Behavioural and Developmental Interventions for Autism Spectrum Disorder: A Clinical Systematic Review

Background Much controversy exists regarding the clinical efficacy of behavioural and developmental interventions for improving the core symptoms of autism spectrum disorders (ASD). We conducted a systematic review to summarize the evidence on the effectiveness of behavioural and developmental interventions for ASD. Methods and Findings Comprehensive searches were conducted in 22 electronic databases through May 2007. Further information was obtained through hand searching journals, searching reference lists, databases of theses and dissertations, and contacting experts in the field. Experimental and observational analytic studies were included if they were written in English and reported the efficacy of any behavioural or developmental intervention for individuals with ASD. Two independent reviewers made the final study selection, extracted data, and reached consensus on study quality. Results were summarized descriptively and, where possible, meta-analyses of the study results were conducted. One-hundred-and-one studies at predominantly high risk of bias that reported inconsistent results across various interventions were included in the review. Meta-analyses of three controlled clinical trials showed that Lovaas treatment was superior to special education on measures of adaptive behaviour, communication and interaction, comprehensive language, daily living skills, expressive language, overall intellectual functioning and socialization. High-intensity Lovaas was superior to low-intensity Lovaas on measures of intellectual functioning in two retrospective cohort studies. Pooling the results of two randomized controlled trials favoured developmental approaches based on initiative interaction compared to contingency interaction in the amount of time spent in stereotyped behaviours and distal social behaviour, but the effect sizes were not clinically significant. No statistically significant differences were found for: Lovaas versus special education for non-verbal intellectual functioning; Lovaas versus Developmental Individual-difference relationship-based intervention for communication skills; computer assisted instruction versus no treatment for facial expression recognition; and TEACCH versus standard care for imitation skills and eye-hand integration. Conclusions While this review suggests that Lovaas may improve some core symptoms of ASD compared to special education, these findings are based on pooling of a few, methodologically weak studies with few participants and relatively short-term follow-up. As no definitive behavioural or developmental intervention improves all symptoms for all individuals with ASD, it is recommended that clinical management be guided by individual needs and availability of resources.


Introduction
Autism spectrum disorders (ASD) are neurodevelopmental disorders characterized by a triad of deficits involving communication, reciprocal social interaction, and restricted and repetitive patterns of behaviour, interests and activities [1]. In addition to these core features, a range of other behaviour problems are common, such as anxiety, depression, sleeping and eating disturbances, attention issues, temper tantrums, and aggression or self-injury [2]. Autism is classified within a clinical spectrum of disorders known as pervasive developmental disorders, as defined in the Diagnostic and Statistical Manual of Mental Disorders and the International Statistical Classification of Diseases and Related Health Problems. The spectrum includes conditions such as Autistic Disorder, Asperger's syndrome, Atypical Autism, and Pervasive Developmental Disorder Not Otherwise Specified [3].
In clinical practice, professionals may use different terms interchangeably to refer to children with similar presentations. While there are no definitive medical tests to indicate the presence of any form of ASD, diagnosis can be made by three years of age based on the presence or absence of specific behaviours that are used as diagnostic criteria. Prevalence estimates indicate that between 10 and 15 of every 10,000 children are autistic [1,4] but possibly greater than 20 of every 10,000 children have dysfunction which warrants diagnosis at any point along the spectrum [5,6]. Common comorbidities include mental retardation (Intelligence quotient (IQ) ,70) and epilepsy, which are associated with 70% and 25% of autism cases, respectively [7,8]. While no known cure for ASD exists, the general agreement is that early diagnosis followed by appropriate treatment can improve outcomes in later years for most individuals [9]. Consequently, the question of how various interventions may help to increase the individual's ability to function is highly relevant to families, health professionals, and policy makers.
Over the past 20 years, a variety of therapies have been proposed to improve the symptoms associated with ASD. Current treatments include pharmacological therapies and various complementary therapies including diet modifications, vitamin therapy, occupational therapy, speech and language therapy and behavioural and developmental approaches [10]. Interventions that fall within the continuum of behavioural and developmental interventions have become the predominant treatment approach for promoting social, adaptive and behavioural function in children with ASD based on efficacy demonstrated in empirical studies. These interventions may be viewed in terms of their position on a continuum from highly structured discrete trial training behavioural approaches guided by a therapist, to social pragmatic approaches where teaching follows the child's interests and is embedded in daily activities in a natural environment. While therapy may be provided for up to 40 hours per week, controversy exists regarding the intensity required to achieve positive outcomes and the efficacy of one approach compared to another. An umbrella review of systematic reviews of behavioural and developmental interventions for ASD [11] has found that most systematic reviews have methodological weaknesses which make them vulnerable to bias and compromise their validity. There is evidence of positive outcomes for many of the interventions examined in systematic reviews of ASD and therefore, there is a need for further systematic reviews on the effectiveness of behavioural and developmental interventions for ASD which adhere to strict scientific methods.
Clinicians, educators and families of individuals with ASD need to make informed decisions regarding treatment options and therefore, a host of clinical and research questions regarding the benefits of these interventions still need to be clarified and addressed. Considering the importance of, and demand for, behavioural interventions for ASD, as well as the current rising trend in new programs, a rigorous synthesis of high quality evidence regarding the effect of a continuum of behavioural and developmental interventions for ASD will provide much needed information for health care professionals, policy makers, researchers, and families. This systematic review was conducted in order to identify, appraise, and synthesize the evidence on the effects of a continuum of behavioural and developmental interventions for improving core symptoms associated with ASD.

