Cognitive, language and behavioral outcomes in children and adolescents with autism spectrum disorders exposed to early comprehensive treatment models: A meta-analysis and meta-regression

Background Early comprehensive treatment models (CTMs) have been developed as effective treatments for children with autism spectrum disorder (ASD). Numerous studies have suggested that CTMs can improve short-term outcomes, but little is known about outcomes in mid-childhood and adolescence. The current meta-analysis reviewed studies reporting broader outcomes in children and adolescents with ASD who had ever participated in a CTM and examined the predictors of developmental gains. Methods We searched eight databases up to June 13, 2019, for relevant trials and natural experiments. Longitudinal studies were selected if they investigated the outcomes of CTMs. Two meta-analyses were undertaken to provide a summary estimate of change in treatment outcomes and to evaluate the effect of CTMs; one used the standardized mean change between the pretest and posttest, and the other was a classical meta-analysis. Stratied and random-effects meta-regression analyses were performed to search for outcome differences among studies. Results Eighteen intervention studies (involving 495 children with ASD) met all the inclusion criteria: 12 used early intensive behavioral intervention (EIBI), and two used the Early Start Denver Model (ESDM). Outcomes were categorized into three parts: cognitive, language and behavioral (e.g., adaptive functioning and symptomatology). Overall, most children with ASD who had ever participated in an early CTM made gains in many areas of functioning, especially in terms of symptom- and language-related outcomes. Stratied analyses indicated that the ESDM displayed the largest effect on IQ improvement (ES=1.37, 95% CI: 0.95-1.80), while EIBI was more effective for symptom reduction (ES=-1.27, 95% CI: -1.96 to -0.58). Further meta-regression suggested that interventions with parent involvement and longer treatment hours yielded greater improvements in IQ and social adaptive functioning, respectively. Conclusion The results demonstrate a positive association between CTMs and better prognosis in mid-childhood and adolescence, especially regarding symptoms and language. However, most extant research involves small, nonrandomized studies, preventing denitive conclusions from being drawn. Clearly, the outcomes of children with ASD are still far from normal, especially with respect to adaptive functioning, and the three mediating variables pertaining to treatment elements can affect their gains, including approach, implementer

often di cult and stressful for individuals with ASD and their families, there is a pressing need for systematic knowledge of the outcomes in midchildhood and adolescence and their predictive factors in children with ASD who have been exposed to a CTM to provide timely support [23].
Above all, the present study aims to extend previous reviews by conducting a meta-analysis and meta-regression of longitudinal studies from early childhood to adolescence. The study aimed to (1) report outcomes for speci c domains of functioning and behavior (including cognition, language, adaptive functioning and symptomatology); (2) discover whether there are signi cant improvements in those outcomes for children with ASD and the effect of the CTMs; and (3) examine the in uence of childhood pretreatment characteristics, study characteristics and intervention elements on gains.

Methods
The protocol for this meta-analysis was registered in the PROSPERO database of prospectively registered systematic reviews (www.crd.york.ac.uk/PROSPERO; CRD42019146859), and the completed study conforms to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [24].

