ABA-Based Approaches

One fair quality RCT examined the use of the Learning Experiences and Alternative Program for Preschoolers and Their Parents (LEAP) protocol in preschool classrooms in the United States.⁸⁴ The study compared 27 classrooms (n children=177; mean age: 50.1 months±4.6 months) with teachers trained in the full LEAP curriculum (including peer mediated social skills, incidental teaching, pivotal response training, the Picture Exchange Communication System (PECS), and positive behavior support) to 23 classrooms (n children=117; mean age: 50.7 months±4.2 months) where teachers received the LEAP manual but no formal training. Both groups received an average of 17 hours per week of intervention over two years. Relative to the manuals-only group, children in the full LEAP training classrooms showed significant (p< .05) improvement on investigator-rated Childhood Autism Rating Scale (CARS) scores, language, cognitive, and social skills measures. The students of teachers rated as having better intervention fidelity showed better outcomes on all measures.

Five additional studies (reported in multiple publications) examined the use of school-based ABA programs (one fair quality nonrandomized controlled trial and four fair quality prospective cohort studies. ^{76-77, 78, 81-83} All five compared standard special education preschool curriculums to special education preschools with some sort of enhanced intervention modality, including general ABA,^{81, 82} individual UCLA/Lovaas-based behavioral intervention,^{76, 78} Treatment and Education of Autistic and Communication related handicapped Children (TEACCH)- or LEAP- programs,⁸³ and a mix of behaviorally-based operant conditioning techniques.⁷⁷ Mean treatment intensity ranged from 13.8–28.38 hours per week, with length of enrollment varying from 8–24 months. Mean child ages ranged from 25.1–53.5 months.

The effects of enhanced school-based interventions relative to standard special education preschool curricula were mixed. Some studies^76-78 found that the enhanced intervention groups showed greater gains in cognitive outcomes and parent-reported adaptive skills. Other studies found that children in all groups improved on cognitive, adaptive, and ASD symptom measures^{76, 81-83} regardless of intervention type, although in some cases treatment groups showed greater improvements.⁷⁶ Others found declines in both groups on standardized scores of motor skills.^{81, 82} Intervention efficacy was associated with baseline cognitive scores in one study of TEACCH classrooms,⁸³ with lower baseline cognitive scores associated with more improvement. Lower baseline ASD severity was associated with parent-reported cognitive and adaptive growth for children who received eclectic vs. ABA intervention, but not with standardized test scores.^{81, 82} Additional UCLA/Lovaas-style intervention over-and-above classroom involvement was associated with reduced ASD symptoms as rated by clinicians on the CARS⁷⁸ but not as rated by parents using the Scale of Pervasive Developmental Disorder in Mentally Retarded Persons.⁷⁶ Where examined, total hours of intervention per week were not associated with cognitive or adaptive outcomes, although hours were similar across intervention groups within each study (e.g., comparing half-day programs to other half-day programs).

Four studies (reported in multiple publications) compared ABA-based early intervention to eclectic treatment as usual.^{73, 79, 80, 85-91, 110} One good quality RCT compared ESDM to community-based interventions.^{73, 85} It randomized children into two groups based on gender and IQ. For two years, 24 children in the ESDM arm (mean age: 23.9±4.0 months at study entry, mean IQ: 61.0±9.2) received 1:1 therapist-delivered manualized intervention (mean of 15.2±1.4 hours/week) as well as parent-delivered treatment (mean 16.3±6.2 hours/week). The comparison group of 21 children (mean age: 23.1±3.9 months, mean IQ: 59.4±8.6) received individual (mean 9.1 hours/week) and group (mean 9.3 hours/week) therapies, including speech-language therapy, occupational therapy, and developmental preschool enrollment. The ESDM intervention targets social communication and engagement as well as general child development. After one year of treatment, The ESDM group showed significantly greater improvement in IQ but not adaptive behavior. After two years of treatment, the ESDM group continued to show significantly more IQ improvement as well as receptive and expressive language. Both groups improved in all domains of adaptive behavior but socialization, with greater improvements in the ESDM group. Neither group showed significant differences in Autism Diagnostic Observation Schedule (ADOS) severity scores or repetitive behavior, although the ESDM group demonstrated a diagnostic shift toward a milder diagnosis (PDD-NOS) at followup. Electroencephalography (EEG) measures of engagement and cognitive processing for children in the ESDM group with usable data were comparable to typically developing children.

