The effectiveness of applied behaviour analysis interventions for people with Autism Spectrum Disorder

Publication Details

This systematic review considers the evidence for the effectiveness of interventions grounded in the principles of applied behaviour analysis for people with autism spectrum disorder.

Author(s): Marita Broadstock and Anne Lethaby, New Zealand Guidelines Group.

Date Published: 19 December 2008

Please consider the environment before printing the contents of this report.

This report is available as a download (please refer to the 'Downloads' inset box).  For links to related publications/ information that may be of interest please refer to the 'Where to Find Out More' inset box.


Search Results

There were 1517 references identified by the full search strategy. These included 55 unique articles (after excluding 96 duplicates) from the 151 which the Ministry of Education asked to be considered for eligibility in the review. These 151 papers were suggestions from submissions made in the consultation process following release of the draft ASD Guideline.

Of the total 1517, 1329 were excluded based on application of selection criteria to titles and abstracts, leaving 188 potentially eligible articles. These articles were retrieved as full text and 145 were found to not fulfil inclusion criteria and were also excluded, leaving 43 eligible publications. 

It should be noted that the additional 151 additional references provided by the Ministry of Education identified 12 of the included papers. Eleven of these were independently identified by the bibliographic search strategy developed for the review, and the other study was also identified in reference checking.

All 1474 excluded papers are listed in Appendix 2, appended by key reason for exclusion. The vast majority of studies (n=1409) were excluded for being wrong publication, commonly because they were not primary, controlled group studies relevant to the research question. Fifty-six were excluded as they considered the wrong intervention; that is, an intervention which was not predominantly ABA-based. Six studies were excluded as relating to the wrong population (eg, a sample of people with Rett’s Disorder). Three were excluded as investigating the wrong outcomes (eg, limited to measuring the impact on carers of people with ASD). No studies (meeting all other selection criteria) were excluded as “wrong study design” as all represented study designs at NHMRC (2008) evidence level III-2 or above (studies with a parallel control group). Note, single case experimental design studies and case reports/series were excluded as wrong publication.

The 43 eligible publications reported on 41 studies, which were critical appraised. Included publications are listed in Appendix 3.

Description of Studies

Included studies and levels of evidence

Of the 41 studies that met the review’s selection criteria, 21 were systematic reviews or evidence-based guidelines and 20 were primary studies. 

Of the 21 secondary studies, 10 reviewed RCTs and/or systematic reviews (Level I evidence), including 9 which had no design limitations on studies for inclusion, and 1 which sought to find ‘critical reviews’ only.

Eleven secondary studies were reviews of level III-2 evidence (and were classified as level III-1 evidence). These included nine reviews/guidelines which synthesised single case experimental design (SCED) studies, one which synthesised non randomised studies and one which included only higher quality evidence (systematic reviews and RCTs, but observational studies were also accepted if no other evidence was identified).

Of the 20 primary studies, there were eight RCTs (level II evidence), one quasi randomised study (level III-1 evidence), and three cohort studies and eight non randomised experimental studies (level III-2 evidence). For two of the primary studies, there were two articles published reporting on each study, with the second publication either reporting on outcomes at longer follow up (Eikeseth et al. 2002; Eikeseth et al. 2007), or reporting on different outcomes than the first publication (Yoder and Stone 2006a; Yoder and Stone 2006b).

Details on the included studies are provided in the full Evidence Tables (Appendix 4) and a summary of participants, settings, interventions, comparisons and outcomes in included studies (Table 2). These tables are ordered according to level of evidence (higher level studies reported first) and within each study design type, in reverse chronological order (most recent publications first), and, where necessary, in alphabetical order within year.


Participants in the included studies were diagnosed either with autism, ‘autistic disorder’, ASD or PDD NOS. One study included children with a range of psychiatric problems such as anxiety disorders, depression and attention deficit disorder, but results were available separately for children with autistic disorders. One review required at least one participant in the studies considered for inclusion to have ASD and another review required at least 75% of participants in included studies to have ASD. One study specified that participants have autism or ASD and specifically excluded participants with PDD and/or ASD. One study required participants to have both autism and mental retardation.