Search Strategy and Selection Criteria
A systematic literature search was performed in eight electronic databases: PubMed, EMBASE, PsycINFO, Scopus, the Cochrane Library, OVID, ERIC, and Web of Science. Each database was initially searched for relevant literature in English from its inception through June 13, 2019. We developed a search strategy for PubMed based on MeSH (Medical Subject Headings) terms and text words from key research that we identi ed a priori (see Table S1 for the full search strings). We reviewed the reference lists of key publications and relevant narrative reviews to identify studies that might have been missed in the database searches. To check for possible publication bias, we also undertook a gray literature search in clinical trial registries (http://www.ClinicalTrials.gov) using identical inclusion criteria to identify unpublished trials.
After the removal of duplicates, two independent investigators performed title scans and abstract reviews, and they screened the full-text articles to assess their eligibility for inclusion. Concordance among the investigators was satisfactory, with a positive agreement of 0.83; any disagreements between the authors were resolved by consultation with the third investigator. A number of prespeci ed inclusion and exclusion criteria were used to select key studies. The inclusion criteria were as follows: (a) randomized controlled trials (RCTs), quasi-experimental studies (i.e., nonequivalent control group design, one-group pretest/posttest design), and natural experiments (a form of observational study in which the researcher cannot control or withhold the allocation of an intervention to particular areas or communities; thus, natural or predetermined variation in allocation occurs); (b) longitudinal studies with at least one assessment in early childhood and one in mid-childhood or adolescence; (c) mean age of participants at rst assessment ("early childhood") <5 years; (d) mean age of participants at last assessment ("mid-childhood or adolescence") between 5 years and 18 years; (e) professional/clinical diagnosis of ASD, autism, PDD-NOS, or Asperger syndrome based on DSM criteria; (f) English-language articles published in a peer-reviewed journal (dissertations were excluded); and (g) articles assessing the effectiveness of a CTM and reporting primary outcome variables focused on child functioning.
The following exclusion criteria were applied: (a) studies including children with medical complications or who were receiving drug treatment; (b) pharmacological or dietary interventions, focused intervention practices (FIP, e.g., Pre-school Autism Communication Trial (PACT), Joint Attention, Symbolic Play and Engagement Regulation (JASPER)), and other interventions with unclear evidence according to National Institute for Health and Care Excellence (NICE) guidance, such as secretin, chelation, or hyperbaric oxygen therapy; (c) studies reporting on a CTM that was not present in at least two other studies, that is, "isolated intervention approaches"; and (d) studies for which pre-and posttest means and standard deviations were not available after attempts to contact the authors and could not be calculated from the descriptive data or statistical tests in the study manuscript. For multiple studies on the same cohort, we selected the publication with the longest follow-up, provided it included results with detailed demographic and intervention information.

Data Extraction and Quality Assessment of the Included Studies
Pairs of investigators independently performed data extraction with a predesigned standardized form, and discrepancies were resolved by repeated discussion until consensus was reached. To ensure the accuracy and completeness of the extracted information, the third investigator repeatedly veri ed the extracted data abstraction for all the included studies. The following information from each included study was extracted: rst author; region, study design, and year of publication; population characteristics at intake, including subtype of sample, age, and sex (% male); intervention characteristics, including intervention approaches (e.g., EIBI, ESDM), setting (clinical/home), implementer (therapists/therapists and parents), intensity and duration in weeks and months; type of comparison (e.g., treatment as usual, implementer, intensity, and no comparison group); assessment times (i.e., pre, post, follow-up); the measures employed in each study; and the outcomes reported in mid-childhood and adolescence (e.g., autism symptomatology, IQ, adaptive behavior, language). Two independent investigators applied the Evaluative Method for Determining Evidence-Based Practices in Autism to assess the quality of the included studies [25], which is available for many study designs. A previous study suggested that this tool can be applied to evaluate intervention studies and produce valid assessments of the empirical evidence on practices in children with ASD [26]. Six primary and eight secondary quality indicators were applied and are annotated in Table S3, including the characteristics of the participants, independent variables, dependent variables, comparison conditions, random assignment, blinding of raters, and delity. Divergence between the two investigators who evaluated the quality of the studies was resolved by discussion. The quality of a study was assessed as "strong" when all the primary indicators received high quality ratings and there were four or more secondary indicators; "adequate" when more than four primary indicators received high ratings, with no unacceptable ratings and evidence of at least two secondary indicators; and "weak" otherwise.