Search Strategy
The systematic review followed a prospective protocol that was developed a priori. Peer-reviewed comprehensive searches were conducted up to May 2007 in 22 psychological, educational and biomedical electronic databases for commercially published literature, as well as dissertations, and conference abstracts (e.g., MEDLINEH, EMBASE, ERIC, CINAHLH, Cochrane Central Register of Controlled Trials, ProQuest Dissertations and Theses, PsycINFOH, BIOSIS PreviewsH, and Web of ScienceH). We identified additional studies by contacting experts in the field and by searching reference lists of primary studies, review articles, and textbook chapters. Details of the complete search strategies are available in Supplement S1.

Study Selection
Studies were included if they were: randomized controlled trials (RCTs), controlled clinical trials (CCTs) or observational analytical studies (i.e., prospective or retrospective cohort studies with comparison groups); published in English; and reported data on the effects of a behavioural or developmental intervention in individuals with ASD. Individuals with Rett's disorder or Childhood Disintegrative Disorder were not considered for this review as they do not conventionally fall within ASD due to their significantly different clinical course. Studies involving participants with dual diagnoses (i.e., any ASD plus attention deficit/ hyperactivity disorder, obsessive compulsive disorder, or learning problems) were also considered for inclusion. The primary outcome of interest was the change in core features of ASD (i.e., communication, reciprocal social interaction, and restricted and repetitive patterns of behaviour, interests and activities) as indicated in the Diagnostic and Statistical Manual of Mental Disorders criteria [1]. Other outcomes that were examined included changes in non-core behaviours, developmental changes, cognitive changes, adaptive behaviours, challenging behaviours, play skills, educational performance, and family-related outcomes. One reviewer screened titles and abstracts of potentially relevant studies. Inclusion criteria were applied independently by at least two reviewers. The primary reason for exclusion of articles was documented. A complete list of excluded studies and reasons for exclusion are available in Supplement S2.

Quality Assessment and Data Abstraction
Two reviewers independently assessed the methodological quality of the studies. Disagreements were resolved by consensus. We assessed the methodological quality of the studies with two pre-tested checklists (one for clinical trials and the other for observational studies) that included items from other published scales and checklists [12][13][14][15][16][17][18]; these items address specific aspects of design, execution, and analysis of the studies. The trials checklist included questions related to bias reduction such as allocation concealment [19,20], randomization, blinding (subject, provider, and outcome assessor blinding), and description of dropouts and withdrawals [21,22]. Other variables that were evaluated included description of selection criteria, therapeutic regimens, intervention providers, and treatment fidelity. The checklist for the observational studies included items that evaluated the methods of selection of exposed and non-exposed cohorts, ascertainment of outcome and exposure, and how the study handled confounders in the design or analysis. Finally, information regarding the source of funding was collected [23]. Information regarding the study design and methods, the characteristics of participants, interventions, comparison groups, and outcomes of interest were extracted using a pre-tested data extraction form. One reviewer extracted the data using a pre-tested form, and a second reviewer verified the accuracy and completeness of the data. Discrepancies in data extraction were resolved by consensus between the data extractor and the data verifier. Interventions were categorized based on a classification scheme previously described by other researchers in this field [24].

Analysis and Presentation of Results
There is considerable overlap between and across various models to classify and describe interventions that fall within the continuum of behavioural and developmental interventions for

Description of Studies
The studies evaluated the effect of eight broad types of interventions for ASD: Applied Behaviour Analysis (ABA) interventions, communication-focused interventions, contemporary ABA, developmental approaches, environmental modification programs, integrative programs, sensory motor interventions, and social skills development interventions (Figure 2). The studies were published between 1977 and 2007, with 2002 as the median year of publication. Data from a total of 2566 participants (median sample size = 22 per study; interquartile range: 15 to 36; n = 99) were reported in the studies. The median chronological age of participants in the studies was 62 months (interquartile range: 42 to 105 months; n = 84). Seventy-six percent of the studies included populations of infants or toddlers (less than 6 years of age), 44 per cent included school age children (6 to 12 years of age), 25 percent included adolescents (13 to 18 years of age), and only 11 percent included adults (older than 18 years of age). Studies included participants with conditions described as autistic disorder (93 percent), progressive developmental disorder (23 percent), Asperger's syndrome (14 percent), high-functioning autism (5 percent), atypical autism (2 percent), not yet diagnosed autism (1 percent), and other (3 percent) such as autistic savant, or autistic-like conditions. The majority of the studies (67 percent) did not report on the level of severity of autistic symptoms in the study population. Participants with severe symptoms of ASD were included in 20 percent of the studies, whereas 19 percent included participants with moderate symptoms. Those with mild symptoms were not frequently included in the studies (15 percent). Summaries of the study characteristics and details of individual findings are presented in Table 1.