A good quality non-randomized trial from Europe⁹¹ compared children (mean age=62.52±16.96 months) with ASD and co-occurring intellectual disability receiving school-based ABA therapy (n=20) to a matched control group of children receiving care-as-usual (e.g., enrollment in TEACCH classrooms, PECS; n=20). The intervention group received one-to-one treatment (mean=4.98 hours, SD=1.45; range: 1.32–7.11) from master's level interventionists. Eleven participants received 2 years of treatment and 9 participants received only one year due to funding loss and school noncompliance. Independent ratings indicated high treatment fidelity (90.3%). Monthly meetings between therapists and parents and teachers provided them with strategies on skill instruction and maintenance.

Cognitive, adaptive, and language skills, and ASD symptoms, were assessed at 12- and 24-months after starting treatment. Both the treatment and control groups showed significant improvement from baseline to 12-month and from 12-month to 24-month followups in developmental age, adaptive skills, and receptive language, with the treatment group showing significantly more improvement than the control (p values<.04, effect sizes ranging from 1.09–2.61). The treatment group showed significant improvements in IQ (p<.001, effect size=.40) between baseline and the 12-month assessment but not the 24-month assessment, whereas the control group did not show significant IQ gains at either time point. Fifty-five percent of treatment group participants showed reduced levels of intellectual disability post-treatment versus 5 percent in the control group. The treatment group (but not the control group) also had reduced ASD symptoms as measured by the CARS and ADOS (p values<.01, effect sizes 1.50–1.51). Neither group significantly improved in expressive language. Multiple baseline child factors were significantly correlated with progress over time, including developmental age, hours of treatment per week, IQ, adaptive and play skills, and receptive language.

A fair quality Canadian retrospective cohort study^{86-90, 110} matched children receiving a large-scale, publicly funded, community-based early intensive intervention program that incorporated ABA, discrete trial training, and naturalistic approaches (n=61, mean age=42.93±11.53 months) to waitlisted children receiving care-as-usual (n=61, mean age=42.79±10.51 months). The intervention group received treatment (mean 25.81±3.44 hours/week) conducted by trained instructor therapists in specialized centers, preschools, and the home environment. The control group received a mean of 17.9±12.3 hours/week of school-based services and <10 hours/week of behavioral intervention conducted by community-based interventionists in community settings. The approaches included low-intensity ABA, speech and occupational therapy, and behavioral consultation. Children in the treatment arm were enrolled in treatment longer (mean=27.84±8.11 months) than children in the waitlist group (mean=17.01±2.81months), and analyses controlled for this difference.

ASD severity improved for the treatment group compared with control, as did Vineland composite standard and ratio scores and IQ estimates (p values≤.033, effect sizes ranging from 0.53 to 0.83). Although treatment group participants had cognitive scores an average of 19 points higher than controls at followup, this should be interpreted with caution due to a lack of baseline cognitive data. Outcomes were related to age at enrollment, treatment duration, and higher baseline adaptive scores, with duration becoming nonsignificant after accounting for group membership (correlation of duration, group=.57, p<.01). A significant interaction emerged between age at enrollment and group membership, with younger starting age influencing outcomes for the treatment group but not control. Analyses including participants in the cohort study and additional participants found that younger age at intake, higher initial developmental levels⁹⁰ and treatment intensity^{88, 90} were related to treatment outcomes.

