Given the tremendous implications of this issue at the personal, family, and societal level, in our opinion the identification of predictors of treatment outcomes should be at the top of the autism research agenda. In this paper, we shall argue that information about predictors of treatment response is limited because the current theoretical and methodological approaches are not adequate to address this issue.
Furthermore, we will offer some directions for research, and discuss the implications for clinical practice. While there are a number of treatments in the field of ASD, only a small proportion of these have scientific evidence for their efficacy. Within this framework there are different programs, which vary according to the specific curriculum and teaching procedures 19 — A number of reviews and meta-analyses have examined group-level outcomes in response to early intervention programs 6 , 10 , 22 , indicating that EIBI programs appear to be effective for increasing adaptive behavior and IQ in young children with ASD.
Nevertheless, there are limitations to available research, including lack of randomization in most published studies, small sample sizes, and possible biases due to awareness of treatment status in parents and providers. Moreover, the definitions of EIBI vary across studies, ranging from very specific definitions e. Importantly, available evidence suggests that EIBI programs and early educational programs more in general are not equally beneficial for all treated children 8. In the following, we focus on the factors associated with individual-level variability in response to current early educational intervention approaches, including strictly defined EIBI programs based on the work of Lovaas, as well as programs that are based on behavioral techniques but are not necessarily implemented at a high level of intensity [e.
Analysis of available evidence indicates that pre-treatment cognitive abilities IQ and language abilities are the most often reported correlates of gains in early intervention studies 26 — 33 , although not all studies concur in their conclusions 34 — Several studies also indicate pre-treatment level of adaptive behaviors as a relevant correlate of treatment gains 6 , 37 , Whilst there is some evidence that children who are younger and less severely affected might be more responsive to early intervention, other studies report mixed or negative findings [ 31 , 39 — 43 ; see also Ref.
A number of studies have identified more specific abilities associated with positive treatment outcomes, including play skills 24 , 45 — 47 , interest in objects 48 — 50 , joint attention 46 , 47 , 50 , imitation 38 , low social avoidance 51 , and response to social versus non-social reinforcement In addition, there have been mixed reports on family factors such as the level of maternal education 53 , 54 and family stress 55 , Despite such findings on predictors, knowledge of the factors underlying positive treatment outcomes is limited and inconclusive.
The major limitations to current research approaches are outlined below. Variables associated with change in treated groups do not necessarily reflect actual predictors of outcomes i. If factors other than treatment that might contribute to change are not considered and controlled, correlates of changes within a treated group might or might not tell the full story about prediction of treatment response. The selection of predictor variables is often not theory-driven 59 , In most papers reporting on predictors of outcomes, no rationale is provided on why specific variables are selected for analyses.
It is apparent that, in most cases, the predictor variables are the measures used to characterize samples at trial commencement, with the purpose of identifying any important differences between randomized groups. Equally, the types of measures used often indicate broad constructs and thus lack the specificity required for a predictor variable. Analyses of specific behavioral predictors that could account for individual or subgroup differences are rarely included in research on intervention and prognosis Most intervention studies focus on overall group-level outcome data, and analyses on those factors associated with identifiable subgroups of children or individual differences are either not conducted or not reported, despite the fact that treatment response is so variable in ASD.
However, as current research emphasizes the heterogeneity of ASD at every level of analysis, and the results of trials and meta-analyses show variable effect sizes, there is a clear need to move to the next stage of intervention research, with a focus on the specific individual predictors of treatment outcomes, and a focus on moderators and mediators of treatment response.
Few studies compare predictors of outcomes across different intervention programs The impact of child characteristics is likely to vary according to the intervention program implemented, so that a child who is not responsive to program X e. Since different programs utilize different instructional techniques for example, different emphasis on external versus social reinforcements, different emphasis on verbal versus non-verbal instructions , it is likely that children with varying intrinsic cognitive and learning profiles will respond preferentially to different teaching approaches, just as do children who do not have ASD 60 , 62 , However, the majority of studies on predictors of outcomes include data on response to one program only.
Moreover, the inclusion of only one treatment leaves open the possibility that the change predicted by the pre-treatment factors is due to factors other than treatment Intervention studies recruit different samples from those that present at clinical services. Exclusion criteria in many intervention studies result in children with associated medical conditions, seizures, or a low IQs [e. While these exclusion criteria are set to create homogeneous samples for research purposes, they reduce generalizability to those who are seen in clinical and community-based settings, providing little information about which programs should be recommended based on different pre-treatment characteristics in the wider ASD population.
Therefore, research on their responses to different programs is crucial to inform clinical practice, so that enrollment to an intervention service is most likely to offer benefits and cause no harm. Current guidelines in clinical trials suggest that excluding patients with associated disorders from research should not occur when these comorbidities are common or when they affect treatment response and prognosis Standard predictor measures currently used in research are very broad.
Omnibus factors such as tested IQ, speech and language assessments, and adaptive behavior have predominated as both predictors and outcome measures in intervention research 53 , The use of such broad measures in intervention studies as predictors is problematic for a number of reasons.
Low scores in IQ, language, and adaptive behaviors reflect a variety of distinct underlying processes, making it difficult to understand the specific mechanisms underlying the intervention response.
Given that performance on IQ tests is, in itself, a measure of learning abilities e. To avoid circular reasoning children who have more difficulties in learning, as measured through pre-treatment IQ testing, are the ones who will have more difficulties in learning from educational treatments , research needs to focus on more proximal predictor variables.
These should reflect specific and clearly defined processes that might explain difficulties in responding to educational strategies [e. Moreover, broad variables such as IQ and communication scores are not robust measures in children younger than 3 years 71 , and different tools used to measure IQ might provide different results depending on the instruction formats e.
Family factors are seldom considered as predictors in outcomes studies. The motivation of parents to pursue and persist with intervention programs, which involve considerably more effort than administering medications or using complementary and alternative therapies, may be an important factor in treatment outcomes. Lack of research in this area is surprising, given that 1 parents are frequently expected to engage as the main therapists for their children in many interventions, even those that are not called parent-mediated interventions, and 2 family factors have been found to impact on treatment response across a number of intervention programs for children with other conditions [see Ref.
Moreover, family factors such as higher parenting stress, negativity and depression, and low SES are ubiquitous factors in poorer outcomes across a range of child mental health interventions 74 — Another study 79 reported that higher distress in mothers pre-treatment was associated with lower adaptive behavior outcomes post-treatment, although the effect was not statistically significant [see also Ref. Similarly, Osborne et al. Despite the relevance of this literature, the available evidence on family characteristics moderating treatment outcomes is limited and inconclusive.
de.iniqoduwuhuk.tk On the basis of the limitations in the intervention outcome literature to date, a number of recommendations are made here. Selection of putative predictors should be theory-driven, and predictors should be proximal and specific rather than broad. In order to match specific learning profiles to specific teaching programs, it is important to conduct a fine-grained analysis of child characteristics. Doing so will enable us to determine not only what the child needs to learn which will inform treatment objectives but also how the child learns which will inform treatment strategies.
In addition, we also need to develop a fine-grained understanding of how each treatment works; that is, what are the processes or the active ingredients of the intervention that interacts with the child characteristics to promote learning in that child? The analysis of the active ingredients of treatment involves a conceptual distinction between moderators of treatment effects, and mediators through which the intervention supposedly works Moderators of treatment outcomes are the pre-treatment characteristics that might determine the degree of effectiveness of treatment versus control, but do not change as a consequence of the intervention, such as chronological age, gender, or maternal education.
Conversely, mediators of treatment outcomes are the factors through which a treatment exerts its effects: they are subject to change as a consequence of the treatment, and these changes, in turn, affect treatment outcomes.
Similarly, it is plausible that changes in the propensity to imitate others mediate outcomes in Reciprocal Imitation Training 24 , and that changes in the ability to understand and follow visually mediated task instructions would be relevant in response to the TEACCH program The study of mediators of treatment outcomes requires the knowledge of the learning processes upon which the instructional techniques of the teaching program are based, or, in other words, understanding of the active ingredients underlying treatment-related changes.
Without such knowledge, selecting among the many variables that are potentially associated with response to treatment in ASD is a difficult task.
A clear definition of the processes through which the child is able to learn in response to the particular instructional techniques is therefore one starting point for defining a specific set of putative moderators and mediators. Furthermore, it is crucial to focus on proximal factors that are known to support learning, with different predictor variables reflecting distinct and defined processes, so that the specific weight of these putative predictors mediating treatment response can be measured. All of these processes reflect different facets of social cognition that are known to support social learning, and which are associated with developmental outcomes in ASD as well as in typical development [e.
Preliminary evidence provides encouragement on the value of these factors in predicting response to intervention [e.
For example, Watt et al. Data on the relevance of RRBs in response to treatment are scant and equivocal 52 , 97 , 98 , so more research is needed to investigate how individual differences in the extent of RBRs affect response to intervention. Other factors that are not specific to ASD might also be associated with treatment outcomes across intervention programs. These include attention in particular sustained attention, e. Rate of learning e. Research on the predictive value of these factors in response to treatment is scant.
As standardized tests are not available for many of the processes listed above, it is necessary to develop and utilize novel, fine-grained experimental measures and observational protocols that are suitable for young children with ASD across the spectrum of severity [e. Family factors should be investigated in treatment studies. As family involvement is a recommended component of early intervention , future research should systematically investigate the family characteristics associated with responses to treatment for children with ASD.
Importantly, different early intervention programs involve instructional techniques e. High levels of child behavior problems are prominent in families with children with developmental disabilities and contribute to greater stress in families than cognitive delay , Generally, mothers of children with ASD report higher levels of distress than those of non-disabled children or those with children with other disabilities, although there is considerable variability across families , In a recent paper, Benson has shown that social network attributes, including the range and function of emotional support, are related to perceived social support, which in turn can bring about a decrease in depressed mood in mothers of children with ASD.
Formal and informal social support networks can have the effect of enhancing quality of life, confidence in parenting and optimism for families with children with disabilities, and might be crucial factors predicting treatment outcomes. Accurate measures that appropriately capture individual differences between families in terms of specific values, attitudes, and resources may need to be developed. Future research should investigate the role of family factors in a systematic way, selecting specific, theory-driven variables, developing novel fine-grained measures and comparing variables associated to outcomes across different programs.
When examining a broad array of putative factors e.
Studies looking at longitudinal outcomes in at-risk populations often use cumulative risk approaches in which a discreet number of risk indexes are created to capture the level of risk within a number of predefined theoretically driven domains. To illustrate, a number of studies [e. This approach can be helpful in examining whether risk factors in family related and sociodemographic domains play a role in response to treatment in the ASD population. The analysis of individual differences i. Moreover, when predictor variables are measured, their association with the outcome measures should always be reported.
The tendency to report on predictors of outcomes only when results indicate significant associations makes it hard to draw clear conclusions on the robustness and consistency of factors associated with treatment response. Thus, it should be mandated that all data on predictor and outcome variables are reported, as in the case of outcome measures in clinical trials Reporting all available information on the association between predictors and outcomes can also be useful in studies with small sample sizes, which comprise the majority in the field of ASD, so that data from multiple small trials can be pooled for meta-analyses.
While data on correlates of treatment gains per se do not allow for definite conclusions on predictors of treatment outcomes, this information can be critical to inform subsequent research designed to control for the predictive value of factors that appear to be relevant in relation to treatment changes. Research on predictors of treatment outcomes should compare responses across different programs.
The IEP should not be limited to academic skills, but should include focus on language, social skills, and developing friendships. This trusting relationship enables the child to become a well-adjusted adult later in his or her life. It also forms a stable base for all future relationships. For example, if a child picks up a toy, the adult may show what can be done with it by demonstration and prompts. When hungry, he can ask for food; if he needs help he can call his mother instead of crying. Although the articles describing intervention processes include participants, a meta-analysis is not possible due to the lack of comparable inclusion and characterization criteria.
As mentioned above, the analysis of correlates of gains in a treated group might not be informative on predictors of treatment outcomes, as not all observed change in a treatment study can be attributed to the treatment. Research on predictors of treatment outcomes should use designs that compare different programs, thus allowing for the analysis of the interaction between selected participant characteristics X treatment group.
Different early intervention programs, while sharing many similarities, involve distinct instructional techniques that are based on different theories and which tap into different learning processes. Similarly, programs vary across a number of teaching processes e. Since these different teaching practices require different learning processes on the part of the learner, it is possible that different types of learners will respond to different types of teaching procedures.
Given that remarkable differences are present both in the teaching procedures of the different programs and in the social, cognitive, and learning profiles of children with ASD, it is imperative that future research compares the profiles of response to different early learning programs that use different instructional techniques. This approach would allow identification of the pre-treatment child characteristics that support learning when a specific set of teaching procedures is used.
Only a few studies to date have provided information on the specificity of predictors of outcomes to one versus another treatment program [see Ref. Research on predictors of outcomes should involve large heterogeneous samples.
Many children with comorbid conditions are typically excluded from research studies, such as those with severe intellectual disability and associated medical conditions or seizures. However, as discussed above, these children represent a substantial proportion of the ASD population referred to community early intervention centers. It is particularly important that research focuses on children with very low IQ to determine whether the currently available early intervention programs are appropriate for this population, and which specific factors are predictive of positive outcomes e.
Table 1. Little is known about predictors, moderators, and mediators of treatment response in children with neurodevelopmental and neuropsychiatric conditions , , and the field of ASD is no exception. With recent advances in research documenting the positive impact of early intensive behavioral programs for children with ASD, the critical issue facing researchers, clinicians, and practitioners in the field is not as much a lack of evidence-based treatments, but rather an inability to predict which treatment will work best for each child.
The variability of response to EIBI and other early intervention programs is a phenomenon that very likely reflects the heterogeneity of the ASD population. While some crude prognostic variables, such as IQ level, hold some value in predicting which children will respond best to early intervention, to date, the available information on moderators and mediators of treatment response is inconclusive. This model is illustrated in Figure 1.
Current understanding on the interplay of these factors in determining treatment outcomes is limited, and we have offered here a number of recommendations to advance knowledge in the field.