (35 boys), aged 20 to 41 mon. Inclusion criteria included that children had a primary, previous, and psychological diagnosis of autistic disorder or pervasive development disorder confirmed by ADI-R, pretreatment IQ above 35 on the Bayley Scales of Infant Development-Revised (BSID-R), chronological

age between 18 and 42 mon at diagnosis and under 48 mon at treatment onset, residence within 60 kilometers of the treatment agency, and parental agreement to active participation. Exclusion criteria included children who had a severe medical limitation or illness, including motor or sensory deficits, that would prevent a child from participating in treatment for 30 h a week, and children who had undergone more than 400 h of prior behavioral intervention. The main outcomes of this experiment suggested a significant difference in the IQ scores and adaptive behavior for children who had undergone the EIBT, and a significant increase in EIBT children in regular education as compared to the control group. However, there were no significant between-group differences in language comprehension or nonverbal skills.

Kasari, Freeman, and Paparella [40] examined the efficacy of JA- and SP-targeted interventions. The randomized controlled study investigated a sample size of 58 (46 boys), aged 36 to 48 mon. Inclusion criteria included that children had a diagnosis of autism on the ADI-R and ADOS, were of 5 years of age or younger, and were accessible for follow-ups. Exclusion criteria included no seizure disorders or additional medical diagnoses, and children whose parents demonstrated refusal of final assessments or who left the program unexpectedly. The main outcomes of this experiment demonstrated improvements of JA and SP within the respective experimental groups, as well as significantly greater growth in expressive language for the individuals within these groups.

Eikeseth, Hayward, Gale, Gitlesen, and Eldevik [41] investigated the outcomes of varying intensities of early behavioral intervention for children with autism. The open-design study initially analyzed a sample size of 23 (17 boys), aged 28 to 42 mon. Inclusion criteria included diagnosis of autism according to the ICD-10 (International Classification of Diseases), chronological age at intake between 24 and 42 mon, the absence of other severe medical conditions as certified by a medical practitioner, and if the child resided outside of the catchment area for the clinical-based services. Exclusion criteria included an increased intensity of supervision due to lack of acquisition (as was the case for one child). The main outcomes of this experiment demonstrated a correlation between the intensity of supervision with changes in IQ scores and visual-spatial IQ after 14 mon. However, there was no significant correlation with the intensity of supervision and adaptive functioning.

Many of the studies that fell within the restrictive exclusion criteria category demonstrated positive outcomes of early behavioral interventions on various developmental skills including autism severity, verbal communication, social interaction, and other markers of development in comparison to control groups. Thus, these studies demonstrated promising results in improvement of many skills for young children with ASD. However, the restrictive nature of these studies limits the applicability of their outcomes to a wider audience of children with ASD who present with some form of comorbidity.

### *3.2. Loosely Defined Exclusion Criteria*

Of the studies discussed in this review, 15% (*n* = 4/26) utilized loosely defined exclusion criteria for their early-intervention behavioral treatments. Studies with loosely defined criteria included children who experienced ASD with comorbidities but excluded subjects based on other factors, such as primary language and accessibility to testing sites, or severe motor or sensory deficits.

Yoder and Stone [42] evaluated two different communication interventions: Responsive Education and Prelinguistic Milieu Teaching (RPMT) and the Picture Exchange Communication System (PECS) in preschool children with ASD. The randomized group experiment included 36 children with a diagnosis of ASD or PDD-NOS aged 18 to 60 mon, who demonstrated communication deficits and passed hearing screenings. Of the participating children, 31 were boys. Participants were excluded from the study if they demonstrated severe sensory or motor deficits or if English was not the primary language spoken in the home. Of the 120 children who were screened for participation in the study, only 60 met inclusion criteria. Results demonstrated mixed results, with RPMT demonstrating better effects with generalized turn taking and generalized joint attention initiation as compared to PECS. Conversely, PECS demonstrated better effects with generalized requests in children who arrived to the study with little initiation of joint attention.

Oosterling et al. [43] strove to understand the efficacy of non-intensive parental training in combination with standard care for children with autism. The randomized, controlled trial investigated a sample size of 75 (52 boys), aged 12 to 24 mon. Inclusion criteria included children with a clinical diagnosis of ASD or PDD-NOS, a demonstrated developmental potential at 12 mon, and a developmental quotient below 80. Exclusion criteria included family problems that may interfere with parental training and insufficient parental proficiency in the native language, Dutch. The main outcomes of this experiment suggested that additional non-intensive parental training did not have any influence on language and global clinical improvement outcome variables.

Wetherby et al. [44] sought to compare the effects of two parent-implemented Early Social Interaction (ESI) interventions. The randomized, controlled trial investigated a sample size of 82, aged 16 to 20 mon. Exact numbers of male and female participants were not given, but the individual ESI group contained 81% male participants, and the group ESI contained 92.5% male participants. Inclusion criteria included children who had received an ASD diagnosis between ages 16 to 20 mon and lived within 50 miles of either research site. Exclusion criteria included children who demonstrated participation in other interventional research studies. The main outcomes demonstrated that children within the individual social intervention groups improved their social communication, daily living, receptive language, and social skills, while children within the group intervention groups demonstrated worsening or no significant change in these measures.