Most of the participants in the studies were preschool children. Two studies did not report age criteria, participants in 8 studies were described as ‘children’ without clarification, 6 studies included children in either preschool or elementary school, 7 studies included students of school age (one up to age 21 years) and 18 studies were limited to preschool children. Of these 18 studies, only 2 required children to be aged less than 3 years; another 3 studies included children aged between 1½ or 2 years and 3½ years.

Download: Table 2
     Summary of Quality Scores, Participants, Settings, Interventions, Comparisons & Outcomes
           in Included Studies [PDF 65KB]


Most of the included studies for the secondary studies used a variety of settings and results were not grouped according to setting, with the exception that one review was restricted to studies that implemented social skills interventions in a school setting (Bellini et al. 2007). Of the primary studies, all of the RCTs were undertaken either in university clinics or school/preschool classrooms. Most of the non randomised studies (n=7) were undertaken partly or wholly at home, sometimes with a mixture of parents and therapists implementing the intervention(s).


Primary studies

Interventions in the included studies varied widely. Twelve studies assessed the effects of different variants of ‘early intensive behavioural interventions’ (EIBI), but programme delivery varied. Five of the studies assessed the effects of clinic-directed EIBI. Another four studies assessed the effects of EIBI delivered in the home (usually by parents). One study assessed the effects of a ‘minimal intensity’ early behavioural intervention (Eldevik et al. 2006), and one study assessed the effects of a EIBI delivered in a mixed setting, home, community and school (Howard et al. 2005). The remaining EIBI study considered EIBI with either the addition of joint attention or symbolic play. Three studies evaluated the Picture Exchange Communication System. One of the PECS studies had two intervention groups; immediate PECS and delayed PECS (Howlin et al. 2007). The design of this study enabled the researchers to assess whether the effects were maintained in those having immediate PECS therapy. Other interventions included video modeling by direct teaching (n=1), the Early Social Interaction project, a parent implemented intervention aimed at increasing communication opportunities within daily routines (n=1), an approach that used parents to implement a ‘psycholinguistic and social-pragmatic approach’ that included behavioural techniques (n=1), a ‘behavioural treatment’ intervention focusing on vocabulary acquisition (n=1), and the Autism Preschool Program which offers carers support in behavioural and language development methods (n=1).

Secondary studies

The aim of six reviews/guidelines was to assess the effectiveness of any treatments for children with autism and it was possible to extract from the overall findings specific results pertaining to behavioural interventions. Four reviews assessed the effects interventions of ‘behavioural’ treatment. Four assessed the effects of various forms of EIBI. Some reviews of SCED studies focussed on particularly types of behavioural intervention that were broadly similar: video (self) modelling (n=3), functional communication training (n=1), interventions based on targeting social skills and communication (n=3). 


Twelve of the 20 primary studies had variants of an ‘eclectic’ approach. These were mostly defined as using a combination of approaches, and many included (but not exclusively) ABA approaches. One of these studies had two comparator groups: intensive eclectic therapy and non intensive eclectic therapy. Another of these studies compared low intensity eclectic therapy with low intensity EIBI. 

The comparators for the other eight primary studies were: educational software programme (n=1), EIBI (without the interventions joint attention and symbolic play) (n=1), low intensity EIBI (compared to high intensity EIBI) (n=1), unstructured play activities (within an EIBI approach) (n=1), parent directed EIBI (compared to clinic directed EIBI) (n=2), responsive education and prelinguistic milieu teaching (n=1), and no treatment (n=1). 

All 21 studies of secondary research (systematic reviews or guidelines) did not specify comparators in their evaluations of behavioural therapies, although conclusions in some of these reviews were based on a synthesis of controlled primary studies with comparators.