Calculation of Effect Sizes (ESs)
Because the instruments for evaluating a given outcome differed across studies (e.g., Wechsler Intelligence Scale for Children vs. Merrill-Palmer Scales of Mental Tests), we used standardized ESs to obtain standardized measurements of the effect of the intervention on the outcome variables. According to the methodology of Reichow and Wolery [27], two types of ES were computed: the standardized mean change ES (g c ) and the standardized mean difference (SMD) ES (g d ). We took two steps to ensure the most conservative ES. First, ESs were calculated only when the data necessary for the calculation were available. If an outcome variable was missing the necessary data for the calculation of an ES, no ES was calculated for that outcome of the study. Hence, no data were extrapolated or interpolated for the calculation of ESs. Second, ESs based on small samples are known to be biased [28], so we multiplied them by the small sample correction factor [29].
The rst ES analyses were calculated for the intervention groups in all the included studies and examined the differences between the average gains made by distinct samples. This comparison revealed the absolute difference within a sample from preintervention to mid-childhood or adolescence without regard to the comparison group in between-group studies. We calculated the g c by dividing each adjusted mean change by the pooled standard deviation.
For the between-group studies, the g d was used to show the magnitude of the difference between the group receiving a CTM and the comparison group.
The ES (g d ) was calculated by dividing each adjusted mean difference by the pooled standard deviation.

Meta-analytic Procedures
We combined ndings from all the included studies using prespeci ed meta-analytic methods to determine the effect of CTMs in children and adolescents with ASD. Data synthesis involved two steps: (1) Meta-analysis I was performed to estimate longitudinal changes in broader outcomes in children and adolescents with ASD who were exposed to a CTM. (2) Meta-analysis II was performed to assess the effect of CTMs on those outcomes in the test group compared to the control group. The standardized mean change/difference and 95% con dence interval (CI) for each intervention effect were the primary outcome measures in the meta-analysis. Due to the diversity in population characteristics and intervention approaches, we expected a conservative estimation of the ESs. Consequently, a meta-analysis was performed on studies judged su ciently similar and appropriate to pool using random effects models. Cohen's criteria [30] were applied to determine the magnitude of the effect. The magnitude of the effect was assessed as "trivial" when the ES was < 0.2, "small" when the ES was between 0.2 and 0.49, "medium" when the ES was between 0.5 and 0.79, and "large" when the ES was ≥ 0.8.
Prespeci ed and exploratory strati ed analyses were conducted to assess differences in ESs based on the use of (1) EIBI, (2) ESDM, and (3) other interventions to examine the consistency of the intervention approaches. Outcomes reported in fewer than six studies and parental outcomes were discarded from the meta-analysis, and studies were rank-ordered by quality rating in the forest plots.
The I 2 statistic was used to assess the potential heterogeneity of ESs across interventions. An I 2 >50% was considered evidence of heterogeneity. Sensitivity analysis was performed by reanalyzing the data using a xed effects model and by omitting one study at a time to assess the impact of each individual study on the overall pooled estimate. Potential publication bias was assessed in two ways: a funnel plot and Egger's linear regression test. When publication bias was identi ed, a nonparametric trim-and-ll method was used to adjust for the publication bias.

Meta-regression
Although there are certainly variations across the included studies (i.e., varying amounts of time between posttreatment and the collection of follow-up data, total treatment hours), we applied random-effects meta-regression analyses to examine the effect of mediators and moderators on primary outcomes and to explore the potential heterogeneity. A moderator (baseline variable) suggests for whom or under what conditions a treatment might affect the outcome of interest. A mediator (intervention variable) suggests how or why the treatment might work. Three categories were de ned a priori in the protocol: (1) internal validity (risk of bias, sample size), (2) population characteristics (preintervention age, preintervention IQ, time interval between postintervention and follow-up, age at the last assessment), and (3) intervention characteristics (intervention approaches, total intervention hours [duration multiplied by intensity], implementer [therapist, therapist and parent]). To reduce the risk of type II errors, we abstained from performing regression with predictors that were available for < 6 of the included trials, and univariate meta-regression was used for predictors available in 6-10 of the included trials. Only for IQ, which was reported in >10 trials, were all variables that predicted variance (p<0.05) included in a multivariate regression model, and forward elimination was performed. Given the type I errors of the multivariate meta-regressions, we also applied the Monte Carlo permutation test.
A systematic description of eight between-group studies and ten prepost studies (including 495 nonoverlapping participants with ASD) is provided in Table 1. Three of the ten prepost studies with within-subject designs were natural experiments, and the intervention characteristics were reported by parents. Half of the included studies were postintervention follow-ups and thus had a period of time during which the intervention was not being implemented; the outcomes from these studies were de ned as "long-term". These studies used a wide range of measures to assess autism symptom severity, cognitive and language abilities, and adaptive behavior (Table S2). Most employed standardized measures and researcher-developed interviews, and all the repeatedly measured outcomes were standard scores. Moreover, six studies (33%) received the highest rating (strong), two (11%) received the middle rating (adequate), and ten studies (56%) received the lowest rating (weak ; Table S3) based on the assessment of research report rigor. Table 1. Characteristics of the studies included in the meta-analysis reporting multiple outcomes in children and adolescents with ASD.