Additional analyses of some children in these earlier Canadian studies (overlap not clear) assessed the effects of baseline age and IQ on cognitive and adaptive outcomes in 207 children, and, in a separate analysis of matched older and younger children, effects of baseline age on the same outcomes.¹¹⁰ In the initial retrospective analysis of 207 children, participant ages at intake ranged from 2 to 14.5 years, IQ from 10 to 104, and mental age from 3 months to roughly 7.5 years. Higher baseline IQ and younger age were significantly associated with greater cognitive rate of change (pre-post change in mental age/time in intervention) and with higher IQ at followup (all p<.001), but change in IQ was not significantly associated with higher initial IQ. Higher baseline IQ was also associated with higher adaptive behavior scores at followup (p<.001), but age was not a significant predictor. Longer duration of intervention was associated with slower rate of IQ and adaptive behavior development (p values ≤.01); however, as this analysis was not prospective, the children who received more intervention could have been making slower progress. In the analysis of older (n=60, age 6–13.58 years at baseline) and younger (n=60, age 2.08–5.92 at baseline) children matched on developmental trajectory (i.e., number of intervention hours, baseline IQ and adaptive behavior), younger children had significantly better followup IQ outcomes compared with the older group. Younger children gained an average of roughly 17 IQ points (effect size=0.80) while older children gained an average of 2 points. Cognitive rate improved significantly for younger (effect size=3.19) but not older children. Both groups improved over time in adaptive behavior, but differences between groups were not significant (improvement of 4 points in younger children and 5 in older).

Finally, a poor quality UK study^{79, 80} compared the long-term effects (2 years post-treatment) of 1:1 home-based early intervention (both university-provided and privately-provided) to community-based treatment-as-usual, including PECS, TEACCH, and medication. The early intervention group included 23 children (mean age=35.7±4 months; mean IQ=61.43±16.43 months), and the community-based group included 18 (mean age=38.4±4.4 months; mean IQ=62.33±16.64) at the two-year followup, with children in the community-based group significantly older at the start of treatment (p< .05). For 24 months, children in the early intervention group received an average of 25.6 hours/week of ABA-based intervention using discrete trial training in the home environment, whereas children in the community-based arm received an unspecified amount of eclectic treatment. After 24 months of intervention, IQ, mental age, and language comprehension/expression improved significantly for the ABA group versus community-based (p ≤.05; effect size for IQ change=0.77). At the two year followup, IQ gains were only maintained for children who received privately-provided ABA-based intervention. IQ remained stable for children in the community-based group and significantly declined for children who received university-provided intervention (effect size=.49). This result is confounded by nonrandom assignment and the fact that at baseline, the university-based group had higher levels of ASD symptoms, lower levels of adaptive behavior, and fewer total intervention hours.

Parent Training Approaches

One good quality RCT examining parent training⁹⁶ was conducted in Australia and compared two variations of the Building Blocks® program—home or center-based-- to waitlisted controls. The program targeted social and communication skill development. Mean child ages at enrollment ranged from 41.5 to 43.7 months. Mean IQs ranged from 57–66. Treatment duration was 40 weeks. Not all enrolled children had autism spectrum diagnoses; the breakdown was 100 percent of the home-based group, 82.8 percent of the center-based group, and 78.6 percent of the control group. To be enrolled in the center-based group, children had to have a baseline level of “social maturity,” a lack of “high levels” of problem behavior, and parents willing to attend sessions. The home-based group (n at followup=27) received individualized 2-hour visits every 2 weeks in the home environment. Center-based children (n=29) received weekly manualized, 2-hour, center-based sessions in small groups of 4 to 6 children, as well as parent training and a parent support group. The control group (n=28) comprised a non-randomized treatment comparison waitlist. All groups received concomitant additional interventions classified as educational (home-based: 2.37 interventions±1.28; center-based: 2.41±1.50; control: 3.11±1.64) or ASD-specific (home-based: .22±.42; center-based: .14±.35; control: .54±.79). Providers were multidisciplinary teams of teachers, speech-language pathologists, occupational therapists and psychologists.