Summary of primary study comparators

The majority of the 20 primary studies (n=12) compared different variants of behavioural therapy with a more varied eclectic approach. Six studies had a behavioural control group of some type: one compared EIBI with an educational software programme using a computer, one compared EIBI using either programmes for joint attention or symbolic play skills with EIBI without these additional programmes, one compared high intensity EIBI with low intensity EIBI, one compared a video modelling intervention with an unstructured play behavioural intervention, two compared EIBI delivered in the clinic with EIBI directed by EIBI-trained parents. The remaining two designs included a comparison of ESI with no treatment (n=1), and a comparison of PECS with responsive education and prelinguistic milieu teaching (n=1).


Of the primary studies, most undertook to assess the effects of interventions on a range of skills and behaviours (commonly, cognitive/intellectual functioning, language skills, adaptive functioning, non verbal intelligence, academic achievement such as class placement, and communicative and social skills). Five studies also included results from outcomes not relevant to the scope of the current review. Seven studies assessed the effects of interventions only on aspects of communication and/or language and play. One study restricted the assessment of outcomes to cognitive functioning and behavioural symptoms. Another study assessed the effects of the intervention on a simple learning task, and attention and motivation to continue in the programme.

Of the secondary studies, 10 studies were either non specific about the outcomes assessed in the included studies or evaluated the effects of interventions on overall functioning and autistic symptoms. Four studies restricted their assessments to social skills and functioning, three solely considered challenging behaviour outcomes, one study assessed the effects of the intervention on both problem behaviour and communication, and three studies assessed intervention effects on a wide variety of outcomes (such as IQ, class placement, behavioural functioning, functional skills, social interaction and play, adaptive functioning).

Study Quality and Risk of Bias

Level I (systematic reviews, including RCT) studies were mostly of very good (++) or good quality (+): four had ++ quality scores, five had + quality scores and one had a fair (-) quality score. 

Of the RCTs (level II evidence), one had a ++ quality score, five had + quality scores and two had – quality scores.

Of the reviews of SCED studies (level III-1 evidence), two had ++ quality scores, seven had + quality scores, and one had – quality scores. Also providing level III-1 evidence, the quasi-randomised study (Kroeger et al. 2007) was of fair (-) quality. 

Most of the level III-2 studies had fair quality scores; 10 had fair (–) quality scores and 2 had good (+) quality scores.

Effects of Interventions

The clinical question underpinning this review broadly seeks to determine whether interventions and strategies based on ABA have effects on a wide range of outcomes in people with ASD. As many of the identified studies reported on the multiple outcomes, it was considered more efficient to group results according to type of intervention in the first instance rather than outcome, therefore avoiding repeating critiques of the same studies for different outcomes. The results from the primary studies are reported first as some of these studies were published more recently than the secondary studies and therefore may not have been included in the identified reviews.

Primary studies

Early intensive behavioural intervention studies compared with standard care

A number of studies compared EIBI with a control group of children receiving standard care or an eclectic approach (mostly a mixture of approaches and many also including some ABA approaches). One non randomised experimental study (level of evidence: III-2) of good quality (Eikeseth et al. 2002; Eikeseth et al. 2007) compared EIBI (based on the UCLA treatment model except that no aversives were used) with eclectic treatment designed to reflect best practices for serving children with autism. Intensity of treatment between EIBI and eclectic treatment groups was similar (initially, an average of 28 and 29 hours per week, respectively). Significant differences were reported for the EIBI group at one year follow up in IQ, language and adaptive behaviour when compared with the control group but not in daily living and socialisation. At 2½ years follow up, significant benefits were found for the EIBI group in IQ, adaptive functioning, communication, daily living skills, aggressive behaviour and socialisation when compared to control. There were no differences reported in socioemotional functioning between groups. Most of the gains appeared before one year follow up. There were also large individual differences in gains in the EIBI group (but none of the demographic variables predicted these individual differences).

Another non randomised controlled study (level of evidence: III-2) of fair quality (Zachor et al. 2007) compared an EIBI programme with an average intensity of 35 hours/week with an eclectic programme (hours not reported). After a year of treatment, the EIBI group had a significant advantage over the control group in language and communication and there was a trend towards a benefit in reciprocal social interaction.