Participants
Intervention Notes: a Total number of subjects at the last measurement for pre-post studies and subjects in the experimental group for between-group studies.
b Chronological age at which the participants entered the study or started the intervention.
c Intervention characteristics for pre-post studies and the experimental group's features for between-group studies. Male% was not reported in follow-up subjects. We used male% at intake to replace it.
ABA, applied behavior analysis; AD, autism disorder; ASD, autism spectrum disorder; AUS, Australia; CA, chronological age; CA, Canada; DSM, The Diagnostic and Statistical Manual of Mental Disorders; EIBI, early intensive behavioral intervention; ESDM, the Early Start Denver Model; IBT, intensive behavioral treatment; IQ, intelligence quotient; PDD/NOS, pervasive developmental disorder not otherwise specified; T, therapist ; T+P, therapist and parents; N-R, non-random; NR, not reported; NRT, non-randomized trial; NRSI, non-randomized study for intervention; RCT, randomized controlled trial; UCLA, University of California, Los Angeles.

Population and Intervention Characteristics
The mean pre-IQ, reported in 15 studies, was 50-64; the mean pretest age was 24-49 months, and the mean age at the last assessment was 66-192 months. Of the 18 studies included, 12 conducted EIBI (seven applied the UCLA model [39]), two used the ESDM, and four used other interventions.
Other interventions (e.g., community intervention) include the combination of standard interventions. With regard to the intervention characteristics, eight studies were implemented by therapists and parents. The intervention duration and intensity ranged from 6 to 60 months and from 15 to 40 weekly hours, respectively. Six studies reported that participants were receiving supplemental treatments. Moreover, the comparison conditions in the eight between-group studies, which included 6 EIBI programs and 2 ESDM programs, were treatment as usual (k=5), different implementers (k=2) and active comparison (k=1).
Outcomes and Meta-Analysis I: Longitudinal Change in Mid-Childhood and Adolescence Although a number of studies evaluated outcomes across multiple domains, others focused on speci c areas, such as intellectual abilities, adaptive functioning, language outcomes or autism severity. A summary of reported outcomes is presented in Table 2; generally, positive ESs (g c ) suggest that children's performance improved on average after the preintervention stage in multiple dimensions of functioning (see Figure 2 and Figure 3).  f ASD symptom severity was measured by ADOS, ADI-R, and CARS.
g Adaptation composite was measured by VABS. Akshoomoff et al. (2010) reported the subdomains of adaptive functioning and non-verbal/verbal IQ, which are not represented in Table 2.

Meta-Analysis II: Effects of EIBI on Outcomes in Mid-childhood and Adolescence Compared to those in the Control Group
As presented in Figure 4, the majority of the SMD ESs (g d ) were positive, which indicates that the functioning of children with ASD in the EIBI group was generally better than that in the comparison group in multiple dimensions. The controlled ESDM studies and the outcome for ASD symptom severity were discarded from meta-analysis II because of inadequate or isolated data.