Howard, Sparkman, Cohen, Green, and Stanislaw [45] compared the effects of intensive behavior analytic intervention (IBT), intensive eclectic intervention, and non-intensive public early-intervention programs in children with autism. The quasi-experiment investigated a sample size of 61 (54 boys), all less than 48 mon of age. Inclusion criteria included children who were independently diagnosed with Autistic Disorder or PDD-NOS according to DSM-IV criteria, entry into an intervention program before 48 mon of age, English spoken as the primary language within the child's home, no significant and separate medical condition, and no prior treatment of more than 100 h. Exclusion criteria included individuals who had not completed the 7 mon of intervention, and parents who could not be contacted to arrange follow-up testing despite repeated attempts or refusal of testing. The main outcomes of this trial demonstrated that individuals who participated in the IBT group performed significantly higher in tests for IQ, nonverbal and verbal language, overall communication, and social skills.

The studies that utilized loosely defined exclusion criteria provide a stronger foundation to apply certain early-intervention behavioral methods to a wider range of children with ASD, given that they included a more diverse participant pool. However, not only are there a limited number of studies available with this type of exclusion criteria, but the criteria were often so specific to the particular study that it inhibited any potential conclusions that may be drawn in understanding the applicability of these outcomes to a wider range of children with ASD. This could compromise the generalizability of the results of these studies to a wide range of children with ASD.

### *3.3. Exclusion Criteria Not Defined*

Of the studies discussed in this review, 30% (*n* = 7/26) did not specifically list any exclusion criteria for the participants of their early-intervention behavioral treatments, and thus, the results of these studies may be applied to the comparably widest range of children with ASD.

Welterlin, Turner-Brown, Harris, Mezibov, and Delmolino [46] implemented the Treatment and Education of Autistic and Communication Related handicapped Children (TEACCH) program in home-based models for parents of toddlers with ASD. Inclusion criteria for the study were chronological age of less than 42 mon and a diagnosis of Autism. No other exclusion criteria were specified. Twenty children participated in the study and were randomly assigned to receive TEACCH intervention at home or wait-list control. Six children participated in the experimental group and, of these, five were male. Participants were matched for data analysis between the experimental and control groups on the basis of similar age. Results between the experimental and control group did not reach statistical significance, which the authors attributed to low sample size and short time frame.

Reed, Osborne, and Corness [47] conducted a study of 33 children who were nonrandomly assigned to treatment groups. Inclusion criteria were as follows: Age of 2 years, 6 mon to 4 years, 0 mon at the start of their intervention, and a diagnosis of ASD. No details were given about the number of males and females that participated. The only exclusion criterion specified was that the children participating in the study must not have been involved in any other major intervention at the same time as the study. Children were divided into one of three treatment groups. One group received preschool special education, another received special education designed specifically for autism, and the final group received in-home one-on-one behavioral treatment. After 10 mon of intervention, results from the three groups were compared, with some improvement in measures used across both special education groups. Children in the home-based program showed improvement across the Psych-Educational Profile and British Abilities Scale, but not for the Vineland Adaptive Behavior Scales.

Smith, Flanagan, Garon, and Bryson [48] examined Pivotal Response Training (PRT) in an Early Intensive Behavioral Intervention (EIBI) program delivered in the community. Inclusion criteria for the study were: Having a diagnosis of Autism Spectrum Disorder and age below 6 years. Children who met eligibility criteria were randomly assigned to participate in the experimental group. No control group was used. Rather, participants were divided into subgroups for data analysis, based on their scores on measures of intellectual functioning. Results demonstrated that all study participants, regardless of cognitive functioning level, showed significant improvement in communication skills and adaptive functioning, with larger gains found for the children in the moderate and high cognitive functioning groups.

Fernell et al. [49] conducted a naturalistic, prospective study with 208 children aged 1<sup>1</sup> <sup>2</sup> to 4<sup>1</sup> 2 years. No information was given on the number of males and females included in the study. Children included in the study had a previous diagnosis of Autism that was confirmed through further testing for inclusion in the study, but no exclusion criteria were given, beyond parents' language proficiency in Swedish or English. All children in the study received some form of applied behavior analysis (ABA), and participants self-selected into intensive ABA or non-intensive ABA. There was no control group. This study showed that study participants improved in several areas of functioning, and participants in intensive intervention did not show more improvement than participants in non-intensive intervention.

Landa, Holman, O'Neill, and Stuart [50] evaluated the effects of a curriculum aimed to improve socially synchronous behaviors for children with autism. The randomized, controlled trial investigated a sample size of 48 (40 boys), aged 21 to 23 mon. The inclusion criteria specified that the children met criteria on the ADOS, received a diagnosis of ASD from an expert clinician, had a nonverbal mental age of at least 8 mon, had no siblings with ASD, English was the primary language spoken within the home, and no known etiology for ASD. No exclusion criteria were specifically listed. The main outcomes for this experiment included significant between-group differences for socially engaged imitation, but no significant between-group differences for shared positive affect, expressive language, or nonverbal cognition.

Ingersoll [51] evaluated the efficacy of Reciprocal Imitation Training (RIT) in development elicited and spontaneous imitation skills in children with autism. The randomized, controlled trial investigated a sample size of 21 (18 boys), aged 27 to 47 mon. The inclusion criteria mandated that the children receive a clinical diagnosis of autism based on DSM-IV-TR criteria and met the cut-off for ASD on ADOS. There were no exclusion criteria that were explicitly listed. The main outcomes for the experiment included significantly more gains in elicited and spontaneous imitation for both objects and gestures, as compared to the control.

Reed, Osborne, and Corness [52] investigated the efficacy of home-based early behavioral interventions for children with autism. The quasi-experiment investigated a sample size of 27 (27 boys), aged 31 to 48 mon. Children included in the study were within 2 years, 6 mon and 4 years of age, received no other major intervention during the period of assessment, and had a diagnosis of ASD. The exclusion criteria were not listed. The main outcomes of this experiment demonstrated significant

between-group differences in educational functioning, with no significant between-group differences for intellectual functioning, adaptive behavior, and ASD severity.

The studies described above that did not specifically exclude children from participating in the study present results that are generalizable to the broadest population of children with ASD. However, this same lack of any exclusionary criteria also prevents understanding the specific methods of treatment necessary for the many different types of children who are diagnosed with ASD. Thus, the wider generalizability leads to fewer conclusions that can be drawn about the applicability of these results to any one specific child.

### **4. Discussion**

This review evaluated 26 early-intervention behavioral studies of ASD based on their exclusion criteria into three categories: Restrictive, loosely defined, and not defined. These categories carry critical implications, as these categories define which of their outcomes may be applied to various audiences of children with ASD.

There were 15 studies that utilized restrictive criteria risk excluding approximately 75% of children who have ASD with a comorbid condition, including the 10% with a co-occurring psychiatric disorder, and the 4% with a genetic or chromosomal disorder [53]. Others excluded children with common neurological conditions, such as fragile X syndrome or epilepsy, which are strongly associated with autism [25]. Prevalence of ASD in children with epilepsy is around 6.3% with higher prevalence up to 47% in children with other forms of seizure disorders [54]. Other studies excluded children born before 35 weeks, although some studies suggest that about 7% of preterm infants might develop autism [55]. These studies may exclude a large group of individuals with ASD. Although many of these studies categorized the children as having improved, the results suggest that interventions work only for the minority of children who have "pure" ASD. For this reason, it is not possible to conclude that early intervention works in all children with ASD.

The four studies that utilized loosely defined exclusion criteria and the seven studies that did not define any exclusion criteria may have included children with comorbid disorders that could have influenced their findings. Indeed, these studies showed mixed results, with some experimental groups showing more improvement than control groups, and others showing no significant between-group differences. Inclusion of comorbidities makes these studies' results more applicable to a wide range of children with ASD, but also makes it difficult to know which interventions might be efficacious for specific comorbidities with ASD, since inclusion of comorbidities was typically not limited to only specific disorders.

We believe that studies that investigate behavioral interventions for young children with ASD should make more of an effort to recruit and include study participants with comorbid conditions in addition to ASD, which could make their results more applicable to a wider range of children with ASD. It will also be important for these comorbid conditions to be explicitly listed in the participant characteristics so that conclusions can be drawn about how efficacious certain behavioral interventions are for children with ASD and associated conditions. Listing the participants with these descriptors may make it easier to understand what population of children with ASD may be most likely to benefit from the interventions studied.

Current guidelines suggest not to exclude individuals with associated conditions if these are common. Given the number and incidence of comorbid disorders it may be hard to try to identify individuals who only meet criteria for ASD and no other disorders. Moreover, this may not be representative of the population of children with ASD. This review highlights the possible influence of treatment modifiers such as comorbidity in the outcome of behavioral interventions for young children with ASD. Overall, the results suggest that the heterogeneity observed in the response to early behavioral intervention in children with ASD may be related to various comorbid conditions. They underscore the need to systematically screen for the presence of comorbid symptoms and conditions at the time of recruitment of subjects, identify these in their studies, and modify intervention

methods accordingly. How those interventions should be modified remains unclear as there is not yet enough research evidence to suggest what are evidence-based interventions for ASD with comorbid conditions.

A supplementary table, depicting the studies included in this review, grouped by intervention type, is available in Table A2.

### **5. Limitations**

There are some important limitations in this literature review. To begin, this review only included studies that used a group design. This is an important limitation about the results of this review, given that many studies investigating a behavioral intervention for young children with ASD use single-subject research design [56], which has been increasing over recent years [57]. However, group study designs for investigating behavioral interventions for individuals with ASD are an important part of identifying evidence-based practices for ASD [58] and allow for decisions to be made about the efficacy of a particular intervention [57]. In addition, the research databases used (PsychINFO, PubMed) are widely used and represent many research studies, but they are not inclusive of all research being conducted, so it is possible that some studies that could have met this review's inclusion criteria were missed.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.


**Table A1.** Summary of early-intervention studies reviewed.

**Appendix A**



adaptive skills








differences in language

comprehension

 or nonverbal skills

**Table A2.** Studies reviewed grouped by intervention type.






**Table A2.** *Cont.*

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Brief Report* **Increased Neural Reward Responsivity in Adolescents with ASD after Social Skills Intervention**

### **Elizabeth Baker, Elina Veytsman, Ann Marie Martin, Jan Blacher and Katherine K. M. Stavropoulos \***

Graduate School of Education, UC Riverside, Riverside, CA 92521, USA; ebake001@ucr.edu (E.B.); eveyt001@ucr.edu (E.V.); amart135@ucr.edu (A.M.M.); Jan.Blacher@ucr.edu (J.B.)

**\*** Correspondence: Katherine.Stavropoulos@ucr.edu

Received: 22 May 2020; Accepted: 22 June 2020; Published: 24 June 2020

**Abstract:** The reward system has been implicated as a potential neural mechanism underlying social-communication deficits in individuals with autism spectrum disorder (ASD). However, it remains unclear whether the neural reward system in ASD is sensitive to behavioral interventions. The current study measured the reward positivity (RewP) in response to social and nonsocial stimuli in seven adolescents with ASD before and after participation in the Program for the Education and Enrichment of Relational Skills (PEERS®) intervention. This study also included seven neurotypical adolescents who were tested at two time points but did not receive intervention. We examined the RewP across the course of a task by comparing brain activity during the first versus second half of trials to understand patterns of responsivity over time. Improvements in social skills and decreased social-communication impairments for teens with ASD were observed after PEERS®. Event-related potential (ERP) results suggested increased reward sensitivity during the first half of trials in the ASD group after intervention. Adolescents with ASD who exhibited less reward-related brain activity before intervention demonstrated the greatest behavioral benefits from the intervention. These findings have implications for how neuroscience can be used as an objective outcome measure before and after intervention in ASD.

**Keywords:** autism spectrum disorder; EEG; ERP; reward response; RewP; sensitization; social skills intervention; PEERS®

### **1. Increased Neural Reward Responsivity in Adolescents with ASD after Social Skills Intervention**

The cognitive process of habituation can be conceptualized in a variety of ways, but is generally considered a decreased response to stimuli after repeated exposure [1]. Individuals with autism spectrum disorder (ASD), defined by social communication deficits and the presence of restricted interests and repetitive behaviors [2], display altered rates of habituation. Specifically, individuals with ASD do not habituate to social information at the same rate as neurotypical controls, as evidenced through amygdala activation to faces over time [3–7]. In individuals with ASD, repeated presentation of social information elicits activation rates similar to that of novel stimuli for neurotypical subjects [8]. In neurotypical individuals, habituation tends to occur at a lower rate for stimuli that are more salient, intense, or stimulating [1,9]. Salient information may cause sensitization to stimuli, such that heightened responses can be observed over time [1,10]. One explanation for slowed habituation rates in response to faces is that individuals with ASD find processing social information more challenging than their neurotypical peers and thus must employ more cognitive resources. Alternatively, lack of habituation could reflect sensitization in this population.

Beyond reflecting the allocation of cognitive resources, habituation is also an indicator of learning. Reinforcement learning is facilitated by the goal of maximizing rewards and satisfying desired

outcomes. The reward system has been discussed at length in relation to the core symptoms of ASD. According to the social motivation hypothesis, individuals with ASD experience social interactions as less rewarding than their neurotypical peers, which may lead to reduced social initiation during critical periods of social development [11]. Investigations utilizing electroencephalography (EEG) to measure reward-specific event-related potentials (ERPs) suggest that children with ASD tend to find nonsocial stimuli more salient than social stimuli, and that children with ASD have less reward-related brain activity than that of their neurotypical peers in response to faces [12]. Thus, it is not that the reward system in ASD populations is under-active in response to all stimulus types, but that it is selectively functioning for some categories and not others [13]. However, the literature is mixed on whether the reward system is globally hypoactive in individuals with ASD [14,15]. If the reward system is selectively functioning in ASD, this system might be malleable, and behavioral intervention strategies that focus on social reinforcement might increase brain activity in response to social stimuli in this population. This hypothesis is supported by previous literature demonstrating neural changes in participants with ASD from pre- to post-intervention [16–22].

Social skills interventions for individuals with ASD often implement strategies of reinforcement learning, including applied behavior analysis and social skills training [23–25]. The goal of many interventions is to provide training for independent skill acquisition, ranging from a reduction in maladaptive behavior to increasing social engagement at school. Considerations of habituation or sensitization before and after such interventions are pertinent to not only the effectiveness of intervention but also the interpretation of outcomes.

Understanding how reward-related brain activity changes across the course of a task for individuals with and without ASD can increase our understanding of whether habituation or sensitization occurs at a similar rate across populations, and whether such activity is affected by participation in a social skills intervention. One method for measuring change in brain activity across a task is analyzing brain activity during the first and second halves of a task separately. In the current study, we sought to understand processes of habituation and sensitization to social stimuli among adolescents with ASD by examining patterns of reward-related neural responses to social versus nonsocial stimuli across a task (e.g., activity in the first versus second half of a task), before and after participation in a social skills intervention. The ERP task utilized was a reward-based guessing game in which participants were presented with rewards accompanied by incidental face or nonface stimuli.

### **2. Methods**

### *2.1. Participants*

Participants included seven adolescents with ASD, and seven age- and gender-matched neurotypical (TD) adolescents. Detailed information about participant demographics can be found in Table 1. No significant differences in age or IQ were observed between groups (*p's* > 0.70).

For both the ASD and TD groups, exclusionary criteria included a history of seizures/epilepsy, a history of brain injury or disease, or a diagnosis of intellectual disability. For the TD group, immediate family history of ASD or developmental disabilities, or any psychiatric diagnosis for the adolescent was exclusionary. For the ASD group, a diagnosis of ASD was required, though commonly co-occurring disorders were not exclusionary (e.g., ADHD). For the ASD group, history of serious psychiatric illness (e.g., schizophrenia, bipolar disorders) or a recent (within 6 months) psychiatric hospitalization were exclusionary.

The study took place in inland Southern California with a large Latinx population [26]. Participant families were recruited via flyers posted online and via local community organizations. Those who expressed interest were contacted for an initial phone screen. At the initial intake appointment, informed consent and assent (from adolescents) were obtained.



### *2.2. Behavioral Intervention (Program for the Education and Enrichment of Relational Skills, (PEERS*®*))*

PEERS® [25,27–29] is a manualized intervention designed to help adolescents make and keep friends (see [30] for intervention details). PEERS® consists of 16 weekly 1.5 h group sessions with concurrent but separate adolescent and parent groups. Parents learn how to support their adolescents in practicing and maintaining skills outside of the group. All groups were run by PEERS® certified providers.

### **3. Measures**

Cognitive abilities were assessed using the 2-subtest Wechsler Abbreviated Scales of Intelligence [31] (WASI-II); an IQ under 70 was exclusionary for both groups. For adolescents with ASD, diagnosis was confirmed using the Autism Diagnostic Observation Schedule, Second Edition [32] (ADOS-2), and motivation to learn how to make and keep friends was assessed using the Mental Status Checklist [25]. Trained study staff performed these assessments. As these measures were used to confirm eligibility, they were only completed prior to the intervention.

### *3.1. Questionnaires*

Data reported here are part of a larger-scale study. Caregivers completed the Social Responsiveness Scale, Second Edition [33] (SRS-2) and the Social Skills Improvement System [34] (SSIS) both before the intervention began (Time 1), and immediately after intervention completion (Time 2). Times 1 and 2 were approximately 4 months apart. Neurotypical adolescents (TD participants) did not receive PEERS®, but had lab visits at Times 1 and 2, where each visit was four months apart. In addition, all adolescents completed the Test of Adolescent Social Skills Knowledge, Revised [27] (TASSK-R) at both Time 1 and Time 2, which measures acquisition of the concepts taught in PEERS®.

### *3.2. Electrophysiology Stimuli and Task*

The stimuli and task are described in detail in previously published manuscripts [12,35,36]. Briefly, the task was a guessing game in which participants saw a left and right visual stimulus (question marks), and were asked to indicate their guess via button press whether the left or right stimulus was "correct." After this choice, the left and right question marks were replaced with an arrow in the middle pointing towards whichever question mark the participant chose. This was done to reinforce the idea that participants had control over the task and their responses were being recorded.

In previously published manuscripts utilizing this task, participants were told that the reward for each correct answer was a small snack; here, the food reward was an Oreo cookie, or if preferred, fruit snacks or goldfish crackers. Participants were told that if they guessed correctly, they would see a ring

of intact Oreo cookies, and the cookies would be crossed out for incorrect answers. There were two blocked feedback conditions: Social versus nonsocial. Importantly, in both the social and nonsocial feedback trials, the face/arrow information was incidental (e.g., the face/arrow image was not part of the overt task). Thus, differences in brain activity between social and nonsocial conditions were not due to differences in tangible rewards or differences in task structure. Incidental stimuli in the social condition were faces obtained from the NimStim database [37] that were smiling for "correct" answers and frowning for "incorrect" answers. Incidental stimuli in the nonsocial condition were composed of scrambled face elements from the social condition formed into an arrow that pointed upwards for "correct" answers and downwards for "incorrect" answers. The order of social versus nonsocial blocks was counterbalanced between participants.

A computer program predetermined correct versus incorrect answers in a pseudorandom order, such that children got 50% "correct" and 50% "incorrect," with no more than three of the same answer-type in a row. The two feedback conditions (face/"social" trials and arrow/"nonsocial" trials) were tested in separate blocks, each composed of 50 trials.

### *3.3. EEG Recording*

Participants wore a standard, fitted cap (Brain Products ActiCap) with 32 silver/silver-chloride (Ag/AgCl) electrodes placed in accordance with the extended international 10–20 system. Continuous EEG was recorded using a Brain Vision Recorder with a reference electrode at Cz, and re-referenced offline to the average activity at left and right mastoids. Electrode resistance was kept under 50 kOhms. Continuous EEG was amplified with a directly coupled high pass filter (DC), and notch filter (60 Hz). The signal was digitized at a rate of 500 samples per second. Eye movement artifacts and blinks were monitored via horizontal electrooculogram (EOG) placed at the outer canthi of each eye and vertical EOG placed above and below the left eye. Trials were time locked to the onset of the feedback stimulus. To measure reward processing, the baseline period was −100–0 ms, and the data were epoched from −100 to 800 ms. Trials with no behavioral response, or containing electrophysiological artifacts, were excluded.

Artifacts were removed via a four-step process. Data were visually inspected for drift exceeding ±200 mV in all electrodes, high frequency noise visible in all electrodes larger than 100 mV, and flatlined data. Following inspection, data were epoched and eyeblink artifacts were identified using independent component analysis (ICA). Individual components were inspected alongside epoched data, and blink components were removed. To remove additional artifacts, we utilized a moving window peak-to-peak procedure in ERPlab [38], with a 200 ms moving window, a 100 ms window step, and a 150 mV voltage threshold.

For both conditions (face, arrow) and both feedback types (correct, incorrect), mean brain activity was calculated between 275 and 425 ms after feedback onset. The reward positivity (RewP) was defined as a difference wave, wherein brain activity in response to "incorrect" feedback was subtracted from brain activity in response to "correct" feedback. For statistical analysis, mean amplitude of the RewP between 275 and 425 ms was utilized. To compare reward-related brain activity during the first half and second half of trials, the first half and last half of all accepted trials (e.g., trials that were not removed through any of the processes mentioned above) were extracted for each of the two conditions (e.g., faces, arrows). Comparing brain activity during the first and second halves of trials allowed us to better understand patterns of reward-related brain activity throughout the task. To be included in statistical analysis, participants had to have a minimum of 6 trials in each half of each condition.

### **4. Results**

All analyses were conducted using SPSS (version 26, Armonk, NY, USA). Prior to analysis, Pearson correlations between ERP amplitude, age, and IQ were conducted. No significant relationships were observed (*p'*s > 0.421).

### *4.1. ERP Results*

An independent samples t-test was conducted to ensure no significant differences in the number of acceptable trials were present between groups (all *p'*s > 0.638).

A 2 (group) × 2 (condition) × 2 (time) × 2 (half) repeated measure analysis of variance (ANOVA) was run. Condition (social, nonsocial), time (pre-intervention, Time 1; post-intervention, Time 2), and half (RewP amplitude during the first and second halves of the task) were within-subjects variables, and group (TD, ASD) was used as a between-subjects variable. A significant 3-way interaction was found between time, half, and group; *F*(12, 20.76) = 5.20, *p* = 0.042, η*<sup>p</sup>* <sup>2</sup> = 0.30. Pairwise comparisons revealed a significant effect of group, such that the ASD group had significantly larger RewP amplitude compared to that of the TD group in the first half of trials at Time 2; *F*(12, 27.04) = 4.83, *p* = 0.048. Thus, regardless of condition, the ASD group had larger reward-related brain activity in the first half of presented trials at Time 2 (post-intervention) compared to that of the TD group. No other significant main effects or interactions were observed. See Figure 1 for grand average waveforms at Time 2.

**Figure 1.** Grand average waveforms during the first and second halves of trials in participants with and without ASD at Time 2 (post-intervention). Significant differences were observed between the ASD and TD groups during the first half of trials at Time 2 (post-intervention). Note that for the purposes of this figure, the ERP was filtered using a 25 Hz low-pass filter. \* *p* < 0.05.

### *4.2. Behavioral Results*

To understand how behavioral measures changed over time for each group, 2 (group) × 2 (time) repeated measure ANOVAs were conducted on measures of autism symptoms (SRS-2), social skills (SSIS social skills subscale), and PEERS®-specific knowledge (TASSK-R).

For the SRS-2, a main effect of group was observed, *F*(1,12) = 9.51, *p* = 0.009, η*<sup>p</sup>* <sup>2</sup> = 0.96, such that the TD group had significantly lower SRS-2 scores than those of the ASD group. Lower SRS-2 scores indicate less severe social impairments. An interaction between group and time approached significance, *F*(1, 12) = 4.56, *p* = 0.054. Post-hoc follow-up tests using Bonferroni corrections revealed a significant difference between groups on the SRS-2 at Time 1 (pre-intervention), such that the TD group had lower scores than those of the ASD group (*p* = 0.001). The difference between the two groups was no longer significant at Time 2 (post-intervention). Pairwise comparisons revealed a trend-level effect of time for the ASD group, such that SRS-2 scores decreased from pre- to post- intervention (*p* = 0.07), whereas no effect of time was observed for the TD group.

For the SSIS social skills subscale, an interaction between group and time approached significance, *F*(1,12) = 4.20, *p* = 0.063. Post-hoc follow-up tests using Bonferroni corrections revealed a significant

effect of time for the ASD group, such that SSIS social skills subscale scores increased from pre- to postintervention (*p* = 0.035), whereas no effect of time was observed for the TD group. Higher scores on the SSIS social skills subscale indicate better social skills. Pairwise comparisons also revealed a trend-level difference between groups on the SSIS social skills subscale at Time 1 (pre-intervention) such that the TD group had higher scores than those of the ASD group (*p* = 0.071), whereas the difference between groups was not significant at Time 2 (post-intervention).

For the TASSK-R, a main effect of group was observed, *F*(1,12) = 5.4, *p* = 0.038, η*<sup>p</sup>* <sup>2</sup> = 0.31, such that adolescents with ASD had higher scores on the TASSK-R compared to neurotypical teens. Higher scores on the TASSK-R indicate more understanding of PEERS®-specific skills. A significant effect of time was observed, *F*(1,12) = 45.82, *p* < 0.001 η*<sup>p</sup>* <sup>2</sup> = 0.79, such that TASSK-R scores increased from Time 1 (pre-intervention) to Time 2 (post-intervention). A significant interaction between time and group was observed, *F*(1,12) = 25.78, *p* < 0.001, η*<sup>p</sup>* <sup>2</sup> = 0.68. Post-hoc follow-up tests using Bonferroni corrections revealed a significant effect of time for the ASD group, such that scores on the TASSK-R increased from pre- to post-intervention (*p* < 0.001). No effect of time was observed for the TD group. Pairwise comparisons also revealed a significant difference between groups on the TASSK-R at Time 2 (post-intervention), such that the ASD group had higher scores on the TASSK-R compared to those of the TD group (*p* = 0.001), whereas the difference between groups was not significant at Time 1 (pre-intervention). Please refer to Table 2 for behavioral measures at each timepoint.

**Table 2.** Behavioral measures for Time 1 and Time 2 in ASD and TD groups.


### *4.3. Brain and Behavior Correlations*

Within the ASD group, Pearson correlations were conducted to examine how change on the behavioral measures from pre- to post-intervention related to ERP results. Difference scores were calculated for the SRS-2, SSIS social skills subscale, and TASSK-R by subtracting post-intervention scores from pre-intervention scores. A significant negative correlation was observed between the SRS-2 difference score and RewP amplitude in the last half of the social condition at Time 1 (*r* = −0.77, *p* = 0.044), such that participants with ASD who had less reward-related brain activity in response to social stimuli at Time 1 (pre-intervention) displayed larger improvements on the SRS-2 compared to individuals with more robust social reward-related brain activity at Time 1. See Figure 2A.

A positive correlation was observed between RewP amplitude in the last half of the social condition at Time 1 (pre-intervention) and SSIS social skills subscale difference score (*r* = 0.78, *p* = 0.038), such that adolescents with ASD who displayed less social reward-related brain activity during the last half of trials in the social condition at Time 1 exhibited greater improvements in social skills from pre- to post-intervention compared to those who displayed more robust reward-related brain activity prior to intervention. See Figure 2B.

**Figure 2.** (**A**) Correlation between SRS-2 difference score before and after intervention in the ASD group and reward positivity (RewP) mean amplitude in the last half of the social condition at Time 1 (*r* = −0.77, *p* = 0.04). (**B**) Correlation between SSIS social skills difference score before and after intervention in the ASD group and RewP mean amplitude in the last half of the social condition at Time 1 (*r* = 0.78, *p* = 0.04).

Finally, a negative correlation was found between the TASSK-R difference score and RewP amplitude in the last half of the social condition at Time 2 (post intervention) (*r* = −0.79, *p* = 0.035), such that participants with ASD who demonstrated larger increases in their knowledge of intervention-specific knowledge displayed larger social reward-related brain activity in response during the second half of trials compared to participants who had smaller increases in intervention-specific knowledge from pre- to post-intervention.

No significant correlations were observed between behavioral measures and reward-related brain activity in the nonsocial (arrow) condition.

### **5. Discussion**

This study investigated the effect of the PEERS® social skills intervention on both neural correlates of reward processing and social behaviors in adolescents with ASD. Specifically, we sought to understand how reward-related brain activity changed throughout the course of a task by comparing brain activity during the first and second halves of trials.

Prior to the start of the intervention, patterns of reward-related brain activity did not differ between participants with ASD and their neurotypical peers. However, after intervention, participants with ASD were more sensitive or responsive to all reward types (both social and nonsocial) during the first half of the ERP paradigm. Increased brain activity related to reward processing indicated increased reward responsivity in adolescents with ASD, irrespective of stimulus type, after participating in a social skills intervention. A larger reward response is similar to what Kohls and colleagues [14] have described as a "liking" response involving the consumption of rewards that are salient. Initial sensitivity to rewards (e.g., during the first half of trials) may have been heightened after exposure to frequent reinforcement strategies that were utilized throughout the intervention to encourage participant engagement.

Although lack of significant differences in brain activity between groups at Time 1 (pre-intervention) is in contrast with some previous intervention literature utilizing neuroscience methods, e.g., [16], and changes in brain activity from pre- to post- intervention in individuals with ASD has been reported previously [17,18,20,21]. Notably, previous research measuring brain activity before and after intervention in individuals with ASD either did not utilize a neurotypical control group, e.g., [17,18,20,21], or had a neurotypical group but did not test children with ASD and the TD group at two timepoints (e.g., pre- and post-intervention for the ASD group). [16,22]. Collecting data from both teens with ASD and their neurotypical peers, as well as utilizing neuroscience paradigms that

are hypothesized to capture changes directly relevant to the intervention itself, are both important strategies when measuring neural correlates of change after an intervention (for a review, see [39]). In the current study, we hypothesized that increased reward-related brain activity would be observed across the course of the ERP task after teens with ASD underwent an intervention that utilized social positive reinforcement principles to increase success in making and keeping friends. To our knowledge, this is the first investigation of brain activity of both neurotypical teens and those with ASD before and after participation in an intervention (or, in the case of the TD group, before and after a delay in which no intervention took place).

Contrary to our hypotheses, brain activity did not differ in response to condition (e.g., social, nonsocial) for either group. This contrasts with previous findings using this paradigm with young children with and without ASD [12,35]. However, this is the first time that this ERP paradigm has been utilized with adolescents. Thus, differences between the current study and previous research might reflect developmental changes. It is plausible that adolescents with and without ASD are less overtly motivated by food rewards as they would be by other reward types (e.g., monetary), and thus may have found the paradigm less engaging/rewarding than younger children. Future studies should consider utilizing this paradigm in a cross-sectional design with different age groups to better understand the effects of age on reward responsivity.

As expected, at Time 1 (pre-intervention), the ASD group had more severe social-communication impairments associated with ASD (measured by the SRS-2) and poorer social skills (measured by the SSIS social skills subscale) than the TD group. Adolescents with ASD improved on both measures after intervention (Time 2), which mirrors previously reported findings of the effectiveness of the PEERS® social skills intervention [29,30]. No differences were observed from Time 1 to Time 2 in the TD group. This was expected, as the neurotypical teens did not participate in the intervention. Importantly, only one ASD participant remained in the range for clinical concern on both the overall SRS-2 score and SSIS social skills subscale score following intervention. This is important as it suggests that change from Time 1 to Time 2 was not only statistically significant, but also clinically meaningful. Further, no significant differences were observed between groups on the SRS-2 or SSIS social skills subscale at Time 2 (post-intervention), suggesting that both social-responsiveness symptoms and social skills in our sample of adolescents with ASD began to resemble social behaviors observed in our neurotypical participants.

One of the most interesting findings of our investigation was that ASD participants who demonstrated less robust social reward-related brain activity in the second half of trials prior to the intervention (Time 1) evidenced the biggest gains from Time 1 to Time 2 in both social responsivity and social skills. This suggests that perhaps the adolescents who benefitted the most from PEERS® were those who had the most "room to improve" in terms of social reward response. This also provides initial evidence that the neural characteristics of reward responsiveness prior to intervention may serve as an indicator of treatment response. That is, it might be possible to utilize neural correlates of social reward responsivity to predict which individuals with ASD might benefit the most from participating in PEERS®. To further investigate this potential predictor of intervention efficacy, future research with a larger sample size and a randomized control group should be conducted.

### **6. Limitations**

This study is part of a larger investigation of a social skills intervention, and this report serves as an initial analysis. Thus, the current study had a small number of participants. It is important to interpret differences in behavioral measures that were approaching significance with caution. Additionally, randomization of treatment was not performed (i.e., a waitlist control group was not utilized) and ASD participants were aware of their enrollment in the social skills intervention (i.e., parent rating forms were not completed "blind," as parents were actively participating in the PEERS® intervention with their teen). Thus, we cannot rule out the possibility that improvements in parent ratings in the ASD group were due to the expectation of improvements. Finally, findings from this study cannot be generalized to all individuals with ASD, as one of the criteria for participation was that the adolescent was motivated to participate in PEERS® and wanted help making and keeping friends. Thus, this sample consisted of adolescents who were highly motivated to learn social skills.

### **7. Conclusions**

The results of our study have important implications for intervention outcomes in adolescents with ASD. First, these findings add to the existing literature on the efficacy of PEERS® for adolescents with ASD. Second, we found evidence for increased reward sensitivity in adolescents with ASD (compared to their neurotypical peers) after participation in the intervention. This suggests that participating in PEERS® increases reward system sensitivity in teens with ASD. Finally, we found that teens who benefitted the most from the intervention (i.e., had the largest gains in social skills and largest decrease in social-communicative impairments) were those with less reward-related brain activity in response to faces prior to the intervention. This relationship between symptom improvement and brain activity prior to the intervention suggests that PEERS® might be most effective for teens with ASD who have "room to grow" in their social reward responsivity, whereas teens with ASD who already have higher levels of social reward responsivity might benefit less. Finally, neuroscience measures may be reliable predictors of teens' responsiveness to treatment because they are independent of potentially biased parent ratings.

**Author Contributions:** K.K.M.S. designed the experiment. E.B. and K.K.M.S. conceptualized the analysis strategy. E.B. performed the EEG processing and statistical analysis under the supervision of K.K.M.S. J.B. verified the analytical methods and interpretations. E.V. and A.M.M. reviewed and confirmed descriptions of methodology. All authors discussed the results and contributed to the final published manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Bezos Family Foundation Grant "Autism, the 'Social Brain,' and Neuroscience: Treating Underserved Latino Teenagers in the Inland Empire."

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Case Report*