Several non randomised studies (level of evidence: III-2) were identified that compared types of EIBI based at home/community (rather than clinic directed) with types of usual care/eclectic interventions (mostly based at preschool/school or in the community). One study rated as fair quality (Cohen et al. 2006) compared an EIBI programme administered in the community (with an intensity of 35 to 40 hours/week) with eclectic services in public schools (hours per week not reported) in 42 children. Children in the EIBI group had a significant benefit for IQ and adaptive behaviour compared with the control group. No difference between groups was found in language comprehension or non verbal skills. Children having EIBI were more likely to be placed in regular classrooms after three years of treatment. In a study of fair quality by Sheinkopf and Siegel (1998), treatment intensity differed (average of 27 hours for home-based EIBI and 11 hours for school-based eclectic treatment). At an average follow up of 18 months, children receiving EIBI had significantly higher IQ scores and less severe autistic symptoms than children in the control group. However, there were no reported differences between groups on frequency of overall symptoms or proportions remaining autistic. None of the children achieved ‘normal functioning’.

Magiati et al (2007) conducted a good quality cohort study which compared home based EIBI in a community setting with a specialist autism-specific school based nursery programme in preschool children. Programme intensity was high for EIBI and school-based groups (32 and 25 hours per week, on average at baseline, respectively). At two year follow up, there were no significant group differences in cognitive ability, language, play or severity of autism. A trend was found for the EIBI group (approaching significance) for daily living skills. However, there were large individual differences in progress with IQ and language level best predicting overall progress. No child was able to attend a regular classroom at the end of treatment without 1:1 support. By contrast, Remington et al’s (2007) fair quality cohort study reported a significant benefit at one and two years follow up for children receiving home-based EIBI when compared to those receiving usual care in responding to joint attention, IQ, mental age, daily living skills and a trend for motor skills. No group differences were reported for initiating joint attention, adaptive behaviour (composite), socialisation or communication.

One fair quality study (Eldevik et al. 2006) based in kindergarten/school setting (with treatment administered in separate room) compared EIBI with an eclectic intervention, both delivered at minimal intensity (12.5 and 12 hours/week respectively over 18 months). Participants had both autism and mental retardation. The EIBI group had a significant benefit in IQ gains, language and communication when compared with control. However, there were no significant differences reported between groups for non verbal intelligence, adaptive skills (composite), daily living and socialisation.

Another study also of fair quality (Howard et al. 2005) compared EIBI (administered in multiple settings at an intensity of 25 to 40 hours/week) with two types of control group: intensive eclectic services (1:1 ratio for 30 hours/week in public school classrooms) and a non intensive eclectic early intervention programme (small groups, 15 hours/week, using a combination of methods in community special education classrooms). After 14 months of treatment, children receiving EIBI had significantly higher scores on all skill domains (cognitive skills, non verbal skills, language, adaptive skills) except for motor skills than either of the 2 control groups.

Comparisons of EIBI with programme variations

Some of the primary studies assessing EIBI looked at within-behavioural treatment variations; one study compared EIBI interventions with varying programme content; two studies compared the intensity of EIBI (high intensity vs lower intensity), and two studies compared the means of delivery of EIBI (clinic vs parent guided).

EIBI contrasting programme content

A good quality RCT (level of evidence: II) (Kasari et al. 2006) compared the effects of two specific additions to their EIBI, known as the Early Intervention Program for Infants and Toddlers with Disabilities (EIP). The treatment groups had either additional teaching in joint attention skills or teaching in symbolic play skills in their EIP programme and the control group had EIP alone with intensive hours and nearly 1:1 staff ratios. Children in the joint attention plus EIP intervention initiated significantly more showing and responsiveness to joint attention on the structured joint attention assessment and more child-initiated joint attention in the mother-child interaction. The children in the play plus EIP group showed more diverse types of symbolic play in interaction with their mothers and higher play levels on both the play assessment and in interaction with their mothers. 

EIBI studies contrasting programme intensity

The intensity of behavioural interventions is an important factor in the determination of effectiveness and was considered by two primary studies. A cohort study (level of evidence: III-2) (Reed et al. 2007) rated as fair quality compared the effects of 3 different types of high intensity with low intensity EIBI on a variety of outcomes in preschool children after 9 to 10 months of home based treatment. As the numbers of participants in each of the high intensity programmes (Lovaas, Verbal Behaviour or CABAS) in the Reed et al (2007) study are insufficient to meet the inclusion criteria of this report, results are reported for the three high intensity programmes combined. There were significantly stronger gains in the high intensity EIBI group (average 30 hours/week) educational functioning relative to those in the low intensity group (average 13 hours/week) but gains in adaptive behaviour were smaller and not significantly different between groups. The difference between groups in intellectual functioning did not reach statistical significance. There was also no evidence of a change in the rating of autism severity across groups. Regression analysis indicated that there was an inverse relationship between the temporal input and the gains in the high intensity group, suggesting that the original requirement of 40 hours per week suggested by Lovaas (1987) may not be optimal. In this study, there were diminishing returns after an intensity of 20 hours per week. 

A randomised controlled trial (level of evidence: II) (Smith et al. 2000) of good quality (level of evidence: II) also compared the intensity of EIBI but with different implementation; high intensity clinic guided EIBI was compared with lower intensity parent directed EIBI. The study compared an average of 24.5 hours per week of clinic directed EIBI with a lower intensity programme of EIBI directed by trained parents (weekly hours of training not reported) over a period of 4 to 5 years. At follow up, the intensive treatment group outperformed the parent training group on measures of intelligence, visual spatial skills, language and academics, though not adaptive functioning or behaviour problems. The intensively treated children also had less restrictive school placements. There was, however, variation in the gains made by the children with some children making very modest gains. Moreover, because the method of delivery of the behavioural therapy also varied (clinic vs parent directed), it is not clear whether the differences between groups was due to variation in the intensity (hours of EIBI the child received) or the means of delivery (clinic or parent-directed) of the programme. 

EIBI studies contrasting programme delivery

The means of delivery of the behavioural therapy may also be a factor influencing the effectiveness of treatment. Two good quality RCTs (level of evidence: II) (Sallows and Graupner 2005; Smith et al. 2000) compared clinic directed with parent directed EIBI.

The aim of Sallows and Graupner’s (2005) study was to determine whether a community based programme operating without the resources, support and supervision of a university centre could implement an EIBI programme and achieve similar results. After two to four years of treatment, cognitive, language, adaptive, social and academic measures were similar for both the clinic directed and parent directed EIBI groups. Children in both clinic and parent directed groups were treated intensively (average of 39 hours/week and 32 hours/week, respectively, in year 1), although the parent directed group received much less supervision. There was also variation in the gains made by children within both groups. The authors divided all participants into ‘rapid’ and ‘moderate’ learners post hoc in an attempt to assess patterns of improvement and look at overall predictors of achievement. Those in the rapid group (48%) showed dramatic increases in cognitive functioning (from a mean IQ of 55 to a mean IQ of 104) and social skills and were able to succeed in regular education classrooms. Low IQ (below 44) and absence of language (no words at 36 months) predicted limited progress whereas the rate of learning, imitation and social relatedness predicted favourable outcomes. 

As described in the section on programme intensity, Smith et al’s (2000) good quality RCT also compared clinic directed with parent directed behavioural treatment but the parent group had less intensive therapy. The benefits found in this study for the intensive clinic directed EIBI group when compared to the parent directed group may be related either to the less intensive hours in the latter group or the mode of delivery of the programme or both.

Picture Exchange Communication System

Three primary studies (one reported in two publications) assessed the effects of the Picture Exchange Communication System, a programme that teaches children how to interact with others by exchanging pictures, symbols, photographs or real objects for desired items (Carr and Felce 2007; Howlin et al. 2007; Yoder and Stone 2006a; Yoder and Stone 2006b). The goals of PECS are specific: to identify objects that may serve as stimuli for each child’s actions and to learn responses to simple questions with multi-picture systems. This behavioural programme uses ABA to teach functional communication.

One of the two RCTs (level of evidence: II) (Howlin et al. 2007) compared a group of elementary school children receiving immediate PECS or delayed PECS with elementary school children receiving standard care. In this good quality study, PECS was implemented during a five month period for the immediate treatment group until Time 2, the delayed treatment group then had five months of PECS treatment until Time 3 and assessments at Time 3 measured the immediate effects of PECS in the delayed group and the maintenance effects in the immediate treatment group. Rates of initiations and PECS usage increased significantly after treatment compared to the standard care group but were not maintained at Time 3 for the immediate treatment group. There were no group differences in frequency of speech, ADOS ratings (communication and reciprocal social interaction) and language scores. 

A non randomised experimental study (level of evidence: III-2) of fair quality (Carr and Felce 2007) also effectively compared PECS with no PECS treatment, using an instrument especially designed to record communication outcomes. Children aged between 3 and 7 years were allocated to either PECS plus eclectic treatment or eclectic treatment alone, based on whether they were within a 50 mile radius of researchers or not. Children receiving PECS had a significant benefit in frequency of child to adult initiations, frequency of linguistic communications, percentage of adult responses given to child initiated communications and percentage of child responses to adult initiated communications. There was no reported difference in the frequency of initiations giving an opportunity for child response and the frequency of adult to child initiations with no opportunity to respond was significantly higher for children receiving control.

The other PECS study was a very good quality RCT (level of evidence: II) (Yoder and Stone 2006a; Yoder and Stone 2006b). It compared two communication intervention approaches that address the intent to communicate prior to targeting spoken communication, PECS and ‘Responsive Education and Prelinguistic Milieu Teaching’ (RPMT). Outcomes were measured in preschool children after six months of treatment and again six months after treatment finished. There was strong growth on measures of spoken communication at 12 months follow up for both treatments and no significant differences were reported between groups. However, when assessments were made immediately after treatment ceased, PECS was more successful than RPMT in increasing the number of nonimitative spoken communication acts and the number of different nonimitative words used. Considering all measurement periods, growth rate of the number of nonimitative words was faster in the PECS group than in the RPMT group for children who began treatment with relatively high object exploration. However, growth rate was faster in the RPMT group than in the PECS group for children who began treatment with relatively low object exploration. For children who began treatment with at least some initiating joint attention, RPMT facilitated the frequency of generalised turn taking and initiating joint attention more than did the PECS. In contrast, the PECS facilitated generalised requests more than the RPMT in children with very little initiating joint attention prior to treatment.

Other interventions

One good quality RCT (level of evidence: II) compared the Autism Preschool Program (which uses a variety of behavioural and language development methods directed by parents and day care staff) with standard care at day care centres (Jocelyn et al. 1998). Language development was significantly greater for children in the APP group. However, blind assessment by a psychologist of autistic symptoms indicated that these symptoms did not improve over time and did not differ between groups.

An RCT (level of evidence: II) of fair quality (Moore and Calvert 2000) compared the effects of a behavioural programme alone with the same programme using sensory reinforcement using educational software in preschool children. The addition of the computer increased the number of unknown words learned, increased attention to task and increased children’s motivation to continue in the programme but the fair quality study was very small (7 in each group) and follow up very short (1 week).

A fair quality RCT (level of evidence: II) (Drew et al. 2002) compared a group directed by parents trained in a psycho-linguistic and social-pragmatic approach (using behavioural management techniques) with local services (where a mixture of approaches were used, including ABA). The children were required to be less than two years of age. There were no reported differences after treatment between groups in non verbal IQ, symptom severity or parent stress. A trend towards a greater benefit for the experimental group in some language abilities was reported but groups were not equivalent at baseline.

A quasi randomised controlled trial  (level of evidence: III-1) of fair quality (Kroeger et al. 2007) compared direct teaching using a video-modelling format to teach play and social skills with unstructured play without direct instruction. Both groups had a similar staff to student ratio, reinforcement and behavioural management for inappropriate behaviour. Groups met regularly three times a week for five weeks. Both groups improved in their prosocial behaviours. The direct teaching group made significantly greater gains in initiating behaviours, responding behaviours and interacting behaviours. 

Finally, a fair quality non randomised experimental study  (level of evidence: III-2) (Wetherby and Woods 2006) compared the effects of the Early Social Interaction Project (ESI) (a programme designed to apply the recommendations of the National Research Council, UK) with a no treatment group on social communication outcomes. Communicative means and play were similar between groups after ESI treatment. However, children receiving ESI had significantly greater gains in social signals, rate of communicating, communicative functions and understanding than children not receiving ESI.

Secondary studies

Eleven of the 21 secondary studies assessed the effects of behavioural interventions (including EIBI) on global autistic symptoms rather than specific behaviours, 3 assessed video modelling, 3 assessed social skills interventions and 4 assessed the effects of behavioural interventions on challenging behaviour. 

Reviews of interventions assessing global improvement

Of the 11 reviews assessing global improvement, 10 were level I systematic reviews (Bassett et al. 2000; Burrows 2004; Chorpita et al. 2002; Doughty 2004; Finch and Raffaele 2003; Ludwig and Harstall 2001; McGahan 2001; Parr 2007; Roberts and Prior 2006; Scottish Intercollegiate Guidelines Network 2007) and one was a level III­1 review of SCED studies (Smith 1999). All reported positive effects of behavioural interventions on overall functioning, mostly in preschool children. However, eight of these secondary studies emphasised that results from the studies were inconsistent and their evidence synthesis indicated that no single programme had been identified that was more effective than any other programme. Concerns noted included that the extent of the improvement varied across studies, and that there was variation in response within groups which could not be explained by demographic factors. Gaps in evidence reported included that there was insufficient evidence to determine the duration and intensity of treatment, long term effects were unknown, as was the effectiveness of the intervention for older children. 

It is important to note that many of these reviews predated the more recent primary studies described in the previous section. The most recent of these reviews assessing effects on global autistic symptoms (published in 2006 and 2007) (Parr 2007; Roberts and Prior 2006; Scottish Intercollegiate Guidelines Network 2007) gave qualified support for behavioural interventions using ABA techniques. The very good quality review published in BMJ Clinical Evidence 2007 (Parr 2007) stated that there was low quality evidence that ABA type programmes might improve language function and IQ when compared to eclectic programmes but there was insufficient evidence for the other programmes using behavioural techniques, such as the Autism Preschool Program and PECS (although primary studies on PECS have been published since this review was completed). An Australian review of good quality by Roberts and Prior (2006) of the most effective models of practice for children with ASD also acknowledged that early behavioural intervention produces positive outcomes for children with ASD but the variation in the children, families, therapists, contexts and methods makes it difficult to reach definitive conclusions. The recent very good quality SIGN guideline (2007) made an A grade2 recommendation that the Lovaas (1987) programme should not be presented as an intervention that will lead to normal functioning. It made a B grade3 recommendation that behavioural interventions should be considered to address the wide range of specific behaviours in children and young people with ASD, both to reduce symptom frequency and severity and to increase the development of adaptive skills. However, the guideline did not identify the components of a successful behavioural programme.

The other secondary studies were reviews of small SCED studies (level of evidence III-1) and either used specific types of behavioural intervention, such as video modelling or targeted specific behaviours. 

Reviews of video modelling interventions 

Three reviews (of good or very good quality) assessed the effects of video modelling and/or video self modelling, mostly on social and communication skills, functional behaviour and challenging behaviour (Ayres and Langone 2005; Bellini and Akullian 2007; Delano 2007). All three reviews found that video modelling or video self modelling had positive effects on these outcomes but noted that the critical features of the intervention(s) were not identified. Two of the reviews confirmed that there was a moderate effect of the video modelling intervention on maintenance and generalisation of skills and behaviour. One review noted that it was unclear how video modelling compared to other programmes targeting these behaviours.

Reviews of interventions assessing social skills

Three reviews (level of evidence III-1) assessed the effects of all types of ‘social skills’ interventions on collateral skills and social behaviours (Bellini et al. 2007; Matson et al. 2007; McConnell 2002). Two of these were qualitative reviews where the researchers relied on the conclusions drawn by the studies’ authors to determine treatment effectiveness. One other review of good quality (Bellini et al. 2007) attempted to synthesise the studies quantitatively. This review reported a low to questionable treatment effect for social skills interventions, a low to questionable generalisation effect and a moderate maintenance effect leading to the conclusion that social skills interventions were minimally effective. There was no difference in relative effectiveness by different types, features of the interventions, age group or individual versus group except for location. Interventions in classrooms were more effective than those delivered in ‘pull out’ locations. There was insufficient evidence in the reviews to distinguish between the types and features of these interventions.

Reviews of interventions assessing challenging behaviour, and use of functional communication training

Four reviews of SCED studies (level of evidence III-1) assessed the effects of behavioural interventions on challenging behaviour. One of these assessed the effects of functional communication training (FCT) on both challenging behaviour and communication (Mancil 2006) and the other three targeted challenging behaviour alone (Campbell 2003; Horner et al. 2002; Machalicek et al. 2007). Settings in the majority of the included studies were the clinic or classroom.

The first review (Mancil 2006), rated as good quality, reported that FCT (preceded by functional behaviour analysis) implemented in a clinic setting decreased challenging behaviour and increased communication, but there was insufficient information about long term follow up, maintenance and generalisation. It noted that the majority of research of FCT in this context is clinically based and focuses on one communication method. The other three reviews assessed the effects of behavioural interventions without specifying the specific nature of these interventions. The authors reported that the interventions reduced challenging behaviour from between 76% to 90%. However, one review of good quality (Horner et al. 2002) noted that there was little evidence for generalisation and maintenance of intervention effects.

One of the three reviews (level of evidence III-1) relating to interventions for challenging behaviour (Machalicek et al. 2007), which was of fair quality and was limited to school-based studies, found that functional behavioural assessment methods (of any type) were used in half of the studies included in the review, but most (73%) of the interventions reported equally positive findings. 

The other two reviews of good/very good quality (Campbell 2003; Horner et al. 2002) found that functional behaviour assessment (FBA) prior to the implementation of behavioural interventions (across settings) increased effectiveness. The very good quality review by Campbell et al (2003) which employed hierarchical multiple regression analyses reported that FBA generally contributed to intervention success, but also found that those employing experimental functional analysis (EFA) had the most impact compared with other functional assessment methods involving indirect or descriptive methods. EFA is the most methodologically rigorous type of FBA and involves undertaking a number of structured observations under several standardised conditions which are introduced and withdrawn systematically. Notably, no other participant, treatment, or experimental variables contributed to efficacy in the regression equation.


  1. That it is based on at least one high-quality meta-analyses, systematic reviews of RCTs, or RCTs with a very low risk of bias (directly applicable to the target population), or a body of evidence consisting principally of well conducted meta-analyses, systematic reviews of RCTs, or RCTs with a low risk of bias (directly applicable to the target population, and demonstrating overall consistency of results) ( 
  2. That is, based on either a body of evidence including high quality systematic reviews of case control or cohort or studies, high quality case control or cohort studies with a very low risk of confounding or bias and a high probability that the relationship is causal (directly applicable to the target population and which demonstrate overall consistency of results), or extrapolated evidence from well conducted or high quality meta-analyses, systematic reviews of RCTs, or RCTs with a low or very low risk of bias.