Strati ed Analyses
The results for the comparison of the three intervention approaches in the strati ed analyses of meta-analysis I revealed disparate effects. Notably, the ESDM group had a signi cantly higher ES for IQ than the EIBI and other interventions groups (g c =1.37, 0.61 and 1.21, respectively; Figure 2). Regarding other outcomes, the number of ESDM studies is insu cient for comparison. Nevertheless, the opposite occurred for symptom outcomes (ASD symptom severity and social adaptive functioning), as the EIBI group had clearly greater symptom improvement than the other interventions group (g c =-1.27, 0.65 vs. g c =-0.03, 0.19). Additionally, strati ed analyses could not be conducted in meta-analysis II because of the limitations of the controlled studies.

Sensitivity Analysis
Sensitivity analyses suggested that the estimates were not substantially modi ed by any single study. There was an exception for the adaptive composite score, as a small effect with a g c of 0.31 (95% CI 0.002 to 0.62) was shown when Magiati [19] was removed in meta-analysis I. The sensitivity analyses did not yield different ndings after the data were reanalyzed using a xed effects model.

Publication Bias
No sign of publication bias was found in the funnel plots and Egger's test for any outcome.

Meta-regression
Across 9 predictors in univariate meta-regressions (Table 3), four mediators of longitudinal change in mid-childhood and adolescence outcomes emerged: (1) EIBI was more effective in reducing symptom severity than non-EIBI programs, and this explained 64% of the heterogeneity (Coe cient=-1.31, P=0.045).
(2) Higher total and social adaptive functioning were associated with longer total hours of the intervention and explained 78% and 100% of the heterogeneity (Coe cient=0.0001, P=0.021; Coe cient=0.0002, P=0.032, respectively). (3) Higher social adaptive functioning was also associated with a higher risk of bias (Adj R 2 =100.00%, Coe cient=0.78, P=0.026). No potential confounding factors affected the change in DLS.
The multivariate meta-regressions demonstrated a clear effect of implementer (therapist or therapist and parents) on IQ after the p-value was adjusted (P=0.028, Table 4). Speci cally, the involvement of parents in implementing intervention strategies had a more bene cial effect on IQ enhancement than the involvement of a therapist alone. Vineland adaptive composite score -Total treatment hours: Adj R 2 =78.06% Vineland social adaptive score -Total treatment hours: Adj R 2 =100.00% Vineland social adaptive score -risk of bias: Adj R 2 =100.00% Table 4. Results of the multivariate meta-regression analyses by cognitive function.
Notes: a Monte Carlo permutation test was applied to correct type I errors for multiple covariate meta-regressions.