Children in all three groups showed significant improvements in Vineland Communication scores. Compared with the home-based group, children receiving center-based intervention had significantly greater improvement in language comprehension and expression as measured by the Reynell Developmental Language Scales. Waitlisted children had significantly greater improvements in followup Vineland Socialization scores than children in either treatment group. No other significant differences emerged among the three groups on other child outcomes. When analyses were limited only to children with autism spectrum diagnoses, the magnitude of the effects increased, but the presence of statistical significance did not change.⁹⁶

Another good quality RCT from the United States ¹⁰⁹ compared the language development of two groups of children with ASD diagnoses, one whose parents received training in a component of ABA, Pivotal Response Training (n=20, mean age=29.5 months, SD=6.9), and another whose parents received training in PECS (n=19, mean age=28.9, SD=4.2). Exclusion criteria included having more than nine intelligible words and having primary diagnoses of intellectual disability, neurological pathology or major sensory impairment. Participants were matched on word use, age, and cognitive functioning. Over the course of 23 weeks, parents completed weekly or biweekly 2-hour parent training sessions from doctoral students. Participants received one-on-one treatment in the home (mean=247 hours, range=181–263) from undergraduate student therapists. Therapist and parent-educator fidelity was maintained at 80 percent. Participants continued to receive outside interventions (e.g., speech therapy) and this was monitored via weekly parent report, with no significant between-group differences emerging.

Outcome variables were assessed at intake, immediately post-treatment, and three months post-treatment. Not all post-treatment coders were blind to participant condition, but no differences were found across blinded vs. not blinded sites. Data were not available on all participants at followup (n=38 for Mullen Scales of Early Learning, 35 for MacArthur Communicative Development Inventories, 35 for Vineland). No differences emerged between Pivotal Response Training and PECS groups. Both improved similarly on all variables over time, with effect sizes across collapsed groups of .216 for expressive communication (Mullen Scales of Early Learning; p<.001), .486 for words produced (MacArthur Communicative Development Inventories; p<.001), and .110 for adaptive communication skills (Vineland; p=.037). The authors reported significant variability in participants' responses to treatment. Parent satisfaction post-treatment was similar across groups, with the only significant difference being the parent-reported difficulty of PECS (p=.005).

Two prospective cohort studies also received good quality ratings. The first was conducted in Australia¹⁰⁰ and compared professional-led parent training (n=17; mean child age, 36.38 months±7.54; 88.2% male) to a self-directed video-based parent intervention (n=22; mean=35.71 months±6.92; 72.7% male). Nearly 80 percent (77%) of participants were diagnosed with autism and the rest with an ASD. Mean IQ was 53.06±9.06 for the professionally led group and 52.86±6.53 for the video-based group. Exclusion criteria included being enrolled in early intervention, passing the Modified Checklist for Autism in Toddlers (M-CHAT), or receiving more than 20 hours/week of services. No information was provided about manualization.

In the professionally led group, parents attended a two-day group workshop and completed a series of 10 hour-long home visits, which occurred two times a week for 5 to 6 weeks. These visits focused on parental stress and child communication. In the video group, parents received an interactive instructional DVD called “Being Responsive: You and Your Child with Autism.” They independently completed video lessons and accompanying worksheets. Followup assessments were conducted 3 months after treatment finished. All outcomes were based on parent report. Children in the professionally led arm showed significantly greater improvement in social communication than children in the video-based arm, regardless of baseline scores. Parents in the professionally led group also reported reduced child-related stress relative to parents in the video group, with fathers reporting more stress than mothers in both groups. Parents in the professionally led group with low baseline self-efficacy reported higher followup self-efficacy levels than parents in the video arm.¹⁰⁰

The second good quality prospective cohort study was conducted in Italy and reported in two papers.^{101, 102} It compared staff- and parent-led ABA-based intervention ( n=24, 92% male; mean age=55.67±17.63 months) to eclectic community-based therapy (n=20, 95% male; mean age=41.94±13.07 months). Group assignments were not random and were based upon parental preference. Children were excluded based on the presence of major medical issues. In the parent training group, children alternated between one week (average of 25 hours) of therapist-led center-based intervention (discrete trial training, incidental teaching, natural environment teaching) and 3 weeks (average of 14 hours/week) of parent-led home intervention. Treatment focused on individual skills, problem behaviors, and facilitated play and social interactions. In the eclectic group, children received in-home developmental and cognitive behavioral treatments (approximately 12 hours/week) with minimal parent involvement. Treatment goals were based upon staff expertise and preferences.

Compared with the eclectic group, children in the parent training arm showed a significant decrease in ASD symptom severity and increases in language production and mental development. The parents of children in the eclectic group reported that their children showed improved socialization and motor skills, but this was not confirmed by behavioral observation. In the parent training group, older children achieved better adaptive behavior outcomes; younger children made more gains in early language comprehension and production. Children who gained more language comprehension had higher adaptive behavior scores pre-treatment. Pre-treatment language comprehension predicted post-treatment language production. In the eclectic group, higher pre-treatment mental development state and early language skills predicted better outcome on parent-reported adaptive behaviors. Initial higher adaptive behaviors predicted better post-treatment early language comprehension. In both groups, child outcomes on early language skills, mental developmental state, and adaptive behaviors were significantly influenced by self-reported parental stress, children's ability to respond correctly to prompts, the number and difficulty of treatment targets, and children's problem behaviors in sessions. Children who were perceived by their parents as more difficult had less improvement in ASD severity.^{101, 102}

Two studies compared interventions focused on increasing parental responsivity. A good quality RCT from Europe (reported in two papers)^{72, 105} compared treatment-as-usual + a manualized, communication-focused parent training (n=14, median age 48 months) to treatment-as-usual alone (n=14, median age=51 months) over 12 months. The intervention focused exclusively on parents and targeted increased parental response to child communication. The additional targeted treatment consisted of a recommended 30 minutes/day of parent-led intervention. Parents received monthly training for 6 months followed by training every 2 months for another 6 months. The intensity of treatment as usual alone was not reported but approaches consisted of speech pathology and ABA-based community treatments. The additional treatment group showed improvements in ASD symptoms, expressive language, and number of communicative acts during interactions with parents. Parents in the additional treatment group showed increased responsiveness to their children during videotaped interactions, which was correlated with reduced ASD symptom severity. No between-group differences were found in adaptive behavior or parenting stress. Greater language gains were seen in children who were younger with lower functioning levels at baseline.

A second fair quality RCT conducted in the United States also focused on enhancing parental responsivity and child communication.⁹³ It compared Hanen's More Than Words intervention to treatment-as-usual. The More Than Words group ( n=29, mean age=21.11±2.71 months) received eight manualized group sessions with parents only and three in-home individualized parent-child sessions over a span of 3.5 months, whereas the control group ( n=26, mean age=21.61±2.82 months) received no treatment or treatment as usual. There was no treatment effect on parental responsivity. The More Than Words group showed differential effects on child communication depending on children's baseline object interest; children with lower levels of baseline object interest had greater growth in communication skills, whereas children with higher levels of object interest showed attenuated growth.

A good quality RCT conducted in the United States compared the manualized Assessment Evaluation and Programming System for Infants (AEPS) with and without additional joint attention and social interaction opportunities.^{103, 104} Both the AEPS group (n=24, mean age=28.6±2.6 months; mean intervention hours=205.66±18.63) and the control (n=24; mean age=28.8±2.8 months; mean intervention hours=196±21) received identical amounts of classroom-based treatment (10 hours/week), home-based parent training (1.5 hours/month), parent education (38 hours), and intervention methods. However, AEPS children received extra training in “interpersonal synchrony,” targeting the three outcome variables of socially engaged imitation, initiation of joint attention, and shared positive affect. No significant (p<.05) differences emerged post-treatment on variables of interest. At the 6-month followup, the AEPS treatment group engaged in significantly more socially engaged interaction than controls (p<.05), with most of the growth in this skill occurring during the treatment period (p<.05) but not during followup (p=.24). No between-group differences were found for initiations of joint attention, shared positive affect, expressive language, or nonverbal problem solving. The AEPS group showed significant growth over time for all variables (p values<.01), but the control group only showed significant growth for expressive language (p=.01). Combined group analyses including 34 children from both the AEPS and control groups examined long-term outcomes an average of 37.6 months after the end of treatment (mean participant age=72.6±months). In this sample, cognitive skills and Vineland-II communication standard scores increased significantly from baseline (mean change 21.4±22.9, effect size=1.02, p<.001 and 12.7±19.4, effect size=0.81, p<.001, respectively), but there was no significant change in ASD symptom severity based on the ADOS.

A fair quality prospective cohort study⁹⁵ compared outcomes for four different types of intervention after 9 months of treatment: 1:1 home-based, manualized ABA (n=14, mean age=39±6.9 months); special education classroom enrollment (n=21, mean age=41.5±4.0 months); comparatively low-intensity, home-based, manualized behavioral intervention (“portage;” n=18, mean age 39.5±6.3 months), and 1:1 behavioral intervention (“local authority”) that included an intensive introductory 5-day parent training component (n=13, mean age=40.2±6.3 months). The home-based ABA group received an average of 30.4 hours/week of intervention, 28.3 of which were 1:1. The special education group received an average of 12.7 hours/week (3.1 hours 1:1). The portage group received an average of 8.5 hours/week (6.5 1:1), and the local authority group received an average of 12.6 hours/week (12.2 1:1). Participants were not receiving any other teaching interventions during the study.

Post-treatment, mean cognitive and adaptive scores were not significantly different across groups. Children in the home-based ABA group showed significant improvements in educational outcomes as measured by the British Abilities Scale relative to other groups (p<.05). The authors created composite scores based on cognitive, adaptive, and educational functioning, but between-group comparisons only approached significance (p<.06). Baseline ASD severity and total intervention hours did modify effects of treatment significantly. First, baseline ASD severity was inversely related to composite change scores for all but the home-based ABA group and was positively related that group. That is, children with more severe ASD symptoms made more progress in ABA and less in the other intervention groups. Second, more intervention time was negatively related to composite change scores for children in ABA but not in the other groups. More hours of ABA were associated with less progress relative to school enrollment or other home-based interventions.⁹⁵

One fair quality U.S. RCT (reported in two papers) compared parent-delivered ESDM to community-based treatment-as-usual.^{98, 106} The ESDM group included 49 children (mean age 21.02±3.51 months, mean developmental quotient [DQ]=64.88±17.22); their parents completed 12 1-hour sessions that included manualized parent-training and coaching. Both the ESDM group and the community group (N=49; mean age=20.94±3.42 months, mean DQ=63.08±15.93) continued receiving community-based treatment-as-usual services as well, including the Developmental, Individual Difference, Relationship-based (DIR) model, TEACCH, ABA, and occupational and speech therapies (range of hours: 0–15.9), with the community group receiving significantly more intervention hours at the second time point (mean 3.68 vs. 1.48; p<.05). Compared with the ESDM group, children in the community-based arm had more severe social affect deficits, poorer imitation skills, and higher nonsocial orienting scores at baseline (p<.05). After treatment, both groups showed improvement in DQ and ADOS Social Affect scores with no main effects of group assignment. Both groups of parents showed significant increases in parent-child interaction behaviors, with greater increases in the ESDM group (effect size=.57) than the community-based group (.37). ). Parents in the ESDM group reported significantly less parenting stress at followup (p=.04) but did not report more parenting competence. When examining combined groups, two key effects emerged. First, total intervention hours were associated with reduced restrictive and repetitive behavior and nonsocial orienting and improved DQ and vocabulary comprehension. Second, children younger than 24 months showed greater increases in DQ scores (effect size=-1.20, p=.002).

A fair quality RCT from Canada compared a DIR-based model, MEHRIT (Milton and Ethel Harris Research Initiative Treatment Program), (n=25; mean age=42.52 months, SD=8.76) to community care-as-usual (n=26; mean age=46.38 months, SD=8.29).¹⁰⁷ Data were collected 12 months into an ongoing 24-month treatment course. MEHRIT was administered by trained occupational therapists and speech-language pathologists who worked with participants' parents for two hours per week. Community intervention included no more than 15 hours per week of ABA, speech and occupational therapy, social skills groups, and alternative treatments (mean intervention hours: 3.9 per week). Post-treatment, the MEHRIT group showed significantly more initiation of joint attention (p<.001), involvement in activities (p<.01), and attention to activities (p<.05). They also showed more enjoyment in interaction, but this group difference was also present at baseline (p<.05). Both groups showed significantly improved language skills adