NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Warren Z, Veenstra-VanderWeele J, Stone W, et al. Therapies for Children With Autism Spectrum Disorders. Rockville (MD): Agency for Healthcare Research and Quality (US); 2011 Apr. (Comparative Effectiveness Reviews, No. 26.)

Cover of Therapies for Children With Autism Spectrum Disorders

Therapies for Children With Autism Spectrum Disorders.

Show details


Chapter 3 presents the results of our systematic review. Each category of intervention includes first an overview of the content of the literature as a whole, including the range of study designs used, outcomes assessed and participants included. The summary of the literature provides further discussion and analysis, focusing primarily on those studies that received either a good or fair quality rating. Overview tables document the interventions included, availability of literature by study design, diagnostic approaches, timing of final outcome assessments, geographic location of study populations, and final numbers of participants with autism spectrum disorders (ASDs) for each intervention section (Tables 9, 14, 16, 23, 25).

Table 9. Overview of behavioral studies.

Table 9

Overview of behavioral studies.

Table 14. Overview of educational studies.

Table 14

Overview of educational studies.

Table 16. Overview of studies of medical interventions.

Table 16

Overview of studies of medical interventions.

Table 23. Overview of allied health studies.

Table 23

Overview of allied health studies.

Table 25. Overview of CAM studies.

Table 25

Overview of CAM studies.

Studies that received a good or fair quality rating and include a comparison group (randomized controlled trial (RCT), controlled trial, or prospective or retrospective cohort study) also are described in more detailed summary tables in the relevant section of text. For information on studies not included in the summary tables, please see the evidence table in Appendix C; for information on quality scores for each study, see Appendix H.

Article Selection

Of the entire group of 4,120 citations, 714 required full text review (Figure 2). For the full article review, two reviewers read each article and decided whether it met our inclusion criteria, using a Full Text Inclusion/Exclusion form. Of the 714 full text articles reviewed, we retained 183 papers (comprising 159 unique studies) and excluded 531 papers. Reasons for article exclusion are listed in Appendix D.

Searches retrieved 4.120 nonduplicate citations (3,779 from the literature search and 341 from the hand search/gray literature search). Of these, 3,406 articles were excluded for various reasons. Of the 714 full-text articles reviewed, 531 were excluded. The reasons for exclusion were: participants not in age range (293), not original research (135), ineligible study size (406), not relevent to Key Questions (KQs) (285), unable to abstract data (16). The total number of articles in the exclusion categories exceeds the number of articles excluded because most of the articles fit into multiple exclusion categories. The 183 unique full-text articles included in the review comprised 159 unique studies. The numbers of articles pertaining to each KQ are: KQ1, 154; KQ2, 38; KQ3, 4; KQ4, 01; KQ5, 17; KQ6, 0; KQ7, 4.

Figure 2

Disposition of articles addressing therapies for children with ASDs. a The total number of articles in the exclusion categories exceeds the number of articles excluded because most of the articles fit into multiple exclusion categories; KQ=key question (more...)

KQ1. Effects of Treatment on Core and Commonly Associated Symptoms in Children With ASDs: Behavioral Interventions

A wide range of interventions can be classified as behavioral. For this review, we included studies of early intensive behavioral and developmental interventions, which comprised University of California, Los Angeles (UCLA)/Lovaas-based approaches, the Early Start Denver Model (ESDM), and parent training approaches incorporating principles of Applied Behavior Analysis (ABA) to improve outcomes among young children with ASDs; social skills interventions; focal play-based/interaction-based interventions; behavioral interventions focused on associated behaviors; and a small group of other behavioral interventions assessing other interventions in core/associated areas (e.g., sleep workshops). Table 9 summarizes critical aspects of all studies of behavioral interventions, those addressing key question (KQ) 1 as well as behavioral studies discussed in the KQ2 and KQ7 sections of the report.

Early Intensive Behavioral and Developmental Interventions

Early intensive behavioral and developmental interventions include interventions based on:

  • ABA-based approaches including the UCLA/Lovaas method and variants,
  • naturalistic/developmental principles (i.e., ESDM)
  • parent/family-based training (e.g., Pivotal Response Training, Hanen More Than Words, and social communication training).

We adopted a similar approach to the operationalization of this category as Rogers and Vismara12 in their review of “comprehensive” evidence-based treatments for early ASDs. Interventions in this category all have their basis in or draw from principles of applied behavior analysis (ABA), with differences in methods and setting. ABA is an umbrella term describing principles and techniques used in the assessment, treatment and prevention of challenging behaviors and the promotion of new desired behaviors. The goal of ABA is to teach new skills, promote generalization of these skills, and reduce challenging behaviors with systematic reinforcement. The principles and techniques of ABA existed for decades prior to specific application and study within ASDs.

We include in this category two intensive manualized (i.e., have published treatment manuals to facilitate replication) interventions: the UCLA/Lovaas model and the ESDM. These two interventions have several key differences in their theoretical framework and implementation, although they are similar in the use of high intensity (many hours per week, one-on-one) instruction utilizing ABA techniques. The UCLA/Lovaas method relies heavily on one-on-one therapy sessions during which a trained therapist uses discrete trial teaching with a child to practice target skills, while the ESDMblends ABA principles with developmental and relationship-based approaches for young children.

The other treatment approaches in this section also incorporate ABA principles, and may be intensive in nature, but have not been manualized. We have classified these approaches broadly as UCLA/Lovaas-based given their similarity in approach to the Lovaas model. A third set of interventions included here are those using the principles of ABA to focus on key pivotal behaviors rather than global improvements. These approaches emphasize parent training (e.g., Pivotal Response Training, Hanen More than Words, social pragmatic intervention, etc.) and may focus on specific behaviors such as initiating or organizing activity or on core social communication skills. Because they emphasize early training of parents of young children, they are reviewed here.

We review the results of UCLA/Lovaas-based approaches and parent training approaches focused on pivotal behaviors below; we discuss results of the ESDM in the KQ7 section of the report given the question’s focus on younger children.

Studies focusing on one specific targeted outcome area (e.g., social skills, maladaptive behavior, mental health comorbidities, play) and intervention studies delivered primarily via educational protocols or allied health providers are reviewed in other sections of this report.

Content of the literature. We identified 34 papers100–133 from 30 unique study populations that addressed early intensive behavioral and developmental interventions. A majority of the reviewed literature examined specific early intensive behavioral and developmental approaches, with most using variants of the UCLA/Lovaas model or other ABA-based approaches.101–107,110,111,113–115,118,121,122,124–127,129–133

Four papers evaluated various parent trainings aimed at social communication skills,100,108,109,128 two papers examined Pivotal Response Training,117,120 two studies examined and described eclectic approaches112,119,123 and one study examined a parent training blending Pivotal Response Training and other behavioral approaches (Group Intensive Family Training).116

Summary of the literature. Of the 34 papers in this section100–130 comprising 30 unique studies, 11 were fair, and 19 were poor. Outcomes of RCTs and cohort studies rated fair in quality are summarized in Table 10.

Table 10. Outcomes of early intensive behavioral and developmental interventions.

Table 10

Outcomes of early intensive behavioral and developmental interventions.

Studies of UCLA/Lovaas-based approaches. The one RCT on the UCLA/Lovaas treatment that met inclusion criteria had fair quality.114 This study compared a clinic-based method to a parent program, and targeted children at about 36 months of age. The study114 was the first attempted replication of Lovaas’ manualized intervention to use random assignment, a standardized assessment battery, and explicit accounting of intervention hours. It included 28 children with a mean intelligence quotient (IQ) of 51 randomized to either an intensive treatment group (UCLA/Lovaas model with an average of 25 hours per week of individual treatment per year with reduced intervention over next 1 to 2 years) or a parent-training group (3–9 months of parent training). Gains in IQ were much more tempered than that of Lovaas’ original noncontrolled study.21 Children in the treatment group gained a mean of 15 IQ points in comparison to the relatively stable cognitive functioning of the control group, although average IQ in the treatment group remained in the impaired range. Most of the children who demonstrated large gains in IQ were within the subgroup diagnosed with Pervasive Development Disorder-Not Otherwise Specified (PDD-NOS), whereas children with classically defined Autistic Disorder demonstrated modest improvements.

Two children in the experimental group (vs. one in the control) achieved the “best outcome” or “recovery” status previously defined by Lovaas. No post-treatment group differences were seen in adaptive behavior or challenging behavior. Thus, while replicating improvements in cognitive ability for some children with ASDs within the repeated discrete trial teaching inherent to UCLA/Lovaas method, the study in fact demonstrated a less dramatic impact for the population of children for whom this approach is often recommended (i.e., children with classically defined Autistic Disorder) compared with what was previously reported.

Seven prospective cohort studies and nonrandomized trials were available on UCLA/Lovaas-based methodologies, but none made the same comparisons either in terms of interventions or populations. Hayward and colleagues126,132 examined the progress of children receiving either intensive clinic directed UCLA/Lovaas-based intervention (n=23; mean age=36 months; 37 hours of weekly treatment) or an intensive parent-managed model (n=23; mean age=34 months; 34 hours of weekly treatment) over the course of one year in the United Kingdom. Group assignment was based solely on geographic location. At follow up, both groups had improved significantly in IQ (16 point gain), nonverbal IQ (10 points), language use/understanding, and most areas of adaptive functioning with the exception of daily living skills but there were no differences between the groups.

Two studies compared intensive center-based treatment to community care. Howard and colleagues129 studied preschool-aged children receiving intensive behavior analytic treatment (n=29, 1:1 treatment for 25–40 hours per week), intensive “eclectic” intervention (n =16, higher teacher-student ratio intervention for approximately 30 hours per week), and children receiving general intervention in public early intervention programs (n=16, combined methods, small groups, 15 hours per week). Groups were assigned via educational placement teams that specifically included parent input. Controlling for age at diagnosis and combined parental education, children in the intensive behavior analytic group demonstrated significant improvements in all areas assessed at followup, including an average IQ of 89 (41-point improvement over baseline) and a 24-point difference from the combined mean of the other intervention groups.

Significant differences between the eclectic and generic intervention groups were not present at followup. Findings do suggest substantial improvement via an intensive approach for young children with autism; however, important differences in group assignment at baseline, difficulties with systematic measurement overtime, the lack of reported treatment fidelity or adherence characteristics, and the small number of children in the comparison group limits the interpretation of these findings.

These results were echoed in another study105 of 42 children in which those receiving the Lovaas program had significantly higher IQs (mean=87, gain of 25; mean=73, 14 points) and adaptive behavior skills at outcome, compared with children in undefined community care. Receptive language improvements were observed but were not significant, and expressive language skills and socialization scores on the Vineland Adaptive Behavior Scale (VABS) were not different for the two groups at year 3 outcome. Twelve of the 21 children in the behavioral group had IQs >85 compared with 7 of 21 in the eclectic treatment group at outcome. Likewise, more children in the Lovaas group were in typical schools subsequent to intervention (17 vs. 1); although this specific outcome is potentially attributable to a wide variety of factors including some that might correlate with differences in socioeconomic status and family constellation evident between the groups.

One study125 of two centers compared an eclectic approach (including the Developmental, Individual-Difference, Relationship-Based/Floortime model, Treatment and Education of Autistic and Communication related handicapped CHildren (TEACCH) and ABA-based approaches) to UCLA/Lovaas-based intervention alone. Hours spent in the intervention were consistent at 8 hours per day, and children were assessed over one year. Significant group differences were noted in terms of both language/communication and reciprocal social interaction domain scores on the Autism Diagnostic Observation Schedule (ADOS), with both groups showing decreases in symptom tallies but more substantial decreases in the ABA group. No significant differences in IQ change were reported. While demonstrating impact on certain ADOS symptom scores, these changes were small, and more recent approaches suggest that calculating an ASD severity score may be a more valuable and sensitive way for measuring changes in ASD symptoms in response to intervention.134 In a subsequent study on diagnostic stability124 with unclear sample overlap, most children receiving intervention continued to display scores in the ASDs range on the ADOS (n=53) although some children’s classification did shift.

Finally, one study tried to assess the role of intensity of the intervention on outcomes. Reed and colleagues103 studied the effectiveness of varying intensity of home-based Lovaas-based programs offering primarily one-to-one teaching. High intensity interventions (n=14) were defined as those provided for an average of 30 hours per week. Low intensity interventions (n=13) were provided for on average 13 hours per week. Assignment to the particular intervention modality was based on geographic location, and children in the high intensity group had higher ability and cognitive scores and lower autism severity scores at baseline. Children were assessed 9–10 months after initiation of intervention. Children receiving high intensity intervention demonstrated statistically significant improvements in intellectual and educational functioning from baseline. Children receiving low intensity intervention demonstrated statistically significant changes in educational functioning and nonsignificant improvement in cognitive functioning. The only significant difference between the groups was in improved educational functioning associated with high intensity interventions. No group differences were found in autism severity, cognitive functioning, or adaptive behavior functioning.

Three additional cohort studies101,106,130 of UCLA/Lovaas-based methodologies provided inconsistent data on the benefit of behavioral approaches, but all three had substantial risk of bias and were thereby rated as poor quality in this report. Nonetheless, they suggest that behavioral approaches may have promise for bolstering aspects of cognitive, language and adaptive functioning in preschool children with ASDs.

Case series of early intervention approaches104,113,118,119,131 had mixed results, likely in part due to the substantial heterogeneity of interventions examined even within individual studies, little or no control of concomitant interventions, and poor fidelity to any given approach. Outcomes in these studies were more likely to be parent-reported and not based on validated tools.

Several chart reviews and other retrospective analyses have been used to understand treatment patterns and effects.111,112,115,121–123 Interpretation of findings is most appropriately confined to noting that some children receiving intervention have displayed improvements during intervention in cognitive, adaptive, and autism-specific impairments, that characteristics of starting treatment and baseline abilities are correlated with improvement in some instances, and heterogeneity in terms of improvement is quite common. We do not describe these studies here, but details on all of them are available in the evidence table in Appendix C.

One chart review,122 however, does provide some evidence for the feasibility of providing intensive behavioral interventions on a larger scale as it reviews data on 322 children served in a large service catchment area. Given the methodological limits including lack of a clearly defined intervention characteristics/protocol, lack of a comparison group, retrospective collection, and lack of key measures for certain children at certain times, the intervention results are limited. However, the study suggests the feasibility of providing intensive intervention to a large group of children.

Studies of intensive parent training approaches. Of the seven studies100,108,109,116,117,120,128 on parent training, four100,108,109,120 included comparison groups and had fair100,108,109 or poor120 quality. Three were RCTs,100,108,109 including one pilot study108 with a report of a later implementation of the intervention including different participants.100 Drew et al.109 compared the effects of a home-based, parent-delivered intervention aimed at improving social communication and managing challenging behavior for 12 children with ASDs with a community-based control intervention group of 12 children (mean age 23 months at start of treatment).

Components of the interventions for social communication included developing joint attention, teaching routines, and play activities promoting interaction. Reinforcement techniques, including for alternative behaviors, were used to address challenging behaviors. Training was conducted at home visits (3 hours weekly for 6 weeks), with parents asked to engage in intervention activities for a half to 1 hour daily. One year after treatment initiation, the parent training group reported that their children used more words than the community group. There were no group differences on nonverbal intelligence quotient (NVIQ), autism symptom severity, or words/gestures observed during followup assessment. Unexpectedly, the treatment group lost IQ points during the study; whereas the control group demonstrated relatively stable cognitive abilities. This finding is further confounded by a significantly higher IQ present in the treatment group at initiation of the study.

Aldred et al.108 compared a parent-based intervention focused on advancing social communication skills within interactions (n = 14, median age 51 months) to treatment as usual (n=14, median age 48 months). Parents participated in initial workshops, monthly intervention sessions where videotaped interactions were reviewed, and 6 months of maintenance visits (approximately once every 2 months). Twelve months after baseline, blinded evaluations showed improvements on ADOS scores, with substantial improvement within the social domain, increased expressive vocabulary, as well as improved communication-related behaviors coded during interactions. Language gains were most prominent in younger, lower-functioning children. A lack of standardized measures of developmental performance, including baseline cognitive skills, as well as challenges in understanding and defining “treatment as usual” limit interpretation of the findings.

In a report of a later intervention of this model, 152 children between the ages of 2 and 4 years were randomized to treatment as usual or treatment as usual plus parent training in social communication.100 Time in “treatment as usual” interventions was similar across groups as were the types of interventions employed. Similar numbers of children in both groups experienced diagnostic shifts from core autism to other diagnoses on the ASDs spectrum as diagnosed on the ADOS-G. Teacher ratings of language and communication after intervention were not significantly different between groups, though ratings of parent-child interactions by independent assessors were positive for children in the social communication group. Parent ratings of language and social communication were also more positive for the social communication group.

Stahmer and Gist120 examined the effects of an explicit parent education support group with a parent education program focusing on Pivotal Response Training, a treatment program designed to enhance core skill areas in autism using naturalistic interactions. Parents met with the intervention provider weekly for 12 weeks and were taught techniques for presenting clear instructions, following and supplementing child choice, and providing direct/naturalistic reinforcement. Involvement in the 12-week intervention was successful in changing parenting techniques and perceived language gain. However, the lack of randomization, wide variation in children served, the lack of objectively assessed changes in child behavior, and the small number of participating limit the reported results.

Social Skills Interventions

The social interventions reviewed in this section focus primarily on children at elementary-school ages and those functioning at higher cognitive/developmental levels. They use various approaches to address three primary dimensions of social competence: specific behavioral skills (e.g., greetings, initiating game play, joint attention), affective understanding (e.g., recognizing emotions in self and others), and social cognition (e.g., theory of mind, problem-solving, self-regulation).

Content of the literature. We located 16 unique papers addressing social skills interventions. This number includes two sets of papers with possibly overlapping samples evaluating a Skillstreaming intervention135,136 and a cognitive-behavioral-ecological social skills approach.137,138 The ages of children studied ranged from 4–16 years old. Twelve studies focused exclusively on higher functioning children or included language and/or cognitive requirements among their eligibility criteria.135–146 Three studies provided individual treatment to children,137,145,147 three used a combination of individual and small group formats,138,146,148 and nine employed a small group format only.135,136,139–144,149 In addition, five interventions included some form of parent training or involvement as an adjunct to child treatment.137,139,141,144,146 For the 14 studies with prospective designs, the total amount of training provided ranged from 6.7 hours to 180 hours. Table 10 summarizes additional details.

Among studies of social skills interventions, seven were fair quality and nine were poor.

Summary of the literature. Three RCTs139,141,146 (Table 11) evaluated social skills interventions targeting high functioning children with ASDs using a format that involved training for both children and their parents. The criteria for determining whether a child was high functioning and therefore eligible to participate varied by study, but at a minimum the child had to have a verbal IQ above 60. Different outcome measures were used across the samples, making direct comparisons difficult.

Table 11. Outcomes of RCTs of social skills behavioral interventions.

Table 11

Outcomes of RCTs of social skills behavioral interventions.

The Children’s Friendship Training141 program involves children with and without ASDs, and uses didactic instruction on rules of social behavior; modeling, coached behavioral rehearsal, and performance feedback during treatment sessions; rehearsal at home; homework assignments; and coaching by parents during play dates with a peer. Children were randomly assigned to receive Children’s Friendship Training either immediately or 12 weeks later (Delayed Treatment Control group). Treatment was conducted in 60-minute small parallel group sessions for parents and children, and lasted 12 weeks.

Immediately following treatment, the Children’s Friendship Training group showed significant improvements in social behavior and social cognition compared with the Delayed Treatment Control group. Children in the treatment group also spent less time during the play date engaged in minimally socially interactive activities (such as watching television) compared with the delayed treatment group (p<0.001), but did not spend significantly more time in socially interactive activities (e.g., talking). Parents of children in the Children’s Friendship Training group reported that their children demonstrated increased self-control when provoked by others relative to the control group (p<0.05).

Parent- and teacher-reported reductions in social withdrawal showed nonsignificant changes. Children in the treatment group self-reported decreased loneliness (p<0.025) and increased popularity (p<0.025) relative to the control group. Three months post-treatment significant improvements were maintained in the treatment group on parent reported hosting of play dates, conflict during play dates, time spent in minimally socially interactive activities, assertion, self-control, and social withdrawal compared with the baseline scores. After treatment, findings from the Delayed Treatment Control group largely replicated those of the Children’s Friendship Training group.

Relative to Children’s Friendship Training, the Social Adjustment Enhancement Curriculum has a more comprehensive curriculum targeting emotion and facial expression recognition; theory of mind, the ability to ascribe mental states to oneself and others to understand and forecast behavior; perspective taking; executive functioning, which allows for planning and abstract thinking; problem solving; and conversation skills. Eighteen boys between 8 and 12 years old met eligibility criteria.139

Participants were matched on age and IQ and randomly assigned to an immediate intervention condition or a wait list condition. Parents and children in the treatment condition received the Social Adjustment Enhancement Curriculum at a clinic for 20 weekly 1.5 hour sessions. Children and parents met separately. Child groups of four or five were structured with a high adult-to-child ratio and followed a consistent schedule each week, using a variety of instructional strategies including in vivo teaching, visual templates, games, and role playing.

Immediately following the intervention participants in Social Adjustment Enhancement Curriculum had higher facial recognition scores post-treatment (p<0.05), while the scores of the participants in the wait list control group declined (although not significantly). There was significantly improved executive function skills (covarying Verbal IQ) post-treatment (p<0.05) in the intervention group, while the scores of those in the wait-list control declined. However, when the one child with a PDD-NOS diagnosis was excluded from the treatment group these results were no longer significant. Both the control and Social Adjustment Enhancement Curriculum group demonstrated significant improvements on the Faux Pas Stories Task post-treatment (p<0.001) but not on the Strange Stories Task. Total social problems reported per time reporting dropped significantly from the first eight weeks of the intervention to the last eight weeks of the intervention (p<0.05).

Beaumont and Sofronoff146 investigated a comprehensive social skills intervention that utilized a computer game as well as child and parent small therapy groups to teach emotion recognition and regulation, problem solving, and social interaction skills. Forty-nine children diagnosed with Asperger syndrome between the ages of 7.5 and 11 years old were randomly assigned to the Junior Detective Training Program or a wait list control. Data suggests that this computerized intervention was beneficial for improving knowledge of emotion management strategies and parent-reported social skills, but not emotion recognition, immediately after treatment over a the waitlist control.

Some of these results were also replicated when the wait-list group underwent treatment in pre-post analyses. Pre- post scores were significantly different on the parent-reported measures of social skills for the intervention group immediately, 6 weeks, and 5 months following the intervention, suggesting maintenance of these treatment effects. However this study had substantial risk of bias and was rated as poor quality in this report. Additional details on this and other studies not fully described in this section are available in the evidence table in Appendix C.

One study attempted to ascertain whether the type of feedback that children received during a social skills intervention affected the outcomes. Skillstreaming150 is a comprehensive, structured social skills curriculum that employs systematic procedures for teaching specific social behaviors (e.g., listening, sharing, having a conversation, accepting a compliment, responding to teasing), as well as social cognition (using self-control), and affect (e.g., recognizing and expressing feelings, responding to anger). The Skillstreaming curriculum used in the study was adapted to focus on social skills particularly important for children with autism.

Unlike in the previous RCTs reviewed in this section, this intervention did not include a parent training component. Fifty-four children between the ages of 6 and 13 years with high functioning ASDs were randomly assigned to small-group Skillstreaming intervention that used either a response-cost condition (involving immediate performance feedback and rewards based on specific social skills and behaviors) or a noncategorical feedback condition (involving more general feedback and noncontingent rewards). The only difference found between the response cost and noncategorical feedback intervention conditions post-treatment was that interventionists reported significant improvements measures of atypicality, withdrawal, and behavior symptoms in the response cost group relative to the noncategorical feedback group (p<0.05).

However, both groups combined made significant improvements after treatment on both parent and interventionist reports of social skills, withdrawal, adaptive skills, and behavior symptoms (p<0.001--p<0.05). As for facial recognition, participants in neither group made significant improvements on the Diagnostic Analysis of NonVerbal Accuracy. A previous case series to assess Skillstreaming136 for 21 children between 6 and 13 years old diagnosed with Asperger disorder found similar results: significant improvements in parent reported social skills, adaptability, and atypicality) and on staff reported social skills.

Three RCTs143,145,149 and a related retrospective cohort study148 evaluated social skills interventions focused on improving children’s ability to socially interact with others while playing. The Quirmbach et al.145 study evaluated the effectiveness of using Social Stories to teach seven to 14 year old children with ASDs social skills when playing board games. Social Stories151 are descriptive brief vignettes constructed according to a specific formula that are read to or by individuals with ASDs to convey appropriate behavior expected for a specific situation. Children in one of two social stories groups (standard or directive) showed significant game play skill improvements across the four trials (p<0.001) while the children who received the control story did not. Children in the two experimental conditions maintained the results of the intervention a week later.

These results provide preliminary support for the effectiveness of a short, focused intervention on improving the specific targeted skills. However, further research is needed to ascertain whether these results generalize to other people (such as peers) and other settings, whether these results are maintained when the intervention is discontinued (i.e., the child stops reading the Social Story), and whether other Social Stories are successful at improving the skills that they target. Despite authors’ predictions that the children who read the directive story would improve their game play skills at a faster rate than the children who read the standard story (because the directive story does not include additional information), there were no significant differences in results between participants receiving the standard vs. directive social stories (both groups showed significant improvements across trials).

In the other studies, direct teaching was associated with greater gains in initiating, responding, and interacting behaviors than an unstructured play group,149 in 4 to 6 year olds. Results on LEGO therapy were conflicting, with one a retrospective cohort study148 showing benefit for LEGO therapy over an unspecified “other” intervention on socialization measures, while one RCT143 had inconsistent results on the benefit of LEGO therapy over a Social Use of Language Program and no intervention. The Lego group improved on measures of social skills when compared with the Social Use of Language Program and control groups; and pre-post scores did not significantly differ on these measures for the Lego group. Both the Lego and Social Use of Language Program groups improved on measures of maladaptive behavior over the no intervention group. The Lego group improved in the duration of social interaction on the playground from pre to post treatment.

Seven additional studies used prospective case series designs to evaluate the effectiveness of social skills interventions.137,138,140,142,144,147,152 All studies noted improvements in some social behaviors that, depending on the study, included eye contact, emotion recognition, and interaction with peers; outcome measures were generally parent-reported. The studies also lacked control groups so it is difficult to determine whether improvements are treatment-specific.

Play-/Interaction-Based Interventions

These interventions focused on children’s interactions with either their parents or experimenters and targeted skills including joint attention and play abilities. Most studies were conducted in the context of a play situation, and included children across broad age and developmental ranges.

Content of the literature. We included 15 papers addressing play- or interaction-based interventions153–167 comprising 13 unique populations. Seven studies were randomized controlled trials,153–157,161,163–165 including a trial of the Stepping Stones Triple P program with two publications,153,154 and a trial comparing joint attention and symbolic play interventions with two publications.155,156 Two additional studies assessed joint attention and symbolic play and likely share overlapping participants with this trial.157,158 Three RCTs assessed comparable interventions (imitation compared with contingent responsiveness) using similar procedures;163–165 two of these163,165 may share participants.

Multiple interventions involved parent training or parent interaction components, including Parent-Child Interaction Therapy,161 responsive teaching,166 play-based approaches based on the Floortime model,160 the Mifne model,162 the Stepping Stones Triple P program,153,154 and the Relationship Development Intervention program.159 Participants ranged in age from 12 months166 to 12 years161 across all studies. Duration of therapy in prospective studies ranged from five weeks156,157 to 12 months.160,166 Table 10 includes additional study details. Among the 13 unique studies, three were fair quality and 10 were poor.

Summary of the literature. Among the fair quality studies was an RCT evaluating Parent-Child Interaction Therapy,161 in which parents of children with an ASD were trained to interact with their children using behavior management strategies (Table 12). The Parent-Child Interaction Therapy intervention group consisted of ten children and the wait-list control group included nine; children in both groups were on average 8 years old. Improvements were greater in the intervention group in challenging behavior, behavioral flexibility and atypical behaviors, and hyperactivity, inattention, challenging behaviors, and depression ratings. However changes on each scale fell short of statistical significance in comparison with the control group. A second parent-focused RCT addressed the Stepping Stones Triple P Parenting Program,153,154 which focuses on managing children’s behavior by considering the function of the behavior and uses procedures such as descriptive praise, planned ignoring, skill acquisition, and communication.

Table 12. Outcomes of RCTs assessing play-/interaction-based interventions.

Table 12

Outcomes of RCTs assessing play-/interaction-based interventions.

Parents of the children in the treatment group reported statistically significant decreases in child challenging behavior on the Eyberg Child Behavior Inventory Intensity and Problem Scales. Wait-list controls eventually received the same treatment, and parents of children in this group also reported statistically significant decreases in child challenging behavior on both Eyberg Child Behavior Inventory scales. At six-month follow up, the treatment group maintained gains on both the Eyberg scales.

The additional studies in this section included three RCTs that compared the effects of imitation and contingent responsiveness.163–165 Contingently responsive behavior refers to the adult responding to the child’s initiations by either commenting back or gesturing within the play context. In the first phase, the child entered the room with an adult present holding a neutral facial expression. During Phase 2, the adult interacted with the child by using either imitation or contingently responsive behavior in response to the child’s behavior. The third phase mimicked Phase 1, and the fourth and final phase included a spontaneous play interaction. Each of these four phases was three minutes in duration.

Each of the three RCTs included 20 children randomly assigned to either the imitation group or the contingently responsive group,163–165 Significantly greater effects were seen in the imitation group compared with the contingent responsiveness groups in all three studies. Improvements included spending more time engaged with both objects and adults,163 a greater reduction in motor activity,165 and more social interest.164

Two RCTs,155–157 one of which was fair quality,155,156 and two case series158,167 focused on the potential for interventions based on joint attention or symbolic play. Generally speaking, interventions with a joint attention focus did result in improvements in tasks based on joint attention. In the first RCT,157 all groups improved in coordinated joint looks over time. No differences were found in pointing to a toy or giving a toy to an adult to share in any group. Both Joint Attention and Symbolic Play groups improved in the following areas compared with controls: showing toys to an adult, shared looks between a toy and the child’s mother, and symbolic play skills.

Compared with other groups, the Joint Attention group showed more improvement in responding to joint attention over time. With respect to mother-child interactions (generalization) assessing the same outcome areas, the Joint Attention group had significantly greater improvement than the Symbolic Play group in giving and showing a toy. Children in the Joint Attention group engaged in more child-initiated joint engagement than those in the control group. The Symbolic Play group showed significantly greater improvement on the Structured Play Assessment than did the control group for overall mastered level of play. In the second RCT,157 significantly more children in the Joint Attention group engaged in coordinated looks during the final stimulus presentation (76.5 percent) than in the Symbolic Play group (38.9 percent). Children in the Joint Attention group engaged in significantly longer periods of coordinated looks between the person in the room and the stimulus presentations across the three time periods.

A second RCT155,156 comparing joint attention and symbolic play interventions included 58 children with autism between 3 and 4 years of age. Investigators assessed language development, joint attention and play skills, and mother-child interactions at pre- and post-intervention and 6 and 12 months after the end of the 5 to 6 week intervention. Children in both groups showed significantly greater growth in expressive language, initiation of joint attention, and duration of child-initiated joint attention over time than did participants in the control group (p<.01 to <.05, moderate to large effect sizes). Growth in receptive language was not significantly affected by the intervention from pre-intervention to 12 months post-intervention. Children in the Symbolic Play group also showed significantly more growth in play level than did children in either the Joint Attention (p<.01) or control (p<.001) groups.

In a fair quality case series describing an eight week, 24-session intervention designed to foster joint attention and language skills as well as joint engagement with the mother,167 episodes of distress occurred in an average of 9.4 sessions (range=four to 24 sessions), with children displaying negativity for an average of 20 percent of the time (range=6–52 percent). There were no associations between negativity and children’s mental or chronological age. Both mothers and children showed improvements in behavior regulation over the course of the intervention. Children engaged in behavioral strategies significantly more often during episodes of negativity than in nonnegative episodes (p<.01). The study also reports associations between mothers’ vocal behavior regulation strategies and child-related stress as reported on the Parenting Stress Index; mothers with greater child-related stress used fewer vocal strategies such as vocal comfort and reassurance. Mothers whose children exhibited more externalizing problems (as rated on the Child Behavior Checklist) used more active behavior regulation strategies (e.g., shifting child’s attention away from negative stimulus, hugging child, etc.).

None of the four additional case series that met criteria for inclusion in this section described the same intervention. They described a relationship-focused intervention teaching parents to use responsive teaching strategies to assist their children with acquiring pivotal behaviors,166 the Developmental, Individual-Difference, Relationship-Based/Floortime model,160 the Relationship Development Intervention, 159 and the Mifne treatment model.162 All four report positive outcomes that are difficult to interpret absent a comparison group.

Behavioral Interventions Focused on Associated Behaviors

Several behavioral interventions target symptoms commonly associated with autism, such as anxiety and anger management. Cognitive behavioral therapy-based (CBT) interventions are particularly common and involve teaching cognitive skills and relaxation strategies, promoting recognition of anxious feelings, and providing children with behavioral exposures in which to utilize their new coping skills in the face of anxiety-provoking stimuli, with an ultimate goal of reducing anxiety symptoms over time.29

Parent training protocols, often implemented to help parents deal with challenging behaviors such as noncompliance, tantrums, self-injury, and aggression, attempt to teach parents strategies to curb negative behaviors. Once trained, parents can act as “co-therapists,” shaping behavior toward the goal of reducing challenging behaviors in daily life, where parents by necessity must act as the primary interventionist. Parent training interventions also often have secondary targets of improving parental feelings of self-efficacy and decreasing parental stress.

Many of the studies of behavioral methods used to treat challenging behaviors, such as functional behavior analysis and positive behavior support, included fewer than 10 participants with ASDs and thus were not included in this review.

Content of the literature. We identified 11 studies reported in 12 papers25,26,168–177 that addressed behavioral interventions focused on symptoms commonly associated with ASDs including anxiety and anger management. Six studies reported on CBT interventions,25,26,168–171,176 four used parent training techniques,172,174,175,177 and one used teacher training methods.173

While the overlap among studies is somewhat unclear, sets of studies from the same authors and using the same methodology appear to include the same or overlapping samples.169,176;170,171;174,175 Accounting for this potential overlap, it appears that at least four independent studies of CBT interventions and three independent studies of parent training address anxiety and anger in children with ASDs. All studies examining CBT treatments included children ages seven and older, with means ranging from nine to eleven years of age. In two studies examining CBT treatments, only children with an Asperger diagnosis were included,169,176 while the Wood et al. RCT enrolled children with an ASD and a comorbid anxiety disorder.170,171

Parent training studies included parents of children ranging from age four to twelve with mean ages spanning seven to nine years.172,174,175,178 In three of four parent training studies, only parents of children with Asperger syndrome were included.174,175,178 In the teacher training study, children ranged in age from two to fifteen and all had diagnoses of autistic disorder.173 Table 10 summarizes additional study details. Among all studies, six were fair quality and five were poor.

Summary of the literature. Among the studies assessing CBT approaches, one RCT examined the efficacy of a modified version of the Building Confidence CBT program for treating comorbid anxiety disorders (i.e., separation anxiety disorder, social phobia, or obsessive compulsive disorder) in seven to eleven year-old children with ASDs.170,171 This was the only RCT in which CBT occurred at the individual level.

The intervention program consisted of sixteen 90-minute weekly sessions conducted by clinical or educational psychologists or trainees in these programs. In the first report from the study,170,171 anxiety symptoms were assessed by evaluators blind to treatment condition using the Anxiety Disorders Interview Schedule, Clinical Global Impression (CGI)–Improvement Scale, and both parent and child versions of the Multidimensional Anxiety Scale for Children.

On the CGI, 92.9 percent of children in the intervention condition met criteria for positive treatment response, while only 9.1 percent of children in the waitlist control group met the same criteria; on the Anxiety Disorders Interview Schedule, 64.3 percent of children in the intervention group no longer met criteria for any anxiety disorder, whereas only 9.1 percent of children in the waitlist control group lost their anxiety disorder diagnosis at post-test.

Eight of ten children from the intervention group who returned for a three-month followup did not meet criteria for any anxiety disorder at followup. Maintenance of treatment response was also indicated by CGI and Multidimensional Anxiety Scale for Children scores at followup. The second report from the study171 included 58 percent of participants from the initial report (42 percent new participants), and measured effects of the intervention on autism symptoms using the Social Responsiveness Scale. Significant group differences were observed at outcome in the Social Responsiveness Scale total score as well as the social communication, social motivation, and social awareness subscales, with children in the intervention group showing fewer autism symptoms post-treatment than children in the waitlist control group.

The remainder of CBT-based interventions (Table 13) were conducted in group settings or directed toward parents. Reaven et al.26 conducted a nonrandomized trial of a 12-week CBT-based group intervention for high-functioning (i.e., IQ above 70) children ages eight to fourteen years (mean = 11.83) with ASDs and comorbid anxiety disorders.

Table 13. Studies assessing interventions targeting conditions commonly associated with ASDs.

Table 13

Studies assessing interventions targeting conditions commonly associated with ASDs.

The authors created an original protocol,27 and treatment involved both children and their parents. Ten children received active treatment in this pilot study, while 23 served as a wait-list control. Anxiety symptoms in children participating in the treatment group decreased over time, while symptoms in the control group did not on the parent (but not child) version of the Kiddie-Schedule for Affective Disorders and Schizophrenia or on the Screen for Child Anxiety and Related Emotional Disorders.

Chalfant et al.25 examined children ages eight to thirteen years (mean = 10.8) with ASDs and one or more comorbid anxiety disorder diagnoses including separation anxiety, generalized anxiety, social phobia, specific phobia, and panic disorder confirmed by structured clinical interview using the Anxiety Disorders Interview Schedule. Children were randomly assigned to treatment and waitlist conditions. Treatment involved a 12-session CBT-based group therapy protocol, led by licensed clinical psychologists, with nine weekly two-hour sessions followed by three monthly booster sessions.

The protocol for the study was based on a manualized CBT-based anxiety intervention for children (Cool Kids) with adaptations made to account for the learning style of children with ASDs (e.g., more visual aids and structured worksheets, increased focus on relaxation and exposure, simplification and decreased emphasis on cognitive components of the treatment). Parents of children in the intervention group participated in concurrent parent groups with a manual also adapted from the Cool Kids program.

Measures were collected at baseline and at the completion of intervention (approximately five and a half months later); clinicians administering the pre- and post-intervention measures were the same clinicians who led treatment groups. No group differences were observed on any measure at baseline. However, children in the treatment group improved significantly over time while children on the waitlist did not in the number of anxiety disorder diagnoses present, as well as in the number of anxiety symptoms reported by children on the Children’s Automatic Thoughts Scale Internalising Scales, Revised Children’s Manifest Anxiety Scale, and Spence Children’s Anxiety Scale, by parents in their report on the Spence Children’s Anxiety Scale—Parent and the Strengths and Difficulties Questionnaire Emotional and Externalizing Scales, and by teachers using the Strengths and Difficulties Questionnaire Emotional and Externalizing Scales.

A series of papers examined CBT approaches delivered directly to children and via parent training. CBT provided by graduate students in psychology was assessed in high functioning children with Asperger disorder with comparisons made across two intervention conditions (child-only and parent-plus-child) and waitlist controls.169,176 Significant improvements in Spence Child Anxiety Scale-Parent scores were observed for both intervention groups on the total score and separation anxiety, obsessive compulsive disorder, social phobia, panic, and generalized anxiety scales; significant improvement on the personal injury scale was observed for the parent-plus-child intervention group. No significant differences were observed from baseline to six-week followup in the waitlist control group.

On the Social Worries Questionnaire, there were significant improvement in scores between baseline and six-week followup observed for both intervention groups, but not for the waitlist control group. Similarly, children in both intervention groups generated more strategies to cope with anxiety at six-week followup than at baseline, while children in the waitlist control group did not. Both intervention groups scored better than the wait-list control group at followup, and children in the parent-plus-child condition scored better than children from the child-only groups at followup. A separate study of the same intervention168 to examine the impact of CBT-based treatment on anger management difficulties in high-functioning (i.e., average IQ above 100) children ages ten to fourteen years with Asperger syndrome found similarly positive results.

Parent training in using CBT approaches174,175 for parents of children ages six to twelve years with Asperger syndrome diagnoses includes psychoeducation, comic strip conversations and social stories introduction, and management techniques for externalizing behaviors, rigid behaviors, and anxiety. In studies of this approach, parents who attended a one-day workshop or who participated in six weekly one-hour individual sessions reported fewer challenging behaviors at both one-month post-treatment and three-month followup relative to baseline, while there were no significant differences over time for the waiting list control group.

Parents from both intervention groups also reported significantly fewer challenging behaviors in their children, decreased challenging behavior intensity and improved social skills at both time points. At three-month followup, individual session participants reported significantly lower intensity of challenging behaviors relative to both the waiting list control group and workshop intervention group; the workshop group no longer showed differences from the control group by three-month followup in terms of parental report of child challenging behavior intensity.

The Research Units on Pediatric Psychopharmacology (RUPP) Autism Network first reported on the feasibility of a parent-training program for parents of children with autism spectrum disorders ages 4 to 13 years who were on stable medications for behavior problems.172 The parent training protocol consisted of 11 required sessions covering topics including prevention strategies, schedules, reinforcement, planned ignoring, compliance training, functional communication training, teaching techniques, and generalization. Two home visits were always conducted, four optional sessions were available, and booster sessions were provided to parents in later weeks; parent training was administered according to a structured curriculum.

Outcome measures related to child functioning included the Home Situations Questionnaire, the Aberrant Behavior Checklist (ABC), the Clinical Global Impressions – Improvement Scale (CGI-I), the VABS, and the Assessment of Basic Language and Learning Skills. Results indicated significant decrease in noncompliance on the Home Situations Questionnaire over the course of parent training intervention. Irritability and Hyperactivity/Noncompliance measured on the ABC showed improvement over time. Fifty-three percent of children were reported as very much improved on the CGI-I, and thirty percent were reported to show minimal improvement. Finally, with regard to adaptive behavior, improvement in daily living skills and increase in adaptive skills were also shown over the course of treatment. The correlation between improved compliance and improved adaptive behavior also strengthened over the course of treatment.

Following the initial feasibility study,172 Aman et al.177 conducted an RCT assessing whether risperidone treatment combined with parent training in behavior management was superior to risperidone treatment alone; this study was conducted as part of the RUPP Autism Network. Parents of children ages 4 to 13 years with ASDs and significant tantrums, self-injury, and aggression who were randomly assigned to the combined treatment group received parent training with a behavior therapist according to a RUPP manual. As noted, the manual specified 11 core treatment sessions, three optional sessions, and up to three booster sessions of 60–90 minutes in length.

On average, parents in the combined condition participated in 10.82 sessions. As in the feasibility study, outcome was assessed on the Home Situations Questionnaire and ABC; the Children’s Yale-Brown Obsessive Compulsive Scale–PDD version was also administered both pre- and post-intervention. After 24 weeks of treatment, HSQ scores for 71 percent of children assigned to the combined treatment group and 60 percent of children assigned to the medication-only treatment group declined (i.e., decreased severity), which represents a significant difference between groups over time. In addition, the ABC irritability, stereotypic behaviors, and hyperactivity/noncompliance subscales all showed significant group differences over time, with children of parents who received the parent training showing less severe symptoms in each of the domains.

One case series of a teacher training procedure in reducing challenging behaviors (e.g., aggression, noncompliance, off-task behavior)173 reported a significant reduction in the rate of the target behavior following classroom instruction (45 hours) and practical application, training, and supervision (45 hours) in applied behavior analysis.

Other Behavioral Interventions

Additional behavioral interventions include techniques such as neurofeedback and sleep hygiene education. Neurofeedback, or electroencephalogram (EEG) biofeedback, aims to remediate abnormal brainwave activity associated with disorders such as anxiety, ADHD, and ASDs through training individuals to control brain activity patterns. Neurofeedback involves the placement of electrodes to monitor brain activity while participants interact with specially designed computer games or other modalities designed to promote attention or other skills.179 Behavioral treatments for sleep problems may attempt to affect the timing of sleep, sleep-wake cycle disorders, or promote efficacious sleep behaviors like bedtime routines and positive reinforcement.180

Content of the literature. Three studies181–183 of additional behavioral interventions met our inclusion criteria. Participant ages ranged from three to fourteen years across the studies, and all three occurred in a clinic setting. Jarusiewicz181 and Coben et al.182 used neurofeedback with children directly while the Reed et al. sleep workshops were aimed at parents using a group approach.183 Table 10 includes additional study details. All three studies in this section were considered poor quality.

Summary of the literature. Jarusiewicz’s RCT examined the efficacy of neurofeedback on autistic symptoms as assessed using the parent-rated Autism Treatment Evaluation Checklist as the primary outcome measure.181 Participants included 40 children ages four to 13 (mean=7) with a previous diagnosis of autism; participants were matched on gender, age, and autism severity, and individuals in each pair were randomly assigned to either neurofeedback or a wait list. Diagnostic and randomization procedures were not described.

Neurofeedback protocols varied depending on a child’s autism severity as assessed by the Autism Treatment Evaluation Checklist and parental report of troubling symptoms; children typically received one to three sessions per week. Eight children in the neurofeedback group dropped out of the study due to family considerations or non-ASDs-related illness; the twelve remaining participants completed 20 to 69 neurofeedback sessions (mean=36). Scores on the Autism Treatment Evaluation Checklist improved from eight percent to 56 percent post-neurofeedback training, with an overall average reduction of 26 percent (p <0.001). Scores for control participants improved by three percent overall (ns).

Coben et al.182 sought to extend Jarusiewicz’s findings in a nonrandomized controlled trial of 49 children diagnosed with ASDs; diagnostic techniques were not described. Children in the experimental (n=37) and control (n=12) groups were matched on age, gender, handedness, ASDs severity, and other treatments received. Participants’ ages across groups ranged from 3–14 years; the majority (75 percent) of participants in the neurofeedback group were diagnosed with PDD-NOS or autism. Four children in this group had Childhood Disintegrative Disorder. Outcome measures included the Autism Treatment Evaluation Checklist, Gilliam Autism Rating Scale, Gilliam Asperger Disorder Scale, the Personality Inventory for Children, Behavior Rating Inventory of Executive Function as well as parental ratings of the effectiveness of the treatment and a battery of neuropsychological tests to assess attention, visual-perceptual, executive function, and language skills.

Neurofeedback protocols were individualized for each child based on assessment information and initial quantitative electroencephalogram results; treatment consisted of 20 sessions, with sessions occurring twice weekly for an unspecified duration. Eighty-nine percent of parents reported improvement in the neurofeedback group; 83 percent of control group parents reported no change (z=2.167, p=0.000). Scores on all measures except the Gilliam Autism Rating Scale, improved significantly in the treatment group (p=0.000 to 0.006) as did scores on neuropsychological tests. The authors did not correct for multiple testing, however.

Reed at al.183 employed sleep hygiene workshops targeted at parents and taught by a neurology sleep specialist, pediatrician with an ASDs treatment background, educational psychologist, and nurse educator. Twenty families participated and completed both baseline and followup assessments; the mean age of children in the study was 5.8±2.7 years, and the majority (n=15) had ADOS scores in the autism range. Workshops addressed establishing effective daytime/nighttime routines, minimizing night and early waking, and discussion of techniques to handle individual sleep concerns. Assessments, conducted prior to the initial workshop and approximately one month after the final session assessed sleep changes, repetitive behavior, and parental stress and also included a week of actigraphy measurement of sleep-wake patterns coupled with a parent-maintained sleep diary.

Significant improvements (P<0.05) over baseline scores were seen in subscales of measures assessing hyperactivity, sleep disturbance, self-stimulatory, bedtime resistance, sleep onset delay, sleep duration subscales, and restricted behavior. Items related to reduction of stimulating activities before bedtime and the use of bedtime routines also improved. Actigraphy data, available for 12 children, illustrated a decrease in sleep latency in nine children with difficulty initiating sleep (p=0.039); among all 12 children, time in bed also significantly improved (p=0.039). Parental stress did not change significantly with the workshops.

KQ1. Effects of Treatment on Core and Commonly Associated Symptoms in Children With ASDs: Educational Interventions

Most children with ASDs will receive some treatment in an educational setting. Educational interventions have focused both on traditional areas of academic progression/achievement, but also are often used to address core areas of social, cognitive, and behavioral vulnerability via classroom or specialized instruction. Educational approaches vary in scope from specific intervention techniques attempting to impart change in short periods of time to comprehensive intervention programs aimed at improving many areas of development.

Within the context of this review we examine the available literature focusing on applications of the TEACCH program, broad-based early intervention center- or classroom-based instruction, and computer-based approaches to educational intervention. Table 14 summarizes critical aspects of studies of educational interventions addressing key question KQ1, and Table 15 summarizes key outcomes of studies of good or fair quality.

Table 15. Outcomes of studies assessing educational interventions.

Table 15

Outcomes of studies assessing educational interventions.


Originally founded in the 1970s at the University of North Carolina at Chapel Hill, TEACCH involves primarily “structured teaching.” Structured teaching refers to applying “structure” to the organization of time, space, and sequences of events within the educational environment to promote learning by making activities clearer and easier to perform. Instruction is based on the idea that individuals on the autism spectrum have specific neuropsychological profiles described by strengths regarding processing visual information (compared with language use/understanding), heighted attention to details, significant variability regarding attention, communication difficulties, difficulties with time, attachment to routines, intense interests and impulses, and marked sensory preferences and aversions.184

By applying physical structure, indicating sequences of events, organizing individual tasks, and work/systems within a classroom environment children with autism are thought to engage in more developmentally appropriate behaviors that ultimately promote learning. TEACCH approaches often include a heavy reliance on visual supports (e.g., picture schedule) and arranging the physical environment to support individual learning.

Content of the literature. We identified 15 studies evaluating educational interventions overall (Table 15).185–200 Four of those studies185–188 evaluated implementation of iterations of the TEACCH program. Among studies assessing TEACCH, one was good quality, one was fair, and two were poor.

Summary of the literature. Four studies assessed various outcomes associated with implementation of components of the TEACCH curriculum (Table 15). One prospective cohort study187 evaluated the TEACCH program over 12 months for 18 children involved in a TEACCH classroom, with 16 receiving other types of individualized training (age range, 3–5 years). Evaluations of cognitive/developmental level, nonverbal intelligence, and adaptive behavior, were assessed using the Chinese version of the Psycho Educational Profile, The Merrill-Palmer Scale of Mental Tests, and the Hong Kong-Based Adaptive Behavioral Scales respectively at baseline, six months after initiation of treatment (Posttest 1), and again at 12 months (Posttest 2).

The intervention group demonstrated statistically significant improvement compared with the control group on the Chinese version of the Psycho Educational Profile Developmental Scale in perception, fine motor skills and gross motor skills (p≤0.05) after controlling for age, IQ, and pretest scores at Posttest 1. However, the control group showed more progress than the intervention group in the daily living domain and the Hong Kong-Based Adaptive Behavioral Scales sum of domains standard score (p≤0.05). Although significant improvements were see in the intervention group at 12 months for all scales and subscales in the Chinese version of the Psycho Educational Profile, Merrill-Palmer Scale of Mental Tests (total raw scores and mental age), and the Hong Kong-Based Adaptive Behavioral Scales the 12 month data are not provided for the control group.

The second prospective cohort study compared the effects of TEACCH in a residential center (n=11, mean age = 9.66 years), a specific school setting (n=13, mean age = 8.66 years), and included a comparison group in an inclusive mainstream classroom (n=10, mean age 9.09 years)185 The main components of the TEACCH intervention groups included arranging the environment with visual aids, individualized communication systems, self-care skills training, and daily living skills related intervention. Cognitive/developmental level and adaptive behavior were evaluated for each participant twice with a 3-year interval between evaluations, using the Psycho Educational Profile and the VABS respectively. Both TEACCH groups showed significant improvement on adaptive measures, but the mainstream classroom group did not. The mainstream group improved significantly on the Psycho Educational Profile perception measure only; both TEACCH groups demonstrated cognitive improvements on the Psycho Educational Profile on a variety of additional subscales.

Two case series of TEACCH were identified, one focused on a psycho-educational training program for parents188 and the other on teachers.186 Both had poor quality scores on our assessment. Each included 10 children with ASDs. After the parent training, children improved on the Ankara Developmental Screening Inventory in total development score, language-cognitive subscale, social and self-care subscale, fine motor subscale, and gross motor subscale.

The teacher training program was targeted to slightly older children (mean age 10 years), and evaluated effectiveness with a study-designed tool (the Classroom Child Behavioral Symptoms questionnaire), which was completed by their teachers at baseline and post-treatment (nine months later). Scores decreased from a mean pretest score of 106.4 to a post-test score of 100.8, representing a medium range effect size (d = 0.66).

Broad-Based Educational Approaches

Classroom and center-based approaches include a blend of teaching strategies that rely on ABA principles and techniques including reinforcement-based procedures such as incidental teaching, discrete trial training, and pivotal response training. Other interventions, such as TEACCH and language development interventions may also be incorporated in center-based treatment.

Content of the literature. Eight papers evaluated a variety of broad-based educational strategies.191–199 One study in this category was good, four were fair quality, and three were poor.

Summary of the literature. Several studies have investigated outcomes of children receiving specific or general instruction within early intervention centers or other classroom environment either within a specific curriculum or across multiple types of interventions (e.g., speech therapy, parent education, ABA instruction) (Table 15). A nonrandomized controlled trial197 compared a developmentally based early intervention (N=12, mean age = 42.6 months) to no treatment (N=5, mean age = 37.7 months). The Scottish Centre for Autism developed an individualized treatment program for preschool aged children with ASDs focusing on social, communicative, play, and adaptive behaviors and included a parent training component focusing on behavior management and teaching new skills. After approximately 11 months of treatment, adaptive behavior scores (VABS for socialization, daily living skills, motor, and composite) improved significantly for the intervention group compared with controls. The intervention group also showed a statistically significant improvement in imitation scores on the Pre-Verbal Communication Schedule, as well as joint attention scores, and social interaction skills measured on the Early Social Communication Scales.

One prospective cohort study compared an early intensive, home-based intervention using discrete trial techniques (and Verbal Behavior) (N=28) to a nursery school-based eclectic intervention (based in autism-specific classrooms) (N=16), which included components of TEACCH, the Picture Exchange Communication System (PECS), and other developmental and behavioral teaching strategies.195 The early intensive home-based intervention group and the nursery school-based intervention group had a mean age of 38 and 42.5 months, respectively. Nonverbal intelligence, cognitive ability, language skills, academic achievement/aptitude, and adaptive behavior were evaluated twice, with a 23–27 month interval between assessments.

No statistically significant differences were identified between the groups post-intervention on any of the measures, although both groups demonstrated improvement across measurements on average. A majority of the children demonstrating improvement had initial IQs above 70 and all but one were verbal at pre-treatment. Initial IQ and receptive language scores were correlated with progress over time.

Another prospective cohort study described outcomes for 65 children (ages 2.5–4 years at start of treatment) involved in broad-based eclectic teaching interventions and programs, often including reinforcement-based interventions, special nursery placements, speech and language therapy, and parent education programs. The authors compared groups (based on median split) of children receiving either196 high/low intensity (less than 15.6 hours per week of intervention) intervention whose parents either reported high or low levels of stress. Children were assessed at baseline and then after 9–10 months, including assessments of developmental/cognitive and adaptive behavior. Children receiving more intervention time (high intensity) had greater improvements across all three measures compared with those with less intervention time. Parenting stress did not affect gains with low intensity interventions but reduced the gains made by those in higher time interventions.

A prospective cohort study192 compared home-based tutor-led ABA teaching interventions (n=12) to two other teaching interventions, one of which was a school-based teaching program borrowing components from TEACCH with 6–8 children in each class (Special Nursery Placement) (n=20), and one (Portage) of which was a home-based, tutor-led program (n=16). The ABA intervention group received 1:1 interaction for two to three hours (including Lovaas, verbal behavior, and Comprehensive Application of Behavior Analysis to Schooling interventions). In the Special Nursery group, the children had several two to three hour sessions per week in a structured classroom (outlined by TEACCH methodology) with 6–8 other student and a teacher. In the portage group, a supervisor visited parents once every one to two weeks to demonstrate how to apply a system developed by a portage supervisor. Sessions were 40 to 60 minutes per day and scheduled when the parent believed the child would be at his or her most receptive. Children were taught new skills through the use of questions and tasks, prompts, and rewards.

The children were evaluated using multiple measures of symptom severity and intellectual functioning. Scores on measures related to intellectual functioning (the Psycho Educational Profile) in the ABA intervention group showed an overall gain of approximately 14 points, the nursery, approximately ten points, and little gain was shown in the portage group (~2 points). Authors documented cognitive, behavioral and adaptive behavioral skill improvements within each intervention group, but did not conduct direct comparisons between groups.

In a similar study comparing a one-to-one home based program (Parents of Autistic Children Training and Support), general special nursery placement, and ASDs-specific nursery placement and including children between 2 and 4 years old, participants in all groups showed marginal decreases in autistic severity and improved in educational functioning relative to baseline scores.199 Children in the nursery groups also showed improvements in adaptive behavior.

Rickards et al.194,198 investigated the addition of home-based intervention to a center-based educational program. The center-based program used training techniques like chaining, variety, repetition, sequencing, and a reward system to encourage learning through play, and used communication systems and behavior reinforcement. The home-based program included one of two specialist preschool teachers who visited each family weekly for one to one and a half hours for 12 months to discuss protocols developed at the centers, and develop new goals and strategies.

The home-based program also included parent training and adapting the home environment for the needs of the child. IQ improved by 1.6 points between baseline and Time 2 for the home plus center group and decreased by 4.3 for the center only group (p = .09). Preschool Behavior Checklist scores decreased by 8.4 for the intervention group and decreased by 1.8 for the control group (p=0.054).

Fifty-four children were assessed 12 months after intervention ceased,198 with children who had received the center and home based treatment maintaining cognitive gains compared with children receiving center-based intervention only (p=0.007). Behavioral improvements seen in the center+home based intervention group were not maintained a year later. Children from families experiencing greater parent-reported stress saw greater improvements in IQ at the 12-month followup than children from lower-stress families.

Case series191,193 evaluating the effects of a variety of ABA-based methods used within a teaching context in center-based classrooms were consistent with group design studies, measuring improvements in social and communication behaviors.

Computer-Based Educational Approaches

Computer-based programs use technology to deliver behaviorally-based teaching in areas such as language acquisition and reading skills.

Content of the literature. We identified thee papers evaluating computer-based intervention programs delivered in school settings.189,190,200 All studies in this section were poor quality.

Summary of the literature. Two randomized controlled trials189,200 and one prospective case series190 of computer-based academic interventions were included in this review. One randomized controlled trial compared a computer and teacher-led vocabulary acquisition program using behaviorally based teaching strategies like positive reinforcement,189 with seven children ages 3 to 6 years in each group. The computer program paralleled the teacher led approach with the addition of features such as color, animation and music. Children recalled more nouns after exposure to the computer, program, (mean=4.43, 74 percent) compared with teacher presentation (mean= 2.43, 41 percent), (p < .01) and were more attentive to the computer than to the teacher (mean= 97 percent vs. 67 percent), p<0.01.

A randomized trial of the TeachTown: Basics computer aided instruction program included 47 children randomized by classroom to either TeachTown instruction or regular school day instruction. Children in the intervention group received computer-based instruction for approximately 20 minutes a day for three months. Scores on standardized measures were better overall for children in the TeachTown group, but differences were not statistically significant. Some language scores for preschool children (but not those in kindergarten or first grade) in the TeachTown group improved significantly over those of children in the control group (p=.036). The total amount of time spent using the software was also correlated with the total number of lessons mastered.

In the case series on the use of multimedia computer program using voice, animation, video and sign language for increasing literacy and language,190 children also increased in verbal expressions (p=0.02), seeking help (p<0.05) and enjoyment (p<0.05) from pre to post intervention.

KQ1. Effects of Treatment on Core and Commonly Associated Symptoms in Children With ASDs: Medical Interventions

Medical treatments for symptoms of ASDs comprise a variety of pharmacologic agents including antipsychotics, psychostimulants, and serotonin reuptake inhibitors (SRIs). Modalities such as therapeutic diets, supplements, hormonal supplements, immunoglobulin, hyperbaric oxygen, and chelating agents have been employed to treat ASDs symptoms. We identified a total of 42 studies of medical interventions, of which 27 were RCTs. Table 16 summarizes critical aspects of studies of medical and related interventions addressing KQ1.


Recent clinical trials in children with ASDs have focused on the efficacy of a number of atypical antipsychotic medications for treating challenging behavior as well as other distressing symptoms. Risperidone was the first medication to receive US Food and Drug Administration (FDA) approval for the treatment of irritability in children ASDs. Aripiprazole also recently received FDA approval for irritability in children (6–17 years old) with autistic disorder.

Content of the literature. We identified 17 papers from nine studies that addressed the use of antipsychotic medications in children with autism spectrum disorders. Five of these studies evaluated the effects of risperidone,201–207,208,209–214 two of aripiprazole,215,216 and one of the addition of cyproheptadine to haloperidol.217 Most participants were recruited from non-primary care populations; table 16 summarizes additional information about these studies.

The risperidone literature base includes four RCTs,203,206,210–212 all comparing risperidone to placebo. Three of these RCTs were conducted in academic clinic settings using institutional and grant funding.203,206,212 The pharmaceutical company that owned the patent for risperidone sponsored another RCT.210,211 One prospective case series214 reported on associations between adverse events and efficacy of risperidone and eight candidate genes.

The literature base on the effects of aripiprazole in children with ASDs includes two RCTs,215,216 both conducted by the pharmaceutical company that owned the patent for aripiprazole. Each RCT compared aripiprazole to placebo in multiple study centers including both academic clinics and independent research centers.215,216

The literature base on the effects of cyproheptadine added to haloperidol includes one RCT.217 This study compared haloperidol alone to haloperidol plus cyptoheptadine in an academic clinic setting.

A variety of outcomes are reported in the literature on antipsychotic effects in children with ASDs, but the literature converges on the Aberrant Behavior Checklist-Community Version (ABC-C), a rating scale completed by caregivers of individuals with ASDs. For the purposes of this review, we will emphasize specific domains of behavioral change because most studies with significant differences in overall ratings also showed significant improvements on more specific measures. Potential side effects or harms, including assessment of weight gain, somnolence, and GI symptoms, were also assessed by most of these studies.

The RCTs included a total of 322 participants in treatment arms, and 214 participants in comparison arms. Participants had an average age of 9.0 and 8.7 years, in the treatment and comparison groups, respectively, when excluding one study that did not provide average ages in each arm.206 Both treatment and comparison groups had more male subjects (83.9 percent and 83.2 percent, respectively). Five of the studies included only participants with a DSM-IV diagnosis of Autistic Disorder;203,206,215–217 whereas one included subjects with any pervasive development disorder (Autistic Disorder, PDD-NOS or Asperger Disorder).211 Four of the studies used the ADI-R to corroborate diagnosis;203,206,215,216 whereas one study used the Childhood Autism Rating Scale211 and another study used DSM-IV criteria only.217 Only three RCTs provided IQ data on subjects, with the majority of subjects with IQs in the intellectual disability range.203,206,211

Among the studies of antipsychotics, three were good quality, four were fair, and two were poor.

Summary of the Literature

Risperidone. Of the four RCTs of risperidone, two targeted challenging behavior as the primary outcome (Table 17).206,211 The first study201,202,204–209 was sponsored by the National Institute of Mental Health as part of the RUPP Autism Network. The second study was sponsored by the manufacturer of risperidone.211

Table 17. Outcomes of RCTs of antipsychotic medications for challenging and repetitive behaviors.

Table 17

Outcomes of RCTs of antipsychotic medications for challenging and repetitive behaviors.

These studies included a total of 89 subjects in risperidone arms and 91 subjects in placebo arms. Both studies used a graduated dose titration design over eight weeks, with an average risperidone dose ranging from 1.5–1.8 mg per day, with one study using primarily once daily dosing211 and the other using twice daily dosing.201,202,204–209 In these two studies, baseline ratings of irritability were similar across risperidone (ABC-C-Irritability 18.9–26.2) and placebo (ABC-C-Irritability 21.2–25.5) arms. Decreases in ABC-C Irritability were significantly greater for the risperidone arms in both studies, which saw improvements of 12.1–14.9, compared with the placebo arms, which saw improvements of 3.6–6.5.

Similar improvements for a second measure of challenging behavior, the ABC-C Hyperactivity subscale, which indexes noncompliance as well as hyperactivity, were also seen in both trials. Baseline ratings of hyperactivity were similar, with decreases significantly greater for risperidone compared with placebo (14.8–14.9, in comparison to 4.7–7.4). The RUPP study also reported a number of other outcomes that may correlate with challenging behavior, including the Ritvo-Freeman Real Life Rating Scale Affectual Reactions subscale, which includes abrupt changes in mood, temper tantrums, and crying, the VABS Maladaptive Behavior Domain, and quantitative ratings of parent-rated target symptoms, each of which showed significant improvements in the risperidone group.204,207

Secondary outcomes in the two RCTs of risperidone included measures of repetitive behavior. Both studies included the ABC-C Stereotypy Subscale, which showed greater response in the risperidone arm. Baseline ratings of stereotypy were similar across risperidone and placebo arms. Decreases in ABC-C-Stereotypy were significantly greater for risperidone in one RCT (4.8 vs. 1.7)206 but would not have been significant after correction for multiple testing in the other RCT (4.3 vs. 2.4).211 One study204,206 also used the Children’s Yale-Brown Obsessive Compulsive Scale-PDD version to assess repetitive behavior, finding no baseline differences between the groups but a significantly greater decrease in the risperidone compared with placebo arms (3.9 vs. 1.0). A number of other outcomes were measured in these studies, but none outside of challenging behavior and repetitive behavior would have yielded statistically significant findings once corrected for multiple comparisons.

Two additional RCTs were identified that did not provide specific numerical ratings on either challenging behavior or repetitive behavior.203,212 One of these was an eight-week drug discontinuation RCT with risperidone and placebo arms203 after positive response during four months of open label risperidone treatment following the RUPP risperidone RCT.206 This publication did not provide quantitative outcome data but instead indexed “relapse” using a composite measure of ABC-C-Irritability and clinician ratings of CGI-I,203 finding significantly less “relapse” in the risperidone arm (two of 16 subjects) in comparison to the placebo arm (10 of 16 subjects).

The last RCT was a six-month RCT with risperidone and placebo arms that used a variety of general rating scales to assess response and provided quantitative outcome data on only some of these scales,212 with the primary outcome measures being parent ratings on the Childhood Autism Rating Scale and clinician ratings on the Children’s Global Assessment Scale. The study only reports Childhood Autism Rating Scale median ratings for those participants with at least a 20 percent response.212 Average ratings on the Children’s Global Assessment Scale were similar in the risperidone (29.8) and placebo (32.7) arms with more improvement in the risperidone (11.1) than placebo (2.5) arms.

All of the risperidone RCTs also provided data on adverse events or side effects (Table 19). All studies reported on weight gain,202,203,206,210–212 which was greater in the risperidone arms (2.7–2.8 kg) than in the placebo arms (0.8–1.7 kg), with a statistically significant difference reported in two of the studies.202,206,210,211 Three of the RCTs202,203,206,210,211 provided data on other adverse events in both the risperidone and placebo arms. Somnolence or drowsiness was the most common adverse event in two of these studies, occurring in 53 of 89 subjects in risperidone arms and nine of 91 subjects in placebo arms.202,206,210,211 These studies also reported that this somnolence improved over time.202,206,210,211 Both these studies also reported more extrapyramidal symptoms, including tremor, dyskinesia, and rigidity, in the risperidone arm in comparison with the placebo arm, but these events were categorized and summed differently between the two studies and did not clearly show a statistically significant difference between treatment arms.202,206,210,211

Table 19. Harms frequently reported in studies of medical interventions.

Table 19

Harms frequently reported in studies of medical interventions.

The RUPP study206 also reported a greater rise in prolactin levels in the risperidone arm (27.7 ng/mL) compared with the placebo arm (0.8 ng/mL);209 although it did not report clinical events such as gynecomastia or galactorrhea that could be related to elevated prolactin levels.203,206 The RUPP study specifically assessed cognitive function in a subset of subjects and found no worsening and some evidence of improvement on risperidone that would not be statistically significant after correction for multiple testing.201,206

Case series data, including that from the risperidone arm and open label extension of the RUPP RCT,203 indicated results consistent with the risperidone arms of the RCTs.177,203,206,208,214

Aripiprazole. We identified two eight-week randomized, controlled trials of aripiprazole in children with ASDs (Table 17).215,216 The manufacturer of aripiprazole sponsored both studies. The primary outcome for these studies was challenging behavior indexed by the ABC-C Irritability subscale. These studies included a total of 213 subjects in aripiprazole arms and 103 subjects in placebo arms within the intent to treat analyses.

One study used a fixed dose design with one placebo arm and three arms corresponding to 5, 10, and 15 mg per day of aripiprazole,215 with all subjects beginning at two mg per day with forced titration weekly to the next dose until they reached their goal dose. The other study used a dose titration schedule with weekly progression from 2 mg to 5 mg, 10 mg, and 15 mg per day following clinical judgment.216 In these two studies, baseline ratings of irritability were similar across aripiprazole (ABC-C-I 28.2–29.6) and placebo (ABC-C-Irritability 28.0–30.8) arms. Decreases in ABC-C Irritability were significantly greater for the aripiprazole arms in both studies, with improvements of 12.4–14.4, in comparison to the placebo arms, with improvements of 5.0–8.4. The trial with differing set doses of aripiprazole demonstrated increasing response with increasing dose.215 Overall, the results of the trial that used titration following clinical judgment were more pronounced.216

Additional assessments of challenging behavior were also performed in these two aripiprazole RCTs. Similar improvements for a second measure of challenging behavior, the ABC-C Hyperactivity subscale, which indexes noncompliance as well as hyperactivity, were also seen across both trials. Baseline ratings of hyperactivity were similar across aripiprazole and placebo arms. Decreases in ABC-C-Hyperactivity were significantly greater for aripiprazole in which improvements of 12.7–16.3 were seen, in comparison to the placebo arms, with improvements of 2.8–7.7. The ABC-C Inappropriate Speech subscale also showed significant improvement in one study216 with a supportive trend in the other.215

Secondary outcomes in the two major RCTs of aripiprazole included measures of repetitive behavior.218 Both studies included the ABC-C Stereotypy Subscale, which showed significantly greater response in the aripiprazole arms. Baseline ratings of stereotypy were similar across aripiprazole and placebo arms. Decreases in ABC-C-Stereotypy were greater for aripiprazole, with improvements of 4.2–4.8, in comparison to placebo arms, with improvements of 1.8–2.0. Both studies also used the Children’s Yale-Brown Obsessive Compulsive Scale-PDD version to assess repetitive behavior, finding no baseline differences between the groups but a greater decrease in the aripiprazole compared with placebo arms (2.4–3.8 vs. 0.8–1.7). A number of other outcomes were measured in these two studies, but none outside of challenging behavior and repetitive behavior yielded statistically significant findings once corrected for multiple comparisons.

The two aripiprazole RCTs also provided data on harms (Table 19). Both studies reported on weight gain,215, 216 which was greater in the aripiprazole arms (1.3–2.0 kg) than in the placebo arms (0.3–0.8 kg), with a statistically significant difference reported in both of the studies.215,216 Somnolence and sedation were the most common adverse events in both of these studies, occurring in 66 of 210 subjects in aripiprazole arms and eight of 101 subjects in placebo arms.215,216 Both studies also reported more extrapyramidal symptoms, including tremor, dyskinesia, and rigidity, occurring in 44 of 210 subjects in the aripiprazole arms in comparison with ten of 210 subjects in the placebo arms.215,216 Both studies found a statistically significant decrease in prolactin levels in the aripiprazole arms in contrast with the placebo arms.215,216

Cyproheptadine plus haloperidol. One eight-week RCT compared addition of cyproheptadine versus placebo to haloperidol.217 Each arm contained 20 subjects.217 The medication doses were titrated up from some starting point to cyproheptadine 0.05 mg/kg/day and haloperidol 0.2 mg/kg/day, but no details are provided. Two general outcome measures were used, the ABC-C and the Childhood Autism Rating Scale. Each measure was apparently translated into Farsi, although no details were provided on validation of the translated versions, nor is it clear whether parents completed paper and pencil versions of these measures or were interviewed. Baseline scores on measures were only presented in graphs.

The ABC-C scores shown are markedly lower than the total ABC-C scores obtained at baseline in other medication trials,206,211,215,216 suggesting that a subscale may have been administered; although this is not stated.217 The response to placebo plus haloperidol was smaller217 than the response found in previous haloperidol trials.39,219 The improvement in ABC-C score in the cyproheptadine plus haloperidol arm was larger (10.9) than the improvement in the placebo plus haloperidol arm (3.7).217 Similarly, the improvement in Childhood Autism Rating Scale score was greater in the cyproheptadine plus haloperidol arm (1.85) than in the placebo plus haloperidol arm (0.37).

Serotonin Reuptake Inhibitors

SRIs have come into wide use for the treatment of depression and anxiety and are some of the most commonly prescribed medications for children with ASDs.41–43 Most recent clinical trials in children with ASDs have focused on their potential to decrease repetitive behaviors.49

Content of the literature. We identified five studies that addressed the use of serotonin reuptake inhibitor medications in children with autism spectrum disorders; table 17 includes additional details. Two of these studies evaluate the effects of fluoxetine,220,221 one of citalopram,222 one of escitalopram,223 and one of a variety of SRIs.224

The literature base on the effects of fluoxetine in children with ASDs included one retrospective case series221 and one randomized, controlled, cross-over trial of fluoxetine compared with placebo. Both of these studies were conducted in academic clinic settings using institutional or grant funding.220 The single study on the effects of citalopram in children with ASDs was a randomized, controlled trial conducted in multiple academic centers using institutional and grant funding.222 The single study on the effects of escitalopram in children with ASDs was a prospective case series223 that analyzed outcome by serotonin transporter genotype, and is therefore discussed in detail in the Modifiers of Treatment Effectiveness (KQ 2) section of the report. One retrospective case series224 reported on a number of SRIs.

The two RCTs included a total of 112 participants in treatment arms, and 115 participants in comparison arms. Participants had an average age of 8.8 and 9.1, in the treatment and comparison groups, respectively.220,222 Both treatment and comparison groups had more male subjects (83.7 percent and 81.7 percent, respectively). Both of the studies included subjects with any pervasive development disorder (Autistic Disorder, PDD-NOS, or Asperger Disorder) and used either the ADI-R222 or the ADI-R and the ADOS for corroboration.220

One of the RCTs had a minority of subjects with intellectual disability,222 and the other had an average IQ in the intellectual disability range.220 The three case series included 276 subjects with any diagnosis of PDD.221,223,224 One study used ADI-R to corroborate DSM-IV diagnosis,223 one used CARS plus ADOS,221 and one used DSM-IV only.224 Among the case series subjects, the average age was 6.9 years old and 87.1 percent were male. Only one case series provided IQ measures on subjects223 and had average verbal and nonverbal IQ in the borderline to low average range (76 and 86, respectively). Among all studies of SRIs, one was good quality, two were fair, and two were poor.

Summary of the literature. We review citalopram and escitalopram together because escitalopram is the active component (enantiomer) of citalopram. One 12-week randomized, controlled trial of citalopram was identified.222 This trial focused on repetitive behavior outcomes with a number of secondary outcomes also measured.

The entry criteria for the study were a PDD diagnosis corroborated by both ADI-R and ADOS, moderate illness severity on the CGI-Severity and significant repetitive behavior on the Children’s Yale-Brown Obsessive Compulsive Scale-PDD version.222 It had 73 subjects in the citalopram arm and 76 in the placebo arm.222 Subjects were begun on 2.5 mg of citalopram daily with weekly increases of 2.5 mg per day for the first five to six weeks as clinically indicated, followed by weekly increases of up to five mg per day thereafter, up to a maximum dose of 20 mg per day.222 This dose is lower than the equivalent daily dose of SRIs used in obsessive compulsive disorder in previous studies,225,226 but it is similar to dosing used in an earlier case series in autism.227

No significant difference between citalopram and placebo arms was seen in measures of repetitive behavior, with similar baseline scores on the Children’s Yale-Brown Obsessive Compulsive Scale-PDD version (15.1 vs. 15.0) and similar improvements (2.0 vs. 1.9) in each arm. The other measures of repetitive behavior, including the Repetitive Behavior Scale-Revised, also showed similar baseline scores and similar improvements in each arm with no evidence for an effect of citalopram (Table 18).222 The CGI-Improvement similarly showed no significant difference between the citalopram and the placebo arm. On the other hand, the primary measure of challenging behavior reported in this trial, the ABC-C Irritability subscale, showed an advantage for citalopram.

Table 18. Outcomes of studies of SRIs for the treatment of repetitive and challenging behaviors in ASDs.

Table 18

Outcomes of studies of SRIs for the treatment of repetitive and challenging behaviors in ASDs.

The baseline ratings were not statistically different between the citalopram and placebo arms (13.2 and 11.2, respectively), but more improvement was seen for citalopram (3.2) than for placebo (0.9).222 Adverse effects in this study included a marked increase in what were termed “activation” symptoms, including increased energy, hyperactivity, inattention, disinhibition, and decreased sleep in the citalopram arm in comparison to the placebo arm.222 Diarrhea and dry or itchy skin were also more common in the citalopram arm.222

One prospective case series of escitalopram was identified.223 This ten-week study sought to identify pharmacogenetic modifiers of treatment response in the challenging behavior domain as measured by the ABC-C-Irritability. Fifty-eight subjects with a PDD corroborated by ADI-R and a minimum ABC-C-I score of 12 underwent a forced dose titration of escitalopram from 2.5 mg daily increasing weekly to 5 mg, 10 mg, 15 mg, and 20 mg, essentially twice the dose equivalent of citalopram given that escitalopram is the active component of racemic citalopram. Pre-designated dose-limiting side effects included sleep disruption and an increase in ABC-C Irritability or Hyperactivity subscales of ten points over the previous week. Average daily doses of escitalopram were 10.8–12.4 mg and did not differ across genotype groups, which reflects the fact that most subjects in all genotype groups could not tolerate the maximum dose. Unfortunately, the data are presented in figures only, and raw values cannot be inferred. It is evident, however, that the ABC-C-Irritability for all subjects was 20 or greater at baseline and that improvements were about ten points for three of the four genotype groups.223 Adverse effects were not directly assessed in this study.

One randomized, controlled crossover trial of fluoxetine was identified with two eight-week treatment periods separated by a four-week washout period.220 Thirty-nine subjects with a PDD corroborated with ADI-R and ADOS were included in the final analysis with no minimum required score on a repetitive behavior scale. Five additional subjects were randomized but not included in the analysis for various reasons. Of the randomized subjects, 19 received fluoxetine followed by placebo and 20 received placebo followed by fluoxetine. During each phase of the study, subjects began the first week at 2.5 mg per day of fluoxetine or placebo, followed as clinically indicated by weekly upward titration to 0.3 mg/kg for week 2, 0.5 mg/kg/day for week 3, and 0.8 mg/kg/day for weeks four to eight. During the first eight-week treatment period of the study, subjects randomized to fluoxetine first had baseline Children’s Yale-Brown Obsessive Compulsive Scale scores of 12.8 and those to placebo first had baseline scores of 13.5.

Subjects in the first fluoxetine group showed an improvement of 1.2, and those in the first placebo arm showed an improvement of 0.5. These differences were not statistically significant when considered alone. In the second eight-week treatment period, subjects randomized to fluoxetine second had baseline Children’s Yale-Brown Obsessive Compulsive Scale scores of 12.8 and those to placebo second had baseline scores of 12.2. Subjects in the second fluoxetine arm showed an improvement of 1.2, and those in the second placebo arm showed a worsening of 0.1. When analyzed together with the first treatment period in a repeated measures design, the Children’s Yale-Brown Obsessive Compulsive Scale change in the fluoxetine arms was significantly greater than the change in the placebo arms. No adverse events were significantly more frequent in the fluoxetine group; although more subjects on fluoxetine had their dose reduced due to agitation.220 The two chart reviews of SRIs 221,224 reported in the literature were of poor quality and included general outcome measures that are difficult to compare with the RCT data.

Stimulants and Other Medications To Treat Hyperactivity

Psychostimulants treat hyperactivity and inattention in patients diagnosed with attention deficit hyperactivity disorder (ADHD) and include agents such as methylphenidate (MPH), amphetamine, and dextroamphetamine. Other medications studied for the treatment of ADHD have also been studied for the treatment of hyperactivity in ASDs.50–54

Content of the literature. We identified six publications228–233 from four studies that addressed stimulant and other medications to treat hyperactivity; table 17 summarizes study information. Most participants were recruited from centers in the RUPP network.228–230 Studies assessed the use of MPH in children with PDDNOS and hyperactivity;228–230 psychostimulants;231,233 and guanfacine to target hyperactivity, inattention, and impulsivity in children with PDD.232 Among all studies of these medications, one was good quality and three were poor.

Summary of the literature. The RUPP Autism Network’s double-blind cross-over trial228–230 of MPH included 72 children with autism, PDD-NOS or Asperger disorder who received a one-day placebo followed by two days at each of three (low, medium, high) test doses of MPH; doses ranged from 7.5 mg/day to 50.0 mg/day. Subjects tolerating MPH (n=66) moved on to a 4 week, double-blind crossover phase. Subjects with a positive response in the double blind phase (n=34) completed an eight week open label continuation phase at their best dose. The primary outcome measure was hyperactivity as assessed by the ABC-C teacher-rated hyperactivity subscale; secondary measures included the ABC-C parent-rated hyperactivity subscale. Blinded clinicians also assessed participants using the CGI-Irritability scale; this subscale and the ABC parent and teacher rated hyperactivity subscales were combined to assess response.

In the double-blind crossover phase, all MPH doses demonstrated effects that were statistically superior to placebo, and effect sizes favored the medium dose for parent ratings and high dose for teacher ratings. Parent-rated lethargy/social withdrawal significantly worsened during the high dose of MPH compared with placebo. Significant improvement in parent-rated stereotypy and inappropriate speech were seen at the medium dose of MPH compared with placebo. Hyperactivity/impulsivity also improved more with the medium and high MPH doses than at the low dose (Table 20).

Table 20. Outcomes of studies of stimulants for the treatment of hyperactivity in ASDs.

Table 20

Outcomes of studies of stimulants for the treatment of hyperactivity in ASDs.

Significantly more joint attention behaviors as measured on the Joint Attention from the Early Social Communication Scales in the intervention group were reported both with the best MPH dose and with the low dose compared with placebo. There was improved self-regulation, as assessed in a “competing demands” task in low dose as well as in medium dose MPH compared with placebo. A significant increase in neutral affect was also found for the medium and high dose, which could be either beneficial, in the case of children with a labile mood, or damaging, in the case of children with flattened affect due to a medication side effect. Irritability was the most frequent reason for discontinuation (18 percent) of treatment.

Three chart reviews assessed guanfacine and psychostimulant use in children with ASDs, all of which were poor in quality and provided conflicting results on the effectiveness of stimulants for hyperactivity in children with ASDs.231–233. The three chart reviews had a longer duration of followup visits: 7 days to 4.8 years,232 1 day to 10 years,233 4 ± 3.9 years,231 while the RCT followed patients for 8 weeks.228–230


Interest in secretin for the treatment of ASDs symptoms stemmed from an unblinded, uncontrolled cases series of three children that reported social, cognitive and communicative gains in recipients after the first infusion and after a second infusion given weeks later.59 Multiple studies conducted after the initial case series have evaluated secretin’s utility in treating autistic symptoms, employing single or multiple doses of synthetic human or porcine secretin.

Content of the literature. We identified eight publications234–241 addressing secretin use in eight unique populations; table 17 summarizes additional study information. Among studies of secretin, two were good quality, five were fair, and one was poor.

Summary of the literature. Of the eight studies evaluating the impact of secretin in the treatment of ASDs, one238 was a repeated dose intervention study. Two studies used synthetic human secretin,234,235 three used porcine secretin,237,239,240 and one biologic secretin.236 All were randomized controlled trials except one open label trial of secretin (type unknown) with a prospective case series study design;241 all of the studies evaluated only short-term outcomes with followup periods ranging from 3 to 12 weeks.

No studies showed significantly greater improvements in measures of language, cognition or autistic symptoms when compared with placebo; in those studies that demonstrated improvement over time, they did so equally in both intervention and placebo groups. There also was no benefit by type of secretin (porcine or synthetic).

Dietary and Other Medical Interventions

Additional studies in the medical literature addressed the use of dietary interventions (including special diets) as well as medical therapies for sleep and gastrointestinal dysfunction, hyperbaric oxygen, and other agents.

Content of the literature-dietary interventions. We identified eight studies conducted in the clinic setting that addressed the use of various oral dietary supplements to treat ASDs in children, including ages ranging from one to 18 years; interventions included iron,242 magnesium-vitamin B6,243 melatonin,244 ketogenic diet,245 and fish oil and evening primrose oil.246 Two studies focused on amino acid derivatives, including L-carnosine247 and N, N dimethylglycine,248 and one on a digestive enzyme supplement.249

Studies measured a wide range of outcomes, with little overlap in instruments utilized among the studies. Most studies analyzed outcomes after 18 days to 3 months;242,246–249 one study assessed outcomes after 6 months,245 one study after a mean of eight months,243 and one examined data at a mean of 1.8 years of followup.244 Five studies reported adverse events.242,245,248–250 Table 17 includes additional study information.

Summary of the literature-dietary interventions. Two RCTs explored dietary supplementation with amino acid derivatives in ASDs. The 8-week RCT of daily L-carnosine supplementation was conducted at a specialty clinic and included 31 children ages three to twelve years with ASDs.247 Significant changes on the CGI at 2 weeks compared with 6 weeks were observed in the L-carnosine group; significant effects on receptive vocabulary and GARS scores were also observed before vs. after treatment in this group. However, no significant changes were observed between groups in the Gilliam Autism Rating Scale measure of ASDs severity, Receptive and Expressive Picture Vocabulary scores, or parent-rated CGI measure of overall improvement. Adverse effects in the L-carnosine group included sporadic hyperactivity and were alleviated by dose reduction.

The 4-week randomized controlled trial of daily N, N-dimethylglycine treatment was conducted at an academic clinic and included 37 children with ASDs ages three to eleven years.248 The dimethylglycine group did not improve more than the placebo group on any behavioral measure (VABS, ABC-C). There was no significant effect on neurologic examination of motor skills, muscle tone, or coordination. Adverse effects in the dimethylglycine group were similar to placebo.

A six month, double blind, crossover RCT of a digestive enzyme supplement (Peptizyde) which combines the enzymes peptidase, protease 4.5, and papain, included 43 children between the ages of 3 and 8 years (mean= 69.4 months).249 Most participants (84 percent) were diagnosed with autistic disorder. Participants were randomized to either enzyme for three months followed, after a one-week washout period, by placebo (n=21) or placebo for three months (one-week washout period) followed by enzyme (n=22). Sixteen participants (10 in the enzyme/placebo group and 6 in placebo/enzyme) dropped out of the study for reasons including a parent-perceived increase in negative behavior and child refusal to eat food with contents of enzyme capsules added. In intention to treat analyses, investigators reported significant differences between enzyme and placebo only on a measure of food selectivity, which was not sustained over the study period. Investigators noted no serious adverse effects though four children withdrew from the study because of behavioral deterioration perceived by parents. The authors also assessed potential effects of alternative therapies, multivitamins, prescription medications, and special diets on the effects of Peptizyde but found no changes.249 Table 21 summarizes the results of studies of good or fair quality.

Table 21. Outcomes of RCTs of dietary supplements for the treatment of ASDs.

Table 21

Outcomes of RCTs of dietary supplements for the treatment of ASDs.

The remainder of the studies were case series of poor quality, in which modest effects were observed in improving sleep with iron supplementation242 and melatonin,244 and for affecting general autism symptoms with magnesium and vitamin B6243 and a combination of fish oil and evening primrose oil.246 Modest improvements were seen in some children with a ketogenic diet, but drop out was high.245

Content of the literature-other interventions. We identified eight papers from seven studies conducted in the clinic setting that examined various other medical therapies for treatment of ASDs in children, including ages ranging from two to 19 years; interventions included amantadine,251 piracetam or pentoxifylline added to a risperidone regimen,252,253 hyperbaric therapy,254 oral human immunoglobulin,64 and dimercaptosuccinic acid (DMSA).250,255 Two studies focused on cholinesterase antagonists, including rivastigmine tartrate256 and donezepil hydrochloride.257

These studies typically examined outcomes after three to 12 weeks of therapy.64,251–254,256,257 Seven studies reported on adverse events.64,251–255,257 Table 16 summarizes additional study details. Among all studies of dietary and other interventions, two were good quality, seven were fair, and seven were poor.

Summary of the literature-other interventions. The RCT of amantadine,251 conducted at six academic clinics in the US and UK and including 39 children, showed no effect of daily amantadine over four weeks on parent-rated ABC-C behavior scores and clinician-rated CGI rating of overall improvement (Table 22). However, children in the amantadine arm improved significantly more than those receiving placebo in clinician-rated ABC-C subscales for hyperactivity and inappropriate speech. There were no differences in harms, and no serious complications.

Table 22. Outcomes of RCTs of other medical interventions for the treatment of ASDs.

Table 22

Outcomes of RCTs of other medical interventions for the treatment of ASDs.

Risperidone plus piracetem was associated with more improvement on the ABC-C than risperidone alone in one RCT of 40 children ages 3–11, with similar incidence of extrapyramidal symptoms, and other adverse events.252

A 10-week RCT of pentoxifylline added to risperidone compared with placebo plus risperidone included 40 children between the ages of 4 and 12 years.253 Scores on the ABC-C lethargy/social withdrawal, stereotypic behavior, hyperactivity/noncompliance, and inappropriate speech subscales were significantly better for the pentoxifylline group compared with placebo (P ≤ 0.0001). Scores on the Extrapyramidal Symptoms Rating Scale and frequency of side effects did not differ between groups. Adverse events reported in either group included GI symptoms, restlessness, drowsiness, weight gain/increased appetite, and fatigue.

In one RCT at 12 US centers, oral human immunoglobulin showed no effect on gastrointestinal symptoms, ABC-C scores or clinical rated global impression scores, and there was a significant effect in favor of placebo on the parent-rated global impression scores.64

Hyperbaric therapy was also studied in one RCT and showed some benefit. The study was conducted at six US clinics and included 62 children ages 2–7 years with ASDs.254 Participants were randomized to 40 hourly sessions over 4 weeks of either hyperbaric therapy (room pressure 1.3 atmospheres (atm); 24 percent oxygen) or slightly pressurized room air (1.03 atm; 21 percent oxygen). Both groups had significant improvement in clinician-rated CGI after treatment as compared with baseline; however, while eighty percent of children in the hyperbaric group improved on this outcome, only 38 percent of control participants improved.

The investigators reported no significant difference in ABC-C or parent-rated CGI between groups at outcome. There were significant changes in Autism Treatment Evaluation Scale total scores and some subscales for both groups compared with baseline; significant between-group differences in the amount of change were observed on the sensory/cognitive awareness subscale only. No episodes of seizure or barotrauma occurred during the sessions; other adverse events were rare and included two skin-related events, worsening of asthma, and GI symptoms.

Two studies explored the use of cholinesterase inhibitors in ASDs, both conducted at a specialty clinic. The study of donepezil hydrochloride included 43 children ages 2–10 years with ASDs.257 Children were randomized to six weeks of donepezil or placebo, followed by a six-week open label continuation. Both groups showed significant improvements on the Childhood Autism Rating Scale measure, but there were no between group differences. Nine children withdrew after the first six weeks; six of these children were in the donepezil group when blinding was broken and two of these discontinued due to gastrointestinal problems, while six discontinued due to increased irritability and associated behavioral issues.

One poor study each was identified for rivastigmine tartrate therapy256 and daily oral DMSA.250,255 In a 2-week open label study of rivastigmine tartrate therapy scores at 12 weeks were significantly improved on the Childhood Autism Rating Scale, expressive picture vocabulary, and the Comprehensive Psychopathological Rating Scale measure of oppositional, hyperactive, and inattention-related behaviors; no significant effects were observed on receptive picture vocabulary at 12 weeks. The randomized controlled trial of DMSA demonstrated no advantage of DMSA over placebo.

KQ1. Effects of Treatment on Core and Commonly Associated Symptoms in Children With ASDs: Allied Health Interventions

Among allied health disciplines, several intervention approaches have been developed to address core symptoms of autism and associated difficulties and deficits. We broadly divided allied health studies meeting our inclusion criteria into those focused on language, those employing sensory or auditory integration techniques including music therapy, and those addressing techniques such as horseback riding and occupational therapy. Table 23 summarizes aspects of studies of allied health interventions addressing KQ1.

Language Interventions

Two approaches to increasing speech and language were identified in the included studies: the Picture Exchange Communication System (PECS) and Responsive Education and Prelinguistic Milieu Teaching (RPMT). We have clustered these two approaches for the purposes of the report, and because in the one study that compared different approaches, these were the two comparative conditions.

Content of the literature. Our search identified eight publications focused on speech and language interventions,258–265 representing four distinct study populations. Two of the studies were RCTs; one was described in four papers,258–261 and one in a single study.263 Three of the four studies referred to a protocol, and one measured treatment fidelity. No two studies used exactly the same outcome measure, making it difficult to summarize across studies. Each is therefore described separately below, and Table 23 provides an overview of study information. Among the four unique studies, one was good quality and three were poor.

Summary of the literature. Four analyses from one randomized trial258–261 comparing RPMT to PECS in the United States met our criteria and were included. The 36 preschoolers included in the study had a confirmed diagnosis of Autistic Disorder or PDD-NOS, were between 18 and 60 months of age, and were nonverbal or had low verbal status (fewer than 20 different words used cumulatively during three communication samples). Most (86 percent) of the participants were male, and most (69 percent) were white. Although the mean chronologic age of the children was 33.6 months (range from 21 to 54 months), the mean mental age in the nonverbal children was 18.6 months and in the verbal children was 11.9 months, based on Mullen Scales of Early Learning. Diagnosis of autism was confirmed with the ADOS for all children. Nineteen children were randomized to receive PECS and 17 to receive RPMT.

During the six-month treatment phase, children attended three 20-minute therapy sessions per week, with parents offered up to 15 hours of training. Outcomes were measured at the end of treatment and 6 months after treatment completion in the context of a 15-minute free play session in which the interaction style, toys, examiner and location were all different from those in which the intervention was conducted. Treatment fidelity was assessed monthly, interrater reliability was assessed on at least 20 percent of data points, and coded data were double entered to ensure accuracy.

At Time 2, the PECS group had significantly higher frequency of non-imitative spoken communication at Time 2 than the RPMT group, and higher numbers of different non-imitative words. No overall significant between group differences were observed at Time 3, indicating that the treatment effects did not maintain at six months after the end of treatment. This study included an analysis of initial characteristics of the children, demonstrating that children who were low in initial object exploration benefitted more from RPMT, which explicitly teaches play with objects; while children who were relatively high in initial object exploration demonstrated more benefit from PECS.

These results were maintained at 6 months. An additional analysis based on this study260 showed greater increases in generalized turn-taking and initiating joint attention in the RPMT group than in PECS. The increased benefit for RPMT in join attention was only seen, however, in children who began the study with at least some initiation of joint attention. Specifically, children most likely to benefit from RPMT in increasing joint attention had demonstrated at least seven acts of joint attention in the pre-intervention assessment. RPMT was also superior in this analysis in increasing object exchange turns.

The second RCT263 in this literature was a pragmatic trial focused on the effect of providing expert training in PECS to teachers in specialist schools, under the assumption that although more than half of autism-specific schools in the UK claim to use PECS system, few teachers have been adequately trained to provide it. This study aimed to explore the potential effects of intensive training of teachers on child outcomes in spontaneous communication and speech; thus randomization occurred at the classroom level in three groups – immediate treatment group, delayed treatment group, and no treatment group. Teachers in the delayed treatment group received intensive training 2 terms after the immediate treatment group.

The 83 children included in the study were between the ages of 4 and 11 and had little or no functional language; most (75) met ADOS-Generic criteria for Autistic Disorder, and nine met criteria for another ASD. Treatment and assessment of language and nonverbal developmental quotient (nonverbal mental age equivalent/chronological age X 100) took place in schools with children videotaped and language outcomes codes by non-masked assessors. Videotaping took place during snack sessions, selected because they are time limited and structured to include requesting and other social communication.

The use of PECS in treatment classrooms was substantially increased immediately after training, with children in the PECS training classrooms having 3.90 times (95 percent CI: 1.75–8.68) the odds of being in a higher PECS use category than those whose teachers received no PECS training. The effect, however, was not maintained, and by Time 3, the immediate treatment group children were no more likely to be in a higher PECS rate category than untreated children (OR 1.10; 95 percent CI: 0.46–2.62).

The third trial available on language intervention was also on PECS and was a nonrandomized controlled trial, with selection into or out of treatment based on a geographical limit. This study was reported in two papers.262,265 Intervention was provided to 24 children whose special education classrooms were within 50 miles of the researchers; while a comparison group was selected from special education classrooms outside the 50-mile limit. Children were between 3 and 7 years old, with a diagnosis of Autistic Disorder. Diagnosis was clinically based, and not independently confirmed by the investigators. Children in the study were to have received no PECS training prior to the research.

Children in the intervention group received a total of 15 hours of PECS teaching during various classroom activities over 4 to 5 weeks. The research team designed the instrument by which they assessed communicative interactions. They did not validate the instrument. The investigators followed the Bondy and Frost guidance82 for using PECS in the classroom, but at Phase III randomly assigned children to either the approach specified in the PECS manual, or a modified PECS teaching procedure.265

Observation occurred 6 weeks prior to commencement of PECS intervention, 1 week prior to intervention and during the week immediately following completion of the 15 hours of PECS. The authors contended that the differences observed between the first and second observation periods (during which no intervention took place) would provide a measure of rate of maturation that could be used to differentiate maturation from treatment effects after the intervention phase; however, no evidence that this approach is valid was provided in the paper.

No differences were observed on child-initiated communication either between the two pre-intervention measurement periods or between the two groups immediately prior to intervention. After intervention, however, the children in the PECS manual group had a significant increase in initiations, while the children in the control group had no increase, and the frequency of child to adult initiations was higher overall in the PECS manual group. Total adult-to-child initiations with the opportunity for child response showed an opposite response, with no significant increase in the PECS manual group, but a significant increase in the control group.

Because the study only measured outcomes immediately after intervention in the classroom setting (at 6 weeks), it is impossible to determine whether the outcomes have any durability or to assess the effects outside of the classroom. Furthermore, although reliability of the observations was reported to be high (89.78 percent agreement), it was measured in only 56 percent of observation sessions, and the observers were not blinded to intervention status.

Sensory- and Auditory-Focused Interventions

Sensory integration, a specialized occupational therapy model based on the premise that the brain’s response to basic sensory input must be normalized before higher-order processes can be addressed,85 typically involves one-on-one, child-directed treatment with a trained occupational therapist and a focus on somatosensory and vestibular systems. Ultimately, it is hypothesized that if a child is better able to process, modulate, and integrate sensory information, he will then be better able to acquire higher-order skills.16

Auditory integration training (AIT) relates more specifically to sensory differences in the auditory realm. AIT involves repeatedly presenting children with modulated music according to specific protocols with a therapeutic goal of improving auditory processing, lessening auditory hypersensitivities, and increasing concentration.87 Finally, music therapy is at times employed with children with ASDs, hinging on speculation that children engage more with music. This treatment method is improvisational and unstructured in nature, and practitioners purport that it can improve both verbal and nonverbal communication skills including joint attention abilities, thereby improving core symptoms of autism.88,89

Content of the literature. We identified two studies266–268 that addressed sensory integration interventions, including a prospective case series with two publications presenting nearly identical results267,268 conducted in South Korea. Both were completed in a clinic setting and included children ages 6 years267,268 and older.266 All children in both studies met criteria for autism (i.e., none had diagnoses of PDD-NOS or Asperger Disorder). Both studies used outcome assessments developed within the project and focused on sensory and motor functioning. Both sensory studies were poor quality.

Two papers assessed auditory integration interventions269,270 and targeted children between 3 and 7 years old270 and 7 to 13 years old.269 Both studies were conducted through a clinic associated with an academic institution and enrolled children with diagnoses of autism across a broad range of cognitive functioning levels. Table 24 provides additional study details.

Table 24. Outcomes of RCTs of auditory/music interventions for the treatment of ASDs.

Table 24

Outcomes of RCTs of auditory/music interventions for the treatment of ASDs.

We identified two papers from a single crossover RCT that compared music and play therapy interventions.89,271 The study included children ages 3 to 5 years meeting DSM-IV criteria for autism and meeting criteria on the Childhood Autism Rating Scale; some were also administered the ADOS to confirm diagnosis. Among studies of auditory integration and music therapy, two were fair quality and one was poor.

Summary of the literature. Two studies examining sensory integration training met our inclusion criteria (Table 25).266–268 An RCT from Fazlioglu and colleagues266 examined the effects of a sensory integration protocol on low-functioning children with autism ages 7 to 11 years who had not previously received sensory integration therapy. The intervention program used in this study was based on “The Sensory Diet” and included a prescribed schedule of somatosensory stimulation activities targeting 13 behaviors across sensory modalities and motor skills development and conducted in a specially arranged sensory room. Results indicated that the difference between treatment and control groups was significant at outcome, but not at baseline, with children receiving sensory integration intervention showing significantly fewer sensory problems at followup than children in the control group.

In contrast to Fazlioglu’s more traditional play-based, child-directed sensory integration intervention, Jung et al.267,268 used a virtual reality – tangible interaction system sensory integration training protocol in 12 five- and 6-year-old children with autism The sensory integration components comprised less than a third of the intervention described, and no outcome measures were reported related to sensory integration activities.

To study auditory integration intervention (Table 24), Corbett et al.270 used a double-blind, placebo-controlled crossover design on the effects of Tomatis Sound Therapy on language skills in children with autistic disorder, ages 3 to 7 years who had not previously had auditory stimulation treatments. In the treatment condition, children listened to music passed through an electronic ear for attenuation and modulation for two hours per day in accordance with the Tomatis Method protocol. The protocol was divided into four blocks, each lasting 3 weeks. In the placebo condition, children listened to commercially produced music in the same blocked schedule. No improvements in receptive or expressive language were related to treatment conditions, as the increase in scores over time (i.e., pre-first condition, midpoint, post-second condition) was not different between groups (i.e., Treatment/Placebo and Placebo/Treatment).

Mudford and colleagues studied AIT in children with autism ages 5 to 13 years (mean age= 9.4);269 all children had significant language delays and low adaptive behavior levels. No significant benefit of AIT was found (Table 24).

One study investigated the effects of music therapy in children ages 3 to 5 years (mean = 51.2 months) with autism in comparison to the effects of play-based sessions on joint attention behaviors with results reported in two papers.89,271 A crossover design was used such that all children completed both music and play interventions, with treatment order randomly assigned. Sessions in both conditions were divided into 15 minutes of undirected child-led activities, followed by 15 minutes of directed activities according to a semi-flexible treatment manual developed for the study. There were no significant between group differences on the Pervasive Development Disorder Behavior Inventory, though both groups improved with time. Results from the Early Social Communication Scales, reflecting growth in joint attention skills, suggested that music therapy was significantly more effective than play sessions. Change scores pre- to post-music therapy were significantly greater than change scores pre- to post-play sessions.

In the second paper,89 treatment sessions were coded for emotional and motivational responsiveness (i.e., joy, emotional synchronicity, initiation of engagement) toward attunement promoted by the therapist and for responsiveness (i.e., social invitation and interpersonal demands) toward the therapist’s initiation of interaction during joint attention episodes within selected four-minute segments of four treatment sessions. More joy, emotional synchronicity, and initiation of engagement were observed during music therapy than play sessions. In addition, children had significantly more compliant behavior and significantly fewer episodes of no response behaviors in the music therapy condition.

Additional Allied Health Interventions

A number of studies addressing interventions such as animal-assisted therapies and assistive tools met our review criteria.

Content of the literature. We found eight studies272–279 addressing additional allied health interventions. Two papers report similar interventions and participants recruited from similar locations and may include overlapping participants.275,277 Treatment duration ranged from 47 days278 to 21 weeks,279 and participants ages ranged from 3 to 18 years.273 Table 23 provides additional study details. All studies in this section were considered poor.

Summary of the literature. Bass and colleagues conducted a randomized controlled trial investigating the potential benefits of therapeutic horseback riding sessions in 34 children with ASDs assigned to either a horseback riding group (n=19, mean age 6.95 yrs ±1.67) or wait-listed group (n=15, mean age 7.73 yrs±1.65).276 Children in the riding group attended sessions of 1 hour per week over 12 weeks. Scores on four of the five Sensory Profile subscales (sensory seeking, inattention/distractibility, sensory sensitivity, and sedentary), along with the overall Sensory Profile mean score, showed a significant between-group difference. The overall Social Responsiveness Scale score and social motivation subscale also showed a significant between-group difference.

A nonrandomized controlled trial investigated the effect of movement therapy on behavior among children with ASDs.274 Children were recruited from specialized schools and underwent 30-minute movement training sessions twice a week for 2 months. Behavior was recorded during the first and last meetings for 6 one-minute periods, in 10-second sample units. Eight behaviors were assessed: stereotypical behaviors, wandering, responding to touch negatively, on-task passive, on-task active, eye contact, social relatedness towards the teacher, and resisting the teacher. Of these, four exhibited significant changes: a decrease in wandering, responding to touch negatively, resisting the teacher, and an increase in on-task passive behavior.

Two prospective case series focused on the effect of incorporating colored overlays to assess effects on reading in children with ASDs.275,277 In the initial study,277 participants attended one session with researchers and performed reads of 30 seconds and 1 minute. Children with autism read significantly more words per minute with than without colored overlays. Seventy-nine percent of children with ASDs showed improvement in reading with colored overlays.277 In a later study,275 Ludlow and colleagues conducted three individual tests to assess the therapeutic benefits of colored overlays when reading; results showed more words read using overlays and slightly better performance on a picture matching task using overlays.

One prospective case series aimed at identifying the effect of therapeutic sessions incorporating animal interaction among children with ASDs.272 Therapy sessions focused on facilitating sensory integration, language use, sensory skills, and motor skills, with each participant attending at least two sessions of each type over 15 weeks. Results indicated that participants engaged in significantly greater use of language and social interaction in the therapy sessions incorporating animal interaction than in the standard occupational therapy sessions. In another series, Carmody and colleagues examined the effect of eyeglasses with specialized prism lenses.273 Children were recruited from a child development center, and each child was assessed for 20 to 30 minutes by a single interviewer, with trials of 60 to 90 seconds without glasses and with each of the lens orientations. Participants showed a significant improvement in a ball catching when wearing lenses (64 percent compared with 20 percent without lenses).

A 21-week study included 16 boys with ASDs and assessed the effects of a water exercise program on social skills.279 Participants had either Asperger syndrome (n=8) or high functioning/mild autism (n=8) and showed decreases in antisocial behavior following the swimming program. Finally, Laud and colleagues conducted a prospective case series that included a systematic feeding program and oral motor therapy aimed at improving feeding behaviors among children with ASDs.278 Participants underwent 3 hours of behavioral training and 1 hour of oral therapy at least 5 days per week. Behavioral training consisted of systematic meal sessions with individualized behavior protocols while the oral training was conducted by an occupational therapist or speech pathologist to determine skill and safety while eating. Significant changes from admission to discharge were found in the following areas: increases in acceptance and grams consumed, increase in refusal behavior, and a decrease in negative vocalizations.

KQ1. Effects of Treatment on Core and Commonly Associated Symptoms in Children With ASDs: Complementary and Alternative Medicine (CAM)

CAM Interventions

As noted, studies of CAM interventions meeting our criteria addressed acupuncture and massage, including qigong massage. Table 25 summarizes critical aspects of studies of CAM interventions addressing KQ1.

Content of the literature. We found seven studies280–286 of CAM interventions meeting our inclusion criteria; two studies282,283 likely contain overlapping participants. Interventions occurred in the clinic280,282,283,286 and home,282–284 and studies addressed massage,284–286 qigong massage,282,283 and acupuncture.280,281 Treatment duration ranged from 1284 to 9 months,280 and participants’ ages ranged from 3 to 10 years. Among the seven CAM studies, two were fair quality and five were poor.

Summary of the literature. Studies assessing massage focused primarily on sensory impairments; a series of studies from Silva et al. assessed qigong massage using similar approaches and potentially including overlapping participants, though the exact overlap is unclear. A 2007 RCT283 extended a 2005 case series (not included in the present review) of 8 children with ASDs; the RCT283 included 15 participants with autism diagnosed according to DSM-IV criteria (mean age at first assessment=4 years, 10 months). Children were stratified into three cognitive groups according to Batelle Development Inventory scores and randomly assigned to treatment or control within each group. Massage treatment for the eight intervention group participants consisted of 11 different qigong massage movements. A trained practitioner delivered massage twice a week in the clinic for two 5-week periods separated by 5 weeks of no practitioner-delivered therapy. Parents were trained to provide massage at least once daily during the 5 week practitioner-delivered and interval periods; parents were also tested to ensure accurate administration of massage. The seven control children received special education, and four were concurrently receiving speech therapy.

Total Sensory Profile scores improved in the massage group by an average of 5.4 points, with an average worsening in control group of 2.7 points (p<0.01); average improvement in daily living skills was 9.8 months compared with 0.9 months in the untreated group (p<0.02) and 10.0 months compared with 4.7 months in control participants (p<0.04). No significant differences between groups were seen in VABS language and motor development scores or on the Autism Behavior Checklist. Children with bowel or sleep difficulties in the treatment group improved according to parent report.

A 2008 case series assessed whether individuals completing an 80 hour training program could deliver qigong massage yielding positive adaptive behavior and sensory outcomes.282 The qigong sensory training curriculum for trainers included didactic and experiential instruction; during the 5 month intervention, each of the 18 trainers worked with two families with a supervisor present for several sessions to assess treatment fidelity; several treatment sessions were also videotaped to inform training efforts. Participants showed gains in sensory and adaptive behavior, and most trainers were considered to have mastered the training program.

A later multisite RCT assessing a similar qigong methodology modified for application in early intervention programs285 included 46 children between the ages of 3 and 6 and evaluated participants in both the home and preschool settings. Children were randomized to either Qigong Sensory Training, comprising 10 hours of practitioner-delivered therapy in addition to daily parent-delivered massage. Practitioners, including some early intervention personnel, received training and weekly supervision during the intervention period. Six therapy sessions were also videotaped over the course of the project to assess treatment fidelity. Assessments were conducted prior to the intervention, after the final massage session, and 5 months after the final session (parent-report) to evaluate maintenance of effects.

The Qigong group (n=25, mean age=65.2 ± 20.7 months) and wait list control group (n=21, mean age=53.3 ± 18.7 months) differed significantly in parent-rated Pervasive Developmental Disorder Behavior Inventory Social/Communication and Autism composite scores (p<0.05) as well as teacher-rated Sensory scores (p<0.01) at baseline. Scores from pre- to post-treatment for the Qigong group improved significantly on all measures (p<0.00); only change scores on the teacher-rated Pervasive Developmental Disorder Behavior Inventory maladaptive behavior subscale reached significance for the control group (p<0.01). Parent-rated data illustrated a significant overall treatment effect on outcomes and in adjusted comparisons on the Sense and Systems Checklist and Pervasive Developmental Disorder Behavior Inventory sensory, maladaptive behavior, social/communication, and autism domains; effect sizes (partial η2) were considered to be large. Treatment effect for teacher rated data was also significant overall, and adjusted comparisons found significant effects (moderate effect size) for the Pervasive Developmental Disorder Behavior Inventory social/communication composite and the Autism Behavior Checklist. Descriptive data collected 5 months after the final massage session in 19 participants available for followup showed significant differences in all scores on parent-rated measures at all three assessment times.

In an RCT of Thai traditional massage, 60 autistic children between the ages of 3 and 10 were randomized to receive 8 weeks (16 total sessions) of either sensory integration therapy (n=30, mean age 4.48) or sensory integration therapy plus massage (n=30, mean age=4.84).286 The same occupational therapist conducted sensory integration sessions, which were individualized for each child but followed established principles. One masseuse delivered all massage sessions following a standard protocol. Both groups saw improvements in behavior with significant differences between groups from baseline to week 8 noted in parental report of conduct problems (p=0.03).

In a one-month study of pre-bedtime, parent-delivered massage,284 twenty children with autism (mean age ± SD=5.2 ± 1.8, range 3–6 years) as diagnosed according to the DSM III-Revised criteria were randomly assigned to nightly massage or reading attention control groups. Assessments, conducted on the first and last days of the study, included subscales of the Conners Teacher and Parent Scales, behavioral observations in the classroom and playground by blinded research associates, and parent-maintained sleep diaries. Teachers’ blinded ratings on the Conners Scale showed greater improvement for children in the massage group on the Emotional Index and the DSM-IV criteria for inattentiveness (p<0.05). Parent ratings were significant for these subscales as well as the ADHD and restless-impulsive behavior indices (p<0.05). Observation ratings across the playground and classroom illustrated a decrease in stereotypical behaviors and increase in on-task behaviors in the massage group. Poor sleep behaviors also exhibited greater declines among the massage group according to parent report.

Two studies meeting our criteria addressed acupuncture. An RCT conducted in Egypt280 assessed the effects of scalp acupuncture and language therapy compared with language therapy alone on elements of attention, receptive and expressive semantics and syntax, phonology, and pragmatics. Children included in the 9 month study were between the ages of 4 and 7, were diagnosed with autism and delayed language development, and had Childhood Autism Rating Scale scores ≥30. Children randomized to language therapy plus acupuncture (n=10) received language therapy twice a week and scalp acupuncture twice a week for 2 months, followed by a 2 week rest period for the 9 months of the study. Children randomized to language therapy alone (n=10) received treatment twice a week for the 9 month study; further detail on the content of language sessions was not reported. The Arabic Language Test was used to measure outcomes.

The acupuncture and language therapy group showed significant pre-post gains in attention (p=0.001) and receptive (p=0.001) and expressive semantics (p=0.021). Children receiving language therapy only improved in attention, cognition, and receptive semantics from pre- to post-treatment, but the magnitude of improvement is not reported. Outcomes between groups were significant for attention (p=0.008) and receptive semantics (p=0.034).

A 6 week RCT of the seven star needle acupuncture technique281 similarly assessed changes in language function as well as social and stereotyped behaviors. The acupuncture technique uses a specially designed dermatoneural hammer housing seven needles in the shape of a star. Thirty-two children with ASDs (diagnosis not specified) included in the study were randomized to treatment (five 5–10 minute daily sessions per week for 6 weeks) or a wait list control group. Assessments were conducted before and after the 6 week study period and included a Parent Rating Questionnaire designed for the study and evaluating language, social, and motor functioning and stereotyped behaviors. EEG data, assessed by a rater blinded to treatment status, measured changes in neural processing.

Children in the treatment (n=16, mean age=6.85±1.76, mean IQ=84.06 ± 15.75) and control (n=16, mean age=6.89 ± 1.77, mean IQ=86.82 ± 19.91) groups were matched in age and IQ as measured by the Test of Nonverbal Intelligence. Parents of children in the treatment group reported significantly greater overall improvement than parents of control group children as well as greater improvement in language, especially related to sentence length and speech clarity and frequency. Social interaction, including gains in eye contact and facial expression, was also significantly improved for the treatment as compared with the control group. Changes in EEG data, which were available for 9 control and 7 treatment group children, suggested that the brain’s information processing functions were affected by the seven star treatment.

KQ2. Modifiers of Treatment Outcomes

Understanding the degree to which child characteristics (i.e., specific ASDs related difficulties and skills), treatment factors (e.g., type, duration, intensity), and systems (e.g., family, community) influence response to treatments could improve targeting of treatments to the appropriate children and circumstances. However, with rare exceptions,259,260,287 few studies are designed or powered to allow analysis of heterogeneous effects in order to identify true modifiers of treatment effect. Although we sought studies of treatment modifiers, only one included study 259,260 actually demonstrated true treatment modifiers based upon appropriate study design and statistical analysis. One other study 287 was designed to examine the role of provider on outcomes, but showed no difference, possibly because it was underpowered to do so.

This first study259 included an analysis of initial characteristics of the children demonstrating that children who were low in initial object exploration benefitted more from RPMT, which explicitly teaches play with objects, while children who were relatively high in initial object exploration demonstrated more benefit from PECS. These results were maintained at 6 months. An additional analysis based on this study260 showed greater increases in generalized turn taking and initiating joint attention in the RPMT group than in PECS. The increased benefit for RPMT in join attention was only seen, however, in children who began the study with at least some initiation of joint attention. Specifically, children most likely to benefit from RPMT in increasing joint attention had demonstrated at least seven acts of joint attention in the pre-intervention assessment. RPMT was also superior in this analysis in increasing object exchange turns.

One study287 explicitly sought to examine the impact of provider choice (parent versus professional) using similar interventions in an RCT. The study did not show a difference in outcomes for children receiving UCLA/Lovaas protocol-based intervention in a clinical setting versus at home from highly trained parents. both clinic and parent groups received over 30 hours of intervention weekly and no group differences related to IQ, language, adaptive behavior, or other outcomes were seen. Children in both groups demonstrated substantial gains in a number of areas. Nonetheless, the results do provide further evidence of response to treatments anchored in the UCLA/Lovaas method, with some children demonstrating rapid acquisitions of new skills and change in IQ. Other studies not specifically designed to examine modifiers have also compared parent to clinic-based interventions 126,132,288 and demonstrated equivalent group change when delivered in the same intensity.

Other studies in this section are those in which potential correlates were identified that may act as true moderators, but not in the context of studies designed and powered to identify modifiers. These potential moderators should be assessed in properly designed and powered studies for this purpose.

Behavioral Interventions

Frequency, duration, and intensity. Apart from the studies described above, a number of potential correlates of treatment effect are observed in the existing literature and should be studied further. The most commonly noted characteristics as potential correlates of effectiveness in the study of behavioral interventions are treatment intensity and approach (e.g., parent-led versus clinician-led) as well as baseline measures of child characteristics, including IQ, language and verbal skills and severity of the autistic disorder.

Vismara and colleagues178 found equivalent results across training (distance vs. in person) modalities for providers of ESDM treatment. When examining characteristics of UCLA/Lovaas-based intervention, Luiselli et al. found that months of treatment was significantly related to language gain, but numbers of hours per week and total hours of treatment were not.121 Intensity of supervision within UCLA/Lovaas-based treatment has also been demonstrated to be positively correlated with change in cognitive in cognitive abilities, although not other skills domains, within one treatment study.126,132

Because of the potentially increased efficiency and desirability of having parents provide intervention to their children in their own homes, several studies have reported on varying approaches to preparing parents to provide behavioral therapies. Sofronoff et al.174,175 conducted a parent training intervention with parents of children ages six to twelve years with Asperger syndrome diagnoses. Parents either (a) participated in a one-day workshop, (b) attended six weekly 1-hour individual sessions conducted by master’s or doctoral students in psychology, or (c) were placed in a waitlist control group. Components of the intervention were the same in both treatment groups and involved psychoeducation, comic strip conversations and social stories introduction, and management techniques for externalizing behaviors, rigid behaviors, and anxiety. Parents from both intervention groups reported significantly better social skills in their children than did parents in the waitlist control group at both 1-month post-treatment and 3-month followup; at 3-month followup, parents from the individual session group reported marginally better social skills for their children than did parents from the workshop group, suggesting little modification of effect by intensity (one day vs. weekly training).

Finally, in a case series assessing an intervention intended to improve either joint attention or symbolic play skills, a teacher-led approach was compared with one in which the child took the lead and found positive effects associated with the teacher leadership.158

Child characteristics. Several characteristics of the child have been assessed to determine whether there were identifiable variables associated with positive outcomes in intensive behavioral interventions.

Cognitive abilities/IQ. The most commonly reported characteristic investigated relates to pretreatment cognitive abilities/IQ. Several investigations have noted that pretreatment IQ and language predicts IQ at followup within the context of UCLA/Lovaas-based methodologies.101,104,115,124,287 However, other studies have suggested having a lower IQ at initiation of treatment is related to increased change in IQ over time127 or and change in response to intervention102,126,132 within this same methodology. In contrast to UCLA/Lovaas-based methodologies, parent training interventions for teaching early social communication skills demonstrate that children with lower language levels and/or lower IQ at baseline may actually benefit more from this intervention.108,109 Some data from Pivotal Response Training studies suggest that less impaired children do better in response to offered parent training.117

Language/communication skills. Baseline language/communication skills may also correlate with treatment success, with studies generally suggesting a benefit for communication skills, including changes in ASDs classification associated with baseline language skills in an ABA-based approach.102, 124 In one RCT155, 156 comparing the use of targeted joint attention intervention to development of symbolic play skills, children with initially higher levels of expressive language showed greater growth in expressive language from pre-intervention to 12 months post-intervention. Among children with lower expressive language initially, those in the joint attention group showed significantly greater improvements in expressive language. In addition, joint attention initiations, responding to joint attention, the duration of child-initiated joint attention, average highest level of play, total number of symbolic play types, and initial receptive language age all predicted greater gains at 6 and 12 months post intervention.

Similarly, social skills studies have found verbal skills, either verbal comprehension (using the Verbal Comprehension Index) or expressive communication skills to be associated with social skills at outcome. Children with higher verbal comprehension scores who participated in the Social Story intervention145 made larger gains in the evaluated game play skills, while children with extremely low verbal comprehension scores did not. Social Stories, an intervention program that relies heavily on the child understanding information presented in a written format, may not be as effective for children with low verbal comprehension abilities. In another study139 pre-treatment communication skills, as measured by VABS Communication domain and Verbal IQ, were associated with social skills at outcome (VABS Socialization) in both a Lego treatment group and the treatment as usual control group (but more so in the Lego group).

Autism symptom severity. Some evidence indicates that specific constellations of symptoms related to ASDs may be important in understanding response to treatment. Social responsiveness and imitation skills have been suggested as skills that may predict improved treatment response in UCLA/Lovaas-based approaches,287 whereas “aloof” subtypes of ASDs have been suggested to be associated with less robust changes in IQ,107 and lower baseline symptom tallies have also been demonstrated to be related to specific gains.104 Other studies have seen specific improvement with UCLA/Lovaas-based intervention for children with PDD-NOS vs. Autistic Disorder diagnoses,114 which may be indicative of baseline symptom differences. However, many other studies have failed to find a relationship between autism symptoms and treatment response.

Two social skills studies139,148 looked at the diagnosis of participants (PDD-NOS vs. high functioning autism vs. Asperger syndrome for one study, autistic disorder vs. Asperger/PDD-NOS for the other) as a potential modifier of treatment effects and failed to find any significant direct effects. However in the study evaluating the social adjustment enhancement curriculum,139 the results on a measure of theory of mind were no longer significant when the one participant with PDD-NOS was excluded.

Age at identification/initiation of treatment. Some evidence suggests that children initiating treatment at earlier ages may benefit more from UCLA/Lovaas-based intervention;115,129 however, other explicit comparisons have not found this same relationship for UCLA/Lovaas-based approaches121 and age at initiation of treatment may in fact be confounded by type of treatment initiated.129

Neurobiological and genetic variation. Only one of the included and reviewed studies examined the relations between potential underlying neurobiological markers/variation and this study simply indexed head circumference as a measurement within design102 and this did not appear to be related to outcome.

Family characteristics. Although family characteristics were rarely reported in the behavioral literature, in one study of a parent-directed play interaction, change in child behavior was not significantly predicted by whether parents perceived their child having a causal role in their own behavior or the parent having a causal role in their child’s behavior,154,161 but parent positive affect, measured through behavioral coding was positively related to parental reports of child adaptive behavior and negatively related to parental reports of child challenging behaviors.

Educational Interventions

Child characteristics. In educational interventions, baseline IQ and receptive language predicted rate of progress in one study.195 In a comparison of a home-based intervention plus center-based (intervention) to a center-based only (control) educational intervention,194,198 no girls in the intervention group improved in IQ or on the Preschool Behavior Checklist. In the control group, one girl improved in IQ, and 2 improved on the Preschool Behavior Checklist. Improvement in IQ in the intervention group was higher with low socioeconomic status, younger age, and high family stress. Improvement in Preschool Behavior Checklist scores was associated with younger age in the intervention group. This study also considered the potential effect of family stress and non-English speaking in the home and found no effect on outcomes.

Family characteristics. One study measured parental stress and its association with outcomes in four different teaching interventions (reinforcement-based interventions, special nursery, speech and language therapy, and parent education programs),196 and found that parenting stress was not associated with gains seen in interventions that required less total time but reduced the gains made by those interventions that required more total time. Moreover, evidence suggests that at lower levels of parenting stress, higher time intensity interventions are more effective than lower time intensity interventions. For the lower parenting stress group, higher time intensity interventions significantly improved intellectual functioning and educational functioning but not adaptive functioning as measured by the VABS.

Medical Interventions

No modifiers of treatment outcome were identified in studies of antipsychotic medications in ASDs, though one case series of risperidone use205 reported a correlation between weight gain in the first month and final weight gain. We were, however, able to identify papers that included modifier data for stimulants and SRIs. None assessed measures of frequency, duration or intensity of treatment specifically; nor did they assess training experience of the investigator or clinician providing care.

Child and family characteristics. Some characteristics of the family and child were found to be useful in predicting treatment success, including a history of psychiatric diagnoses in the family, early verbal skills in the child,224 and, potentially, genotype for predicting lack of treatment response or adverse reactions.214,223 Several studies of stimulant use highlighted differences in effectiveness by diagnosis type228–230,232,233 finding that children with Asperger syndrome were typically more responsive to psychostimulant treatment than those with autistic disorder. The presence of co-morbid intellectual disability was associated with lower treatment response in one study.232 Two studies sought to examine differences in treatment response by gender and found none.231,232 Details are provided below.

One open-ended study of fluoxetine treatment in 129 children assessed response qualitatively. Subjects with an average age of 4.5 years were diagnosed with a PDD by DSM-IV criteria corroborated by Childhood Autism Rating Scale and ADOS and were then treated with fluoxetine as felt to be clinically indicated, with up to 72 months of treatment. They analyzed potential predictors of good/excellent response, including family history and subject characteristics. When comparing to subjects with fair/poor response, they found an increased rate of family history of affective disorder (major depressive disorder and bipolar disorder) and “unusual intellectual achievement” in those subjects with a good/excellent response. Subjects with good/excellent response also showed an increased rate of verbal language before starting fluoxetine. Finally, subjects with a good/excellent response were more likely to have hyperlexia, an early or precocious interest in letters or numbers. They did not find a significant relationship between history of regression and response to fluoxetine. They also did not find a relationship between dose of fluoxetine and likelihood of response.

One additional retrospective case series described response to various serotonin reuptake inhibitors, primarily sertraline, citalopram, paroxetine, and fluvoxamine in 89 children and adolescents with ASDs by DSM-IV criteria.224 The CGI-Improvement was used to gauge medication response, with 40 subjects rated as at least “much improved.” Family history of ASDs was significantly associated with positive treatment response. Other possible moderators showed no association with response, including family history of depression or anxiety, subject diagnosis, concurrent medications, specific SRI prescribed, and indication for SRI initiation, whether for anxiety, repetitive behavior, aggression, or depression.

In the double-blind cross-over trial of MPH in 66 children with PDDs,228–230 authors found no effect of age, IQ, weight, or diagnosis on teacher- or parent-rated hyperactivity subscale scores/Swanson Nolan and Pelham-Fourth edition rating scale (SNAP-IV)/Children’s Yale-Brown Obsessive Compulsive Scale-PDD scores. Children with Asperger syndrome/PDD-NOS (n=19) showed a trend of being more likely to be classified as responders to both placebo and MPH than those with autism. Response to each dose of MPH was significantly superior to placebo in the autism subgroup but not for the Asperger/PDD-NOS subgroup.

Similarly, Posey et al., in their retrospective review of 80 children with PDDs treated with guanfacine,232 found that subjects with PDD-NOS and Asperger syndrome showed a greater rate of global response than those with autistic disorder. Those without intellectual disability showed a higher rate of global response to guanfacine (37.5 percent) than those with co-morbid intellectual disability (17.9 percent). They also identified that the responders were less aggressive at baseline by the CGI severity item. Finally, Stigler et al., in their study on the effectiveness of psychostimulants in 195 children with PDDs, found that children with Asperger disorder were found to be more likely to respond to treatment than those with autistic disorder or PDD-NOS. Those children on concomitant medication were also found to be more likely to respond to treatment; the study did not find any association between stimulant type, gender or IQ and response to treatment.233

Of particular interest currently in the study of medical treatment of autism is the possibility of genetic modifiers that might be used to target treatment choices. One prospective ten-week case series of escitalopram sought to identify pharmacogenetic modifiers of treatment response in the challenging behavior domain as measured by the ABC-C-Irritability.223 Fifty-eight subjects with ASDs corroborated by ADI-R and a minimum ABC-C-Irritability score of 12 underwent a forced dose titration of escitalopram from 2.5 mg daily increasing weekly to 5 mg, 10 mg, 15 mg, and 20 mg, essentially twice the dose equivalent of citalopram given that escitalopram is the active component of racemic citalopram. Pre-designated dose-limiting side effects included sleep disruption and an increase in ABC-C Irritability or Hyperactivity subscales of 10 points over the previous week.

Subjects were also genotyped at several polymorphisms in the serotonin transporter gene. Average daily doses of escitalopram were 10.8–12.4 mg and did not differ across genotype groups, which reflects the fact that most subjects in all genotype groups could not tolerate the maximum dose. One genotype group, designated a priori as the low-expression genotype group, showed diminished response to escitalopram,223 with a particularly striking difference in a subgroup of the low-expression genotype based upon previous association with platelet serotonin uptake measures.289 The low-expression genotype group also had verbal and nonverbal IQ scores that were 25 to 26 points lower than the other subjects; although this was not described as a statistically significant difference. Genotype groups also differed with respect to percentage of Caucasian subjects, with the high-expression genotype group only containing Caucasian subjects; although the pattern of results was reported to be the same in a Caucasian-only analysis.

In other medical studies, children treated with DMSA250,255 had greater improvement in core and associated ASDs symptoms if they were older than age 5. Children with lower initial ADOS scores (below the 50th percentile) also had greater improvements than were seen among children with initially higher scores. It is unclear whether either of these modifiers is significant in the context of a trial with no overall difference in response between DMSA and placebo. In one study of a ketogenic diet,245 the two patients with the greatest improvement were those whose baseline condition was classified as mild using CARS scores, and those with severe autistic behavior showed substantially less improvement. In a study assessing omega 3 fatty acid use, a negative correlation between docosahexaenoic acid level and CARS before treatment was observed in nonresponders. Finally, attempts to identify subgroups of children for whom oral immunoglobulin was successful in treating GI symptoms and associated autism symptoms found no effect of age, regression onset of symptoms, or predominant bowel type.64 The treatment was uniformly ineffective.

CAM Interventions

One CAM study285 noted correlations between changes in scores on sensory measures after qigong massage therapy and positive behavioral changes.

KQ3. Early Results in the Treatment Phase That Predict Outcomes

Early Identifiable Changes Predicting Response/Outcome

Information about early response to treatment, or lack thereof, can be essential to guiding treatment selection, implementation, and modification. The reviewed literature offers almost no information about what specific changes predict long-terms outcome and response. Some evidence indicates that early response to both UCLA/Lovaas-based approaches and ESDM intervention in terms of changes in IQ over the first year of treatment predicts, or accounts for, longer-term change in IQ.287,290 However, findings also suggest that while gains in the cognitive domain might be accounted for primarily within the first year of treatment, changes in adaptive behavior in response to these same interventions may occur over a longer time frame110,133,287,290 if they occur at all.105

KQ4. End of Treatment Effects That Predict Outcomes

One study meeting our criteria addressed whether outcomes measured at the end of treatment could predict longer term functional outcomes. An RCT comparing joint attention and symbolic play interventions155,156 included 58 children with autism between 3 and 4 years of age. Investigators assessed language development, joint attention and play skills, and mother-child interactions at pre- and post-intervention and 6 and 12 months after the end of the 5 to 6 week intervention. Children in the symbolic play and joint attention groups showed significantly greater growth expressive language over time than did participants in the control group (p<.01, moderate to large effect sizes). Growth in receptive language was not significantly affected by the intervention from pre-intervention to 12 months post-intervention.

Children in the both the joint attention and symbolic play groups showed significantly more growth in initiation of joint attention and duration of child-initiated joint attention than did the control group (p<.01 to <.05). Children in the symbolic play group also showed significantly more growth in play level than did children in either the joint attention (p<.01) or control (p<.001) groups.

The investigators also assessed differences in the amount (total hours) of intervention services (speech and overall) children in the three groups received post-intervention, with children in the control group receiving significantly more hours of overall services than either the joint attention or symbolic play groups (p<.05 and <.01, respectively); differences in hours of speech interventions received were not significant. Only the duration of child-initiated joint attention episodes was related to hours of intervention received post-treatment, with children with fewer hours of overall services showing greater growth in child-initiated joint attention episodes.155 Hours of speech interventions received did not affect growth in skills.

KQ5. Generalization of Treatment Effects

Parents and clinicians wish to know whether outcomes observed in the treatment setting are likely to also be found in other settings and are thus generalizable. To try to assess generalizability, we recorded the degree to which studies collected outcomes data in multiple settings when it would be appropriate. For example, we noted when studies occurring in the clinical setting also collected data in the home or school. We also noted the period of time for which studies collected data.

For some areas of intervention, outcomes are primarily measured outside of the setting in which the treatment takes place. This includes, for example, behavioral interventions for associated conditions like anxiety, in which treatment occurs in therapy sessions. For these interventions, outcomes are usually measured using parent, self, and/or teacher report at home, at school and in the community. Studies of these behavioral interventions do, in fact, report positive outcomes in children’s natural settings to mirror what is seen in the treatment setting; however, these outcomes are generally identified with parent report rather than the preferred direct observation. Few behavioral interventions continue to monitor children in their studies and so maintenance of the results over time is largely unknown.

In a number of studies of social skills interventions154–155,162–163, parents reported positive outcomes outside of the treatment session, but parents were not blinded to intervention status. Participants in cognitive-behavioral-ecological137,138 and Lego therapy143 were shown to have improved social skills outside of the intervention settings. Although the parents of children involved in Children’s Friendship Training141 reported significant changes in child social behavior at home immediately following the intervention (as well as 3 months later), teachers did not report any changes in the children’s behavior at school. On the other hand, teachers of children involved in a social competency and social skills training program140 reported improvements in student behavior at school.

One study attempted to assess the ability of children to apply new skills across changing intervention conditions. Participants in Social Stories145 were able to generalize the social skills they learned while playing with one set of board games to a different set of board games. However these “generalization” trials were conducted by the same experimenter in the same room as the other assessments, so it is not clear whether the targeted social skills would generalize to more naturalistic settings with peers.

In medical studies, although the treatment is prescribed in a clinical setting, it is generally administered at home, and one would expect effects of medications to be observed in terms of behavior in the home and other settings. However, medical studies may be good sources of information on the duration of perceived effects. Unfortunately, other than case series data for risperidone177,203,204,206,213 demonstrating continued effects and side effects beyond six months. few data are available on longer term outcomes of medical treatment.

Two of the case series studies of SRIs are based upon durations of treatment longer than 6 months, but the general and sometimes qualitative ratings of change in these studies are difficult to compare with the results of the RCTs to understand if medication responses reported in RCTs are likely to be durable.221,224

KQ6. Drivers of Treatment Effects

No studies were identified to answer this key question.

KQ7. Treatment Approaches for Children Under Age Two at Risk for Diagnosis of ASDs

This section presents the results of our literature search and findings regarding the use of treatment approaches in younger children who are at high risk of developing autism based upon behavioral, medical, or genetic risk factors. Studies located typically included participants whose mean age exceeded 24 months; however, the studies address interventions which can be used with children under age 2. The average age for diagnosis of ASDs in the US is not until at least age 3, but a reliable diagnosis may be possible as early as age 2.291–293 Research suggesting that early intervention can improve outcomes has compelled investigators to consider intervening in very young children.115

We identified four papers178,290,294,295 with unique study populations addressing treatment approaches for very young children. Three studies were conducted in the US178,290,294 and one in the UK.295 Two of the studies were prospective case series,178,294 one was a nonrandomized controlled trial,295 and one was a randomized controlled trial.290 Table 26 summarizes outcomes for studies considered to be fair or good quality and employing comparison groups.

Table 26. Outcomes of interventions for children at risk for diagnosis of ASDs.

Table 26

Outcomes of interventions for children at risk for diagnosis of ASDs.

The RCT290 and nonrandomized trial295 were completed in a clinic setting with instruction to continue with parents at home. All children in the RCT met DSM-IV criteria as well as criteria on ADOS and Toddler Diagnostic Interview for diagnosis confirmation; the mean age of participants in the treatment and control groups was 23 months. The nonrandomized trial295 employed parent training techniques to teach social communication skills and included children with a mean age of 38 months in the intervention group and 34 months in the control group.

One of the case series describes an evaluation of techniques to train personnel to provide ESDM-based therapy and included children with a mean age of 33 months.178 The second case series294 was completed in the home and classroom and focused on social-communication and language outcomes; children assessed in the study were between 18 and 36 months. Among studies in this section, 1 was considered good quality, 1 fair quality, and 2 were considered poor.

The Dawson et al.290 randomized controlled trial evaluated the effectiveness of the ESDM for young children with ASDs. ESDM, a comprehensive, manualized intervention that blends ABA with developmental and relational approaches, was designed to be used with children as young as 12 months, delivered in the home, and to utilize parents as well as trained therapists.

After 2 years of intensive intervention (31 hours of intervention per week, 15 from a therapist and 16 from parents) children receiving ESDM treatment displayed significantly larger gains in IQ (when compared with a community sample receiving 18 hours of individual and group intervention ). Children in the experimental group also demonstrated significantly larger gains in terms of adaptive behavior skills (i.e., all areas except socialization) than controls. The authors also reported greater diagnostic shifts (i.e., from Autistic Disorder to PDD-NOS for seven (29.2 percent) children in the ESDM group and for one (4.8 percent) child in the community services group; two (8.3 percent) children in the ESDM group and five (23.8 percent) children in the community services group experienced a diagnosis change from PDD-NOS to ASDs); however, these shifts were not matched with clinically significant improvements in terms of ADOS severity scores nor measurements of restricted and repetitive behaviors (i.e., RBS scores).

While no replication of this study has been conducted, the model had been subject to an early effectiveness trial178 wherein the research team compared distance learning vs. live instruction for community-based therapists implementing intervention and training parents. Results suggest that both modalities of learning were effective in teaching therapists to implement and train parents, with significant child gains over time and across modalities; however, results also suggested that implementation with fidelity required specific and explicit supervision. Thus, while promising in terms of treatment efficacy and extension to a younger population of children with ASDs, training demands for broad implementation appear substantial. Further, the average age for enrollment was very close to 2 years of age. As such, concerns about how this model would apply to children closer to 1 year of age remain.

In another evaluation of an early intervention approach, parents of 51 preschool-aged children suspected of ASDs (mean age, intervention group = 38 months, mean age, control group = 34 months) participated in the Hanen More than Words program, as created by the Hanen Center either immediately (n = 26) or after a delay (n = 25).295 The program focused on weekly group instruction in enhancing interactions and facilitating communication. In addition to 20 hours of group intervention, parents received individual in-home feedback on three occasions. Operationalization of “suspected ASDs” was identification of language delay and some aspect of concern about social behavior by a pediatrician and/or a speech and language therapist.

Ultimately, this resulted in inclusion of children within intervention and control groups without ASDs, with the authors grouping PDD-NOS and other developmental concerns under a category of “non-core autism.” After the intervention period, reported language use was substantially higher for the intervention group, with both the core autism and non-core autism children demonstrating improvements. Parent use of taught strategies was also higher in the intervention group than in the comparison group but only for the children with core autism. No group differences were found for ADOS scores or behavior issues.

Notably, more children in the intervention group had ASDs, and the intervention group also received more “substantial intervention” outside of the treatment context. Thus, while demonstrating potential benefit for parent training in social communication for young children with ASDs, the unique impact of this program for specific children remains unclear.

Wetherby and colleagues’294 prospective case series served as a preliminary study for the Early Social Interaction Project, which emphasizes a parent-implemented individualized curriculum in a natural environment. The authors found significant within-group differences from pre- to post-test for 11 of the 13 social-communication measures on the Communication and Symbolic Behavior Scales Developmental Profile in the Early Social Interaction group (n=17). The post-Early Social Interaction group performed significantly better than the third-year contrast group (n=18) on three measures of social signals, rate of communicating, three measures of communicative functions, and understanding.

The third-year contrast group performed significantly better than the pre-Early Social Interaction group on all three measures of communicative means and on actions to others in play, but there were no significant differences on the three measures of social signals, rate of communicating, the three measures of communicative functions, understanding, and inventory of actions. The percentage of children who were verbal was 5.9 percent in the pre- Early Social Interaction group, 76.5 percent in the post- Early Social Interaction group, and 55.6 percent in the third-year contrast group.

These findings suggest that the Early Social Interaction project has a positive impact on ASDs symptoms, but because the groups were unable to be compared at pretest, we cannot conclude whether the benefits were due to Early Social Interaction or to normal maturation. Another limitation in the authors’ methodology is the lack of documentation of parental implementation in the home, given that the parents’ involvement is a significant factor in the effectiveness of Early Social Interaction treatment.


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...