Discussion
To the best of our knowledge, this is the rst comprehensive study to systematically and quantitatively assess a series of developmental and symptom outcomes for children and adolescents with ASD. Overall, we found positive effects of early CTMs on longitudinal changes in intelligence, language development, communication and social adaptation, and core symptom severity in children with ASD but negligible effects on DLS and total adaptive behavior. In addition, there is preliminary evidence to suggest that children in the EIBI group have made greater gains than children in the control group with respect to intelligence, communication, and social adaptation. It is noteworthy that the outcomes and the risk of bias in most of the included studies are not optimistic. Nevertheless, we demonstrated that the treatment characteristics played a major role in the later outcomes for children younger than 5 years of age, which may also apply to some novel interventions.
The ndings from this study are similar to those of a narrative review that examined the long-term effects of early intervention (EI) in primary school [14]. The review included eight eligible studies, 5 of which were also included in our study. Both this review and the narrative review indicate that most children with ASD who have ever participated in a CTM make gains in many areas of functioning. However, only 9 long-term follow-up studies were found based on our inclusion and exclusion criteria. In other words, the number of well-designed longitudinal studies is still insu cient to determine the long-term effects; therefore, more emphasis should be placed on empirical studies in this eld in the future.
Although favorable effects were apparent across most outcomes, language-related outcomes (IQ, receptive language, expressive language, and communication adaptation) were distinctly superior to social adaptation and ASD symptom severity, with ESs approaching 1.2 for receptive and expressive language. This nding is highly consistent with previous findings from a meta-analysis on the effects of ABA intervention in early childhood that included studies with a minimum intervention duration of one year [48] and has been attributed to the amount of time devoted by most behavioral interventions to language and communication skills [49].
In addition, there is some evidence that EIBI leads to a small to moderate effect in youth with ASD compared to the effect of treatment as usual, EIBI parent-mediated or EIBI minimal treatment controls in terms of IQ and Vineland social, communication, and adaptive composite scores. This is particularly noteworthy because these ESs were smaller than those from a Cochrane Collaboration systematic review and meta-analysis of studies comparing EIBI to treatment as usual in the community [50], which found medium to large signi cant positive effects. The comparison types of the controlled studies varied across the included studies, with nearly half of them involving implementer comparison (therapist vs. therapist and parents); strati cation by comparison type was impossible due to the very small number of studies. Actually, the available evidence has proven the effectiveness of parent-mediated EI, showing improvement comparable with that achieved with therapist-mediated EI [51]. Needless to say, the existence of this comparison type would weaken the ES.
It is generally believed that children participating in early CTMs will have a reduced need for support and programs as they go through school [46], but our study highlighted that despite some improvements, the outcomes of children and adolescents with ASD are still far from normal. Thus, ongoing intervention is necessary, especially for adaptive functioning in real life. Even so, almost 30% of US children with ASD did not receive behavioral or medication treatment [52], and multiple gaps were identi ed across all the stages of intervention development and testing from conceptualization to community implementation [53]. These may be crucial issues to ll to improve outcomes for individuals with ASD in the future.
Furthermore, a systematic review [18] of outcomes in late adolescence and adulthood was selected for comparison with our results to draw more reliable conclusions, and improvements in language and symptom outcomes were found in both children and adult populations. Our results, however, showed a signi cant gain in IQ and negative ndings for adaptive functioning and DLS. Analyses of the distinctiveness of developmental trajectories with respect to these outcomes provided evidence of steady and remarkable improvements in verbal and nonverbal IQ from childhood to adolescence when the pre-IQ range in the included studies was 50-60 [54]. Similarly, individuals with moderate adaptive functioning at baseline (standard score of approximately 75) had a stable trajectory [55]. These ndings suggest that longitudinal change could be in uenced somewhat by the baseline level of participants, and our result explains the prognosis of ASD children with moderate functioning in terms of IQ and adaptation at baseline. Viewed from another angle, we did not nd enough studies reporting the prognosis of lower-and higher-functioning ASD. Regarding the negative ndings for DLS, Di Rezze et al indicated that an improvement in trajectory was associated only with lower and improving ASD symptom severity [56], whereas none of the seven studies reported symptom-related data. We did not nd any statistically signi cant population characteristics in the meta-regression, probably because the mean values of preintervention population variables were relatively concentrated among our included studies. Therefore, we propose that developmental and symptom outcomes could affect each other over time, and the effectiveness of CTMs should be examined by controlled studies designed for multiple subpopulations. Furthermore, the environmental factors that may be associated with continued changes in those outcomes from childhood to adulthood remain largely unknown [57] and may be responsible for the difference in the results.
Due to the variation in changes in mid-childhood and adolescence, we sought to explore the sources. Although the ESDM was the most effective in improving IQ and EIBI showed greater e cacy in ASD symptom severity reduction in affected children, we are still far from establishing an evidence basis for the superiority or inferiority of the ESDM program because of the limited number of appropriately designed relevant studies. However, metaregression provided a clear account of the impact of the implementer and intervention approach and veri ed the results of the strati ed analyses: 1) IQ tended to bene t more from intervention programs mediated by parents and therapists, while the ESDM is an intervention strategy with parental involvement; 2) symptoms tended to bene t more from EIBI programs than non-EIBI programs. We did explore whether the quality and sample size of the studies, initial IQ or age of participants were related to deterioration/improvement in all outcomes over time. Only four signi cant associations were identi ed: implementer, total treatment hours, intervention approach and risk of bias; these derive almost entirely from intervention elements.
Makrygianni et al. have also suggested that the program intensity and duration are important predictors of the effectiveness of treatment on adaptive functioning [22]. Thus, insu cient treatment time may account for the negligible effects on total adaptive behavior.

Limitation
The conclusions of this review should be interpreted with caution in light of its limitations. First, very few high-quality studies have speci cally examined outcomes in mid-childhood and adolescence, and the numerous methodological weaknesses of the studies reviewed here limit the conclusions that can be drawn. Given that the studies varied widely in terms of cohort selection, treatment features, and assessment reliability, we could not establish an unbiased way of taking into account all these factors in judging research quality. We strongly endorse the conclusions of some reviews that rated the overall quality of evidence as 'low' or 'very low' using the GRADE system [6]. Nevertheless, according to the current quality assessment, the quality level necessary to perform meta-regression was met, and most of the changes in the outcomes have nothing to do with the quality.
Unfortunately, the LEAP program [58], which has a rigorous research-based design, was excluded from this review because of insu cient initial data.
Second, to achieve a certain statistical power, this study combined single-group prepost studies with between-group controlled studies to analyze the ES, although this approach is somewhat controversial. However, similar results were obtained when we performed the meta-analysis II among the between-group studies only, indicating the reliability of our results.
Third, we used the group average age data as one of inclusion criteria due to a lack of individual raw data; therefore, it is inevitable that some children were preschoolers at follow-up and some were in their late teens. However, our results showed that the age at the last assessment did not affect the gains. We are looking forward to a time when investigators are willing to share their unpublished data, allowing meta-analyses on this topic to be more complete.
Finally, delity measures and standards cannot currently be assumed for studies in this eld, and most did not provide information about additional treatment received after the ntervention services ended.

Recommendations for Future Research
In sum, recommendations for clinicians and researchers planning to conduct empirical studies in this area include the following: (1) employ study designs that use randomized controlled trials whenever possible and match treatment intensity and duration across groups; (2) record the speci c intervention approaches and components in detail and monitor the delity of the intervention process; (3) collect detailed information on education and intervention strategies applied during mid-childhood and adolescence; (4) due to the current need, explore ESDM programs and lower-and higherfunctioning ASD; and (5) focus on follow-up measurement and record the initial measurement as comprehensively as possible

Conclusion
Page 11/17 Overall, there is some evidence that most children and adolescents with ASD who participate in an early CTM make gains in many areas of functioning, especially with respect to symptom-and language-related outcomes. However, most of the existing research relies on small studies that are nonrandomized, forestalling de nitive conclusions. What is certain is that the mid-childhood and adolescence outcomes of children with ASD are still far from normal, especially with respect to adaptive functioning, and the mediating variables of developmental gains were primarily intervention elements, including approach, implementer and total treatment hours. Furthermore, the ESDM displayed the largest effect in terms of improving intelligence development, and EIBI showed greater e cacy in reducing ASD symptom severity. Declarations BJS, LC and JJ were involved in the conception and design of the review. BJS, WJW and MXD contributed to the data collection. JJZ and JYL contributed to the quality assessment. BJS and JJZ conducted the meta-analyses. BJS and MXD contributed to interpretation of data. The review was conducted by BJS and WJW, who completed initial drafts of the paper. LC, BW and JJ gave critical comments and advice that helped shape the review.
All authors read and approved the nal manuscript.

Ethics approval and consent to participate
The protocol for this meta-analysis was registered in the PROSPERO database of prospectively registered systematic reviews (www.crd.york.ac.uk/PROSPERO; CRD42019146859).

Consent for publication
Not applicable.

Additional File Legend
Additional le 1: PRISMA checklist.
Additional le 2: Table S1. Study inclusion and search strategy.    Effect sizes (gc) for IQ, language and symptom outcomes in children with ASD. Note: Hedges' g effect sizes represented in black and con dence intervals are reported. Random effects models were used on all outcomes, and the studies were rank-ordered by quality rating. Abbreviations: ASD, autism spectrum disorder; CI, con dence interval; EIBI, early intensive behavioral intervention; ES, effect sizes; ESDM, Early Start Denver Model; IQ, intelligence quotient.