Quality of Studies
Details on the methodological quality of the studies are presented in Table 2 and Table 3. Briefly, the majority of trials (83 percent) failed to mention how representative the sample was in terms of the study setting, the selection criteria for enrolling participants, and the operational definition of ASD. A minority of studies (32 percent) reported on monitoring the fidelity of intervention implementation. Although more than half of the trials (64 percent) reported the use of randomization, few trials (seven trials) reported the procedure for separating the process of randomization from the recruitment of participants. The majority of trials (89 percent) failed to clearly report how they concealed the sequence of allocation to the interventions under study. Less than half of the studies (43 percent) reported that blind or independent outcome assessment was conducted. In terms of attrition bias, 33 percent of the trials provided a description of withdrawals and dropouts from the study. Finally, just over half of the trials (54 percent) reported their sources of funding. Thirty-two percent were funded by government agencies, 22 percent received funding from foundations or societies, 19 percent used internal funds, and five percent were funded by private industry.
Overall, the methodological quality of the 14 cohort studies was modest. In general, the cohort studies failed to protect against selection bias: only three studies clearly mentioned how representative the overall sample was in terms of the study setting, the description of the selection criteria, and the operational definition of ASD used for the study. The control for detection bias affecting the ascertainment of both exposure and outcome was moderate in the cohort studies. None of the studies used secure methods for ascertainment of exposure. The majority of the studies provided evidence on the reliability of methods for outcome assessment; however, only half of the studies explicitly stated that outcome assessment was blind to exposure status. Finally, only four observational studies disclosed their source of funding. The methodological strengths and weaknesses of individual studies, presented in Table 2 and Table 3, should be taken into consideration when interpreting the study results and conclusions.

Summary of Findings
Applied Behaviour Analysis. Evidence from 31 studies (12 trials and 9 cohort studies) involving a total of 770 participants was analyzed on the use of discrete trial training and Lovaas therapy for ASD. The effects of discrete trial learning are inconsistent across studies. All the studies that compared discrete trial training to no treatment reported statistically significant findings [95,118,129]. Motor and functional outcomes more often demonstrated positive results compared to speech-related outcomes which were generally negative. All cohort studies demonstrated significant results [118,124,129]. Lovaas therapy was consistently found superior to standard care [99,122] or regular instruction [94,127] in terms of intellectual functioning, language comprehension, and communication skills. Generally, high-intensity Lovaas was found to be superior to low-intensity Lovaas in terms of intellectual functioning, communication skills, adaptive behaviour and overall pathology [71,99,125,128]. The results for Lovaas therapy compared to special education showed variable results at the individual study level and seemed to indicate more effect for the medium-term (12 and 14 months, respectively) [90,94] which was not apparent within the longer-term studies (3 and 9 years, respectively) [88,120]. No significant differences were found within studies comparing Lovaas to Developmental Individual-difference relationship-based intervention (DIR) [53] or Integrative/Discrete trial combined with Treatment and Education of Autistic and related Communication Handicapped Children (TEACCH) [123]. Seven of the eight studies that reported significant findings for Lovaas therapy were non-RCTs [90,94,99,122,125,127,128]. Three of the four RCTs in this category reported no significant findings [53a,53b,66]. This observation has serious implications for the interpretation of evidence from non-RCTs. There is some evidence that results of RCTs and non-RCTs sometimes, but not always, differ, [130] and that non-RCT can be more prone to bias and overestimate treatment effects [131,132].

Communication-focused
Interventions. Ten trials involving 269 participants were identified that evaluated the effects of communication-focused interventions. Positive effects and statistically significant results were produced at the study level for emotional recognition [49,69], close generalization tasks [49], verbal IQ [49], attention [60] and motivation [60]; these studies were all RCTs and had varied control groups including no treatment, as well as active interventions. There is evidence from three trials (2 RCT, 1 CCT) that sign language training provides benefits in terms of communication-related outcomes, such as  DT N = 2 1) There were no significant differences between the use of picture activity schedules and prompting at home or at school in the rate of on-task behaviours.
2) There were no significant differences in the number of on-schedule behaviours between the groups. NT N = 2 1) DT instruction produced less initiation or expansion of social contact with peers, but they responded more to interactions (non-verbal and verbal) with peers than children without DT instruction. 2) The behavioural group showed more gains than the eclectic group on IQ, language and adaptive behaviour. 2) The IBT group had statistically higher mean standard scores in all skill domains than the control groups, except for motor skills. RI N = 9 1) Participants in the home-based behavioural treatment obtained statistically significant higher IQ score at followup than the control. 2) There was a statistically significant reduction of symptoms severity after participating in the home-based behavioural treatment.

UCLA/Lovaas
Smith T, 1997 [128] USA Lovaas N = 10 1) Participants in the intensive behavioural treatment achieved higher IQ and had more expressive speech than the control.
2) Behavioural problems were reduced in both groups.
3) Intensively treated children achieved clinically meaningful gains relative to the comparison group but remained quite delayed. Social skills program N = 13 1) Interactive multimedia significantly improved close generalization tasks.

COMMUNICATION-FOCUSED INTERVENTIONS
2) The interactive multimedia intervention produced significant effects on verbal IQ. Teaching N = 21 1) The joint attention group showed a significant increase in initiation and responsiveness to joint attention and improvements in mother-child interactions.
2) Children in the play group showed more diverse types of symbolic play and higher play levels in interaction with their mothers.
3) There were differences between joint attention and play groups on initiating shows and coordinating joint looks.

Control (ND) N = 17
Wang 2) Parents in the training group scored significantly higher on responsiveness during free play interactions.
3) There were differences between the groups in parental stress levels.

Incidental Teaching
Bloch 2) The ITB training condition did not appear to have any significant impact on the parents' interactional style.
Openden NT N = 4 1) Implementation of the DIR program did not produce significant changes in behavioural repetitive stereotypies when compared to a non-DIR program group. 2) There were significant changes in positive social interactions or negative social interaction skills between the groups. 2) Compared to contingently responsive group, imitation group showed significantly less time being inactive/playing alone and more time showing object behaviours. DCI N = 10 1) An imitation interaction strategy produced a significant increase of both proximal and distal social behaviours compared to the contingency group.

Imitative Interaction
2) The imitation intervention significantly increased children's imitation skills at a more generalized level.

Incidental Teaching
Eagle 2) There were differences between the groups on communicative means and play behaviour. WL N = 25 1) The More than Words program produced a significant advantage in parents' observed use of facilitative strategies and in childrens' vocabulary size.

More Than Words
2) There were significant differences between the groups on children's social score or behaviour, parental stress or adaptation.

Responsive Training
Aldred 2) The intervention elicited positive changes in cognitive performance. SC N = 12 1) There was some evidence that the parent training group made more progress in language development.
2) The language ability of both groups remained severely compromised at followup.
3) Groups did t differ in non-verbal IQ, symptom severity or parental stress at followup. Social skills program N = 12 1) There were significant differences between treatment and control groups in parent-rated and staff-rated social skills, adaptability and atypical behaviour. NT N = 4 1) A social skills training program produced non-significant but considerable improvements on false belief tasks.
2) The intervention did not produce changes in parent and teacher ratings of social competence. SC N = 11 1) A home program intervention improved significantly more in imitation, fine motor, gross motor, non-verbal conceptual skills and overall PEP-R scores.
2) The home program intervention was effective in enhancing development in young children with autism. Unmodified music N = 9 1) The control condition was superior than AIT on parent-rated hyperactivity and direct observation measures of ear-occlusion.
2) There were differences in intelligence measures and language comprehension between groups at followup. Rimland  Unmodified music N = 9 1) AIT produced statistically significant improvements in adaptive behaviour.
2) AIT did not decrease sound sensitivity in children with autism. Smith  Social stories N = 25 1) Creative dance produced significantly greater social gains than in circle time condition for both verbal and non-verbal children.
2) The performance gap between verbal and non-verbal autistic children was smaller in the creative dance group than circle time group. Mason  Placebo N = 7 1) There was a significant reduction in tension and a marginally significant reduction in anxiety for children who received deep pressure compared to children who did not.
2) Deep pressure may have a calming effect for persons with autism, particularly those with high levels of arousal or anxiety. Social Stories N = 15 1) Social stories using either a standard or a directive approach produced significantly higher game play skills than the control group.
2) Standard and directive groups did t significantly differ on rate of play skill improvement.

Control (ND)
Ricciardelli    Table 3. cont. articulation competence, oral language, nonverbal communication, and child-initiated speech [67,82,104]. There is also some suggestion that sign language training may be most effective when combined with other modalities [82]. One CCT of Picture Exchange Communication System versus regular instruction showed a significant increase in communication initiations and dyadic interactions [133].
Contemporary ABA. Evidence on the effects of contemporary ABA was identified from 12 studies (11 trials and one cohort study) with a total of 573 participants. At the individual study level, significant improvements in the child's behaviour management, social skills, and parent's mental health have been reported following cognitive behaviour therapy [72,73,77]. There is limited and inconclusive evidence for various combinations of discrete trial training, incidental teaching, pivotal response training, and milieu teaching [55,57,78,121], and some evidence that pivotal response training may be beneficial for communication and social interaction [58,63,113].
Developmental Interventions. Twelve studies (11 trials and one cohort study) with a total of 256 participants were identified that evaluated the effects of interventions involving the application of developmental principles. Distinct modalities were evaluated within this category (e.g., imitative interaction, milieu therapy, responsive training, DIR, More than Words, and the Scottish Centre for Autism Preschool Treatment program). The available evidence appears to indicate no short-term differences between DIR and a no treatment group in aggression, self-stimulating behaviour, and social skills [134]. No evidence of effect was reported for the comparisons between two incidental teaching-based approaches in social interaction [89], two milieu-based approaches in communication and play behaviour [119], and responsive training versus no treatment in parents' attitude toward ASD, children's aggressive problems, externalizing problems, and depressive or anxiety symptoms [84]. Positive results were reported for the comparisons between milieu therapy and no treatment in cognitive abilities and course of ASD [101]; and for milieu therapy versus Picture Exchange Communication System in communication and play behaviour [81]. The comparison of More than Words to a wait-list control showed positive results in facilitative strategies and vocabulary size, however, no significant differences were observed for social skills, behaviour, or parental stress or adaptation [103]. Response training was superior to standard care in quality of reciprocal social communication and expressive language [32]. Finally, one small trial evaluated the Scottish Centre program versus a wait-list control and demonstrated positive results in joint attention, social interaction, imitation, daily living skills, motor skills, and adaptive behaviour [111].
Environmental Modification. One cohort study involving 44 participants examined work placement versus waitlist [116]. This prospective cohort study reported positive results in terms of significantly greater improvement in nonvocational outcomes and cognitive performance.
Integrative Programs. Evidence on the effects of a variety of integrative programs was obtained from 14 studies (13 trials and one cohort study) containing a total of 382 participants. Interventions in this category included Lego therapy, social skills program, and TEACCH. Lego therapy was evaluated in one large retrospective cohort study, and produced significant improvements in terms of social skills and autistic symptoms [126]. Six studies evaluating social skills programs produced inconsistent findings; there were no identifiable patterns in the outcomes examined [56,59,75,98,106,108]. Individual studies that evaluated TEACCH consistently reported significant findings for a variety of outcomes, including but not limited to fine motor and gross motor skills, cognitive performance, social adaptive functioning, and communication [105,107,114,115].
Sensory Motor Interventions. Fifteen studies (14 trials and one cohort study) were identified that evaluated the effects of interventions that involved the application of sensory motor principles, involving a total of 156 participants. Several modalities were evaluated within this category: auditory integration training, exercise, restricted environmental stimulation therapy, and sensory integration. Six studies of sensory integration versus no treatment groups reported statistically significant results for stereotypic behaviours, off-task behaviours, and touch aversion [43,44,47,54,100,117], but it is not known how sensory integration compares to other active interventions. The results for communication-related outcomes are contradictory, and no effect was reported for intellectual functioning. Two relatively large studies on creative dance [92] and horse riding [102], respectively, demonstrated significant social gains. Studies on the effects of restricted environmental stimulation therapy provided inconclusive evidence [51,76]. No studies evaluated effects over the long-term; therefore the sustainability of these changes is unknown.
Social Skills Development Intervention. Six trials, containing a total of 135 participants, provided evidence on the effects of social skills development interventions, all of which evaluated the effects of Social Stories TM in ASD. Five of the six studies showed statistically significant results for a variety of outcomes related to social interaction at short-term (e.g., 1 day to 6 weeks) [34,35,46,64,110]. There were no studies comparing Social Stories TM to other active treatments.

Meta-analyses
A limited number of meta-analyses were feasible due to variations among the studies in the type of interventions assessed, the comparison groups, and the outcomes of interest. Of the 101 studies included in the review, 13 studies (six RCTs, five CCTs and two observational studies) contributed data to the metaanalysis. Table 4 summarizes the comparisons and outcomes that were suitable for meta-analysis. In a meta-analysis of three CCTs [38,124,127] involving 112 participants, statistically significant results were obtained for Lovaas treatment compared to special education on measures of adaptive behaviour (WMD = 11.8; 95% CI, 6.94 to 16.67), communication and interaction (WMD = 16.63; 95% CI, 11.25 to 22.01), comprehensive language (WMD = 12.84; 95% CI, 6.38 to 19.30), daily living skills (WMD = 5.61; 95% CI, 0.54 to 10.67), expressive language (WMD = 15.05; 95% CI, 6.19 to 23.90), overall intellectual functioning (SMD = 0.95; 95% CI, 0.44 to 1.46), and socialization (WMD = 9.17; 95% CI, 2.16 to 16.19). High-intensity Lovaas was shown to be superior to Low-intensity Lovaas on measures of intellectual functioning in two retrospective cohort studies with a total of 173 participants (SMD = 0.92; 95% CI, 0.61 to 1.24) [40,111]. Pooling of two RCTs [18,135] including 40 participants yielded statistically significant results for developmental approaches based on initiative interaction compared to contingency interaction in the amount of time spent in stereotyped behaviours (WMD = 20.40; 95% CI, 20.73 to 20.07), and the amount of time spent in distal social behaviour (WMD = 2.85; 95% CI, 0.99 to 4.71), but the effect sizes were not clinically significant. Statistically non-significant results were obtained for the comparisons between Lovaas and special education in measures of nonverbal intellectual functioning (three CCTs [38,124,127] recognition (two RCTs [136,137], N = 48; SMD = 0.53; 95% CI, 20.05 to 1.12); and TEACCH versus standard care on measures of imitation skills (two CCTs [57,121], N = 56; SMD of 0.46; 95% CI, 20.07 to 0.99), and eye-hand integration (two CCT [57,121], N = 56s; SMD = 20.24; 95% CI, 20.77 to 0.28).
Because of the very small number of trials available for each comparison, the effect of publication bias on the meta-analyses presented above was not analyzed.

Discussion
Our systematic review of the indexed scientific literature on the effects of behavioural and developmental interventions for ASD has demonstrated a lack of agreement across the studies on the effect that these interventions may have on clinically relevant outcomes. Despite evidence, there is no clear answer regarding the most effective therapy to improve symptoms associated with ASD.
The interpretation and generalization of results summarized from individual studies is complicated by a number of factors. First, ASD is a complex diagnosis that represents a spectrum of symptoms. The varied interventions may target different symptoms or the same symptoms to different extents. As a result, practitioners and decision-makers may need to target their choice of treatment to the uniqueness of each presenting child and the symptoms that are most important for the well-being of each child and their family. In interpreting the literature, the reader needs to consider the findings in light of the study population and the outcomes that were evaluated. There is considerable potential for heterogeneity in the population, intervention, comparator and outcomes of interest, as ASD is a spectrum disorder, therapy is not always reported in detail, comparators are difficult to control for, and outcomes are somewhat subjective. It should be noted that controversy exists regarding the use of intellectual functioning as an outcome, since higher IQ scores may represent true increases or merely a better ability to take the test following the intervention [138]. Second, the interventions themselves are complex and multifaceted. Many of them have components that may be implemented in different ways, different settings, and by different people including both professionals and lay people. In some cases, this prohibits generalizations regarding a specific intervention. Third, consideration needs to be given to the comparison groups. As a general finding, it appears that any intervention is better than nothing. That is, most of individual studies showed benefits when an active intervention was compared to no treatment or wait-list controls. It has been reported that behavioural researchers sometimes include a comparison group that monitors symptoms for a period of time equivalent to the time required for the intervention in the active group before beginning the treatment [139]. Although this approach controls for certain non-specific treatment effects such as regression to the mean, the situation may create a negative expectation of improvement (i.e., no one expects to improve while they wait for treatment) and therefore, artificially inflate the difference between the active and the control (wait-list or no treatment) groups. Therefore, results of studies that compared active versus no treatment or wait-list controls should be interpreted with caution. Fourth, variation in results may occur due to different length of follow-up across studies. The length of follow-up must be appropriate to the nature of the intervention and its mechanism of action (i.e., how long an intervention is required to begin to have an effect) and the outcomes being measured (i.e., length of time to elicit change in a specific outcome). Further, consideration needs to be made for whether any observed changes are maintained over the longer-term. Finally, the results need to be considered in light of the methodological quality of the studies and their potential for bias (i.e., under or overestimation of treatment effects). A particular concern is related to the potential of outcome reporting bias in these studies, in which statistically significant results have a higher chance of being fully reported compared to non-significant results [140]. Although the presence of selective outcome reporting was not formally evaluated in our review, future evaluations of the evidence on the effectiveness of behavioural and developmental interventions should compare trial publications to protocols to verify whether changes or omissions in selected outcomes were introduced from registration to publication of the trial.
A few studies of modest methodological quality were available for meta-analysis, mostly reporting non-significant results. Few statistically significant results favoured Lovaas therapy and developmental approaches based on initiative interaction. The positive results from these meta-analyses need to be interpreted with caution, since biases, such as expectancy bias, cannot be excluded. It is unknown whether the non-significant results obtained for computer assisted instruction, and TEACCH are truly ''negative findings'' (i.e., evidence of no effect) or if there is a lack of power to detect a statistically significant result due to the low number of studies included in the meta-analyses (i.e., no evidence of effect). Finally, we found that 54 percent of the studies disclosed the source of funding, with most of the research being sponsored by government agencies or scientific societies. Only five percent of the studies declared private industry funding. There is evidence that industry funding of biomedical research may bias conclusions toward positive results for their products (sponsorship bias) [141]; however, there is no evidence on whether sponsorship bias extends beyond industry to other sources of funding.

Applied Behaviour Analysis
Evidence was analyzed on the use of discrete trial training and Lovaas therapy for ASD. The evidence seems to provide some support for discrete trial training in terms of motor and functional skills but not for communication skills. Lovaas' therapy showed benefits when compared to ''no treatment'' and evidence from meta-analysis of retrospective cohort studies showed greater effects for High versus Low intensity Lovaas. Results from a meta-analysis of CCTs demonstrated that Lovaas is superior to special education for a variety of outcomes, however, there is no definitive evidence suggesting superiority of Lovaas over other active interventions.
A previous review [142] concluded that overall, studies of behaviour analytic early intervention programs report substantial improvements, but the nature of improvements vary considerably across studies. The authors also recognize the methodological flaws in the earlier studies that preclude drawing definite conclusions related to programming. Other reviews [143][144][145][146][147][148] have also reported on the effects of early and intensive behavioural interventions for ASD. They agree that the majority of recent primary studies of reasonable quality document some improvement associated with behavioural intervention, but it remains to be determined if any one early and/or intensive intervention program is more effective than another. Furthermore, there was insufficient evidence to establish a relationship between the amount (per day and total duration) of any form of treatment program to obtain desirable outcomes. Replication in RCTs is needed to substantiate the use of Lovaas intervention and to determine the effect of treatment intensity on the outcomes of children with ASD.

Communication-focused Interventions
Individual studies reported positive effects in motivation, IQ changes, and emotional recognition associated with communication-focused interventions; however the meta-analysis results were not statistically significant for the comparison between computer assisted instruction and no treatment on measures of facial expression recognition. Two previous reviews [149,150] have examined the evidence from a variety of study designs of interventions that enhance communicative competence of individuals with ASD, such as assistive technology, augmentative and alternative communication methods. In keeping with these reviews, we conclude that future research is needed to better delineate the extent to which these interventions actually enhance outcomes in individuals with ASD.

Contemporary ABA
The evidence supporting the use of contemporary ABA approaches is variable and there is no evidence to suggest that one approach is more effective than another. A previous metaanalysis [151] that included a variety of study designs other than RCTs, CCTs and observational cohort studies indicated that the contemporary ABA approach produces greater gains in cognitive skills than Lovaas or developmental approaches, but both contemporary ABA and Lovaas methods were similarly effective in fostering language and adaptive skills in this population. This remains a question for future research to confirm as the methodological difficulties with the primary studies have made it difficult to be conclusive.

Developmental Interventions
Overall conclusions for developmental interventions are elusive due to the varied nature of the modalities, discrepant results across modalities, and limited evidence for each. Another review [151] has reported that compared to Lovaas and Contemporary ABA approaches, developmental interventions were found to be ineffective in the cognitive development of the participants but were effective in language development.

Environmental Modification
Only one study assessed the effectiveness of interventions labelled under this category. General conclusions for this type of intervention are prohibited due to limited evidence.

Integrative Programs
The evidence to support the use of these interventions is limited or inconsistent across studies. Another review [142] found similar results and emphasized the need to evaluate which components of these multi-faceted interventions are responsible for changes in clinically relevant outcomes.

Sensory Motor Interventions
The evidence is either limited or inconsistent for this group of interventions to support their use in clinical practice. Interpretation of evidence for individual studies within this category needs to be considered in light of the comparison group, the length of follow-up, and the outcomes examined. One systematic review [152] has evaluated the efficacy of sensory and motor interventions for children with ASD. This review included a variety of study designs such as descriptive case studies, single-subject designs, and RCTs. The review concluded that many of the sensory and motor intervention approaches have shown mixed effects at short term for children with ASD through uncontrolled, descriptive studies. Previous reviews that have evaluated the evidence for various types of sensory and motor interventions have similarly reported mixed clinical effects and have concluded that there is insufficient evidence to support use of these interventions at present [11].

Social Skills Development Intervention
The limited evidence supports Social Stories TM for short-term improvement of social symptoms associated with ASD among school-aged children. Past reviews have examined the effect of Social Stories TM [153,154] in children and young adults with ASD [155]. The reviews conclude that the effects of Social Stories TM are highly variable, and empirical foundation regarding its effectiveness is limited. However, the reviews agreed that published research in Social Stories TM has demonstrated positive effects, and therefore provides preliminary support to consider it a promising intervention.

Review Strengths and Limitations
A range of therapeutic approaches currently exists to help alleviate the symptoms of ASD. Due to the lack of a unique classification system to describe the variety of treatments for ASD, an intervention taxonomy system based on previous studies and experts' opinion was developed for the purposes of the review. The categories considered seem to be sensible, but it may not be possible to find a framework that would mandate exactly this set. A potential limitation of this approach is that the therapies examined were pragmatically classified and other therapies such as music therapy, drama therapy, and animal therapy could have been included if other classification approaches had been used. Synthesis of the evidence for other therapeutic approaches not evaluated in our review are available in the scientific literature [156].
Our search strategy is likely to have identified the majority of the available literature on the efficacy and effectiveness of behavioural and developmental interventions for ASD. We particularly targeted the indexed literature, yet we also searched for theses and dissertations, which altogether represent almost onethird of the studies included in this review. However, we acknowledge the possibility that the review may not be fully comprehensive, as we did not include additional grey literature sources in our search strategy. It has been reported previously that on average, published trials show a 9 percent greater treatment effect than grey trials [157,158] and therefore, there is the potential that our meta-analyses report an overestimate of the treatment effect. Further, the search results were initially screened by only one reviewer due to resource limitation. To date, there is no empirical evidence that indicates what the impact of screening by two, as opposed to one, reviewer has on selection bias; however, use of two reviewers during the screening process may have provided additional reassurance of the selection process.
We adopted a comprehensive strategy to appraise the methodological quality of the included studies. Our approach to quality focused mainly on an assessment of the internal validity of the studies, as recommended by several researchers [22,159]; however, some aspects related to the external validity and adherence to the interventions under study were also considered. One of the limitations of this review is the restriction of included studies to English-language publications. We did not include foreign language literature because of the difficulties in translation. Particularly, there is a wealth of Japanese literature available that could prove very interesting. We do not know the magnitude of bias that the exclusion of foreign literature may have produced in the results of our meta-analysis [160,161]. An additional limitation of our review is that studies were included regardless of whether there is evidence to support the psychometric properties of their outcome measures. Therefore, our analysis includes both instruments which are well validated for measuring clinical change in ASD (e.g., Wechsler Intelligence Scale for Children-Revised), as well as those that are commonly used, yet whose psychometric properties have not been studied in the ASD populations (e.g., Reynell Developmental Language Scales) [90].

Clinical Relevance
The research reviewed in this examination of behavioural and developmental treatments for ASD reveals that there are a number of treatment programs, some comprehensive and others with a specific behavioural focus, that have been developed to treat the core symptoms of ASD. Across any one specific intervention approach, the research is lacking scientific rigour, replications are sparse, and outcomes are variable; however, there are some implications for practice. First, it appears that most children with ASD make at least some progress on desired outcomes during their participation in intervention programs. Yet, the progress that individuals make across these programs or treatment approaches varies; some show remarkable progress while others show slow or minimal gains. Further the sustainability of changes over time is unknown. The source of this variation is uncertain given the quality of the research to date. It is unclear how participant characteristics interact with specific treatment programs or specific components of programs. Thus, practitioners need to be mindful to communicate these uncertainties to families seeking intervention services. Second, when selecting a program, practitioners need to select programs that have at least some evidence of support, select programs that are manualized and ensure that interventionists are able to maintain the level and quality of implementation of the program. Manualized programs serve to provide standardization of an intervention, yet uniformity must be balanced with the need to individualize the intervention [14]. The interventions themselves are complex and multifaceted. The variation in the expression of the symptoms of ASD make individualization necessary yet this presents problems for clearly specifying and evaluating the essential components of any given intervention. McMahon has recommended manuals with ''constrained flexibility'', where limited variation in the implementation of an intervention is permitted [162]. It is important to highlight that, although the evidence regarding the efficacy and effectiveness of behavioural and developmental interventions for ASD is currently limited, it does not mean that there is evidence of no effect from the interventions. The findings of this review are consistent with current clinical practice guidelines, which list various treatment options and approaches, yet offer limited guidance regarding choice of intervention [25,163,164]. However, guidelines do provide some parameters regarding which components constitute effective treatment programs, including: daily opportunities to use and increase spontaneous communication, engaging in meaningful learning activities that are functional in multiple settings, ongoing monitoring progress and adjustment of teaching practices to maximize progress, frequent interaction with typical peers and involvement of family members [164]. Until more reliable evidence is available, practitioners and decisionmakers may need to target their choice of treatment to the uniqueness of each presenting child and the symptoms that are most important for the well-being of each child and their family.

Future Research
Based on our review of the literature, there are several recommendations for future research on interventions for individuals with ASD. It is important that investigators make an effort to clearly define and report the procedures for the intervention under scrutiny. Researchers should consider the use of standard care as a comparison group (i.e., the treatment that is normally or optimally provided to people with a given condition). In order to allow for comparisons across studies, researchers should use standardized and validated outcome measures so that reporting on the effect of the interventions in terms of changes in core symptoms of ASD is more consistent. The impact of behavioural and developmental interventions upon family outcomes (e.g., functioning, quality of life, and finances) and their possible negative effects should also be further explored.
Studies on the effectiveness of behavioural and developmental interventions for ASD should continue to make improvements to meet accepted methodological standards for clinical research including: the use of randomization and allocation concealment, the implementation of intervention protocols that capture a wide range of skills and symptoms, blinded outcome assessment, assessing treatment fidelity, and implementing longitudinal designs with sufficient follow-up to evaluate treatment effects. One of the limitations of the existing literature is the small samples within individual studies. This limits generalizability and also raises questions around the interpretation of negative findings: whether such findings were due to inadequate sample size or true lack of effectiveness of the intervention. Better reporting of how the studies were planned, conducted and analyzed is required. Established guidelines to this end, such as the Consolidated Standards of Reporting Trials for reporting trials of behavioural interventions and Strengthening the Reporting of Observational Studies in Epidemiology guidelines, should be followed [165].
Programs can vary in terms of the degree of prescription versus flexibility of the intended approach, the extent to which adult control is necessary in fostering children's development of social, communicative and other abilities, the degree of social and natural context of the intervention, the focus on adult versus child centered procedures, the exposure to more natural interactions and learning opportunities, the role of typical peers, and how the major goals of treatment are prioritized. The success of any approach will depend on the needs of the individual, which vary greatly. Rigorous scientific evaluation of the evidence is necessary to estimate the likely benefits of any particular approach. Studies that assist in determining whether an individual is being helped by a particular therapy might be extremely helpful by sparing the burden of participation if no benefits are identified facilitating a switch to other types of intervention.

Conclusions
The most effective behavioural and developmental treatments for ASD should include interventions that address the behavioural, social, and communication deficits associated with the disorder. Intervention studies suffer from methodological problems that preclude definitive conclusions regarding their efficacy. This systematic review tried to elucidate a question regarding the effects of behavioural and developmental approaches to ASD and drew conclusions as to the potential effects of these interventions based on the results of clinical trials and observational cohort studies. Without better operational definitions of the critical components of interventions, consistency in choice and reporting of outcome measures, and enhanced descriptions of participant heterogeneity, we will see few gains in understanding 'best practices.' While this review suggests that Lovaas may improve some core symptoms of ASD compared to special education, these findings are based on pooling outcomes from a few, methodologically weak studies with few participants and relatively short-term follow-up. As no definitive behavioural or developmental intervention improves all symptoms for all individuals with ASD, it is recommended that clinical management be guided by individual needs and availability of resources. Future studies on the effectiveness of these interventions need to be more rigorous.
Further, the evidence needs to be interpreted in light of the study populations, characteristics and application of the interventions, outcomes examined, and methodological quality. Over the long term, providing more rigorous evidence for interventions for children with ASD will contribute to positive outcomes for this population, enabling these individuals to contribute more effectively to the social and economic life of their communities. The past 40 years have seen many gains in the quality and quantity of intervention research for individuals with ASD. Research in this area has provided hope for many families and provided evidence that many individuals can learn and develop beyond earlier expectations. This systematic review summarizes this research and elucidates the many areas in which we have much to learn.

Supporting Information
Supplement S1 Complete literature search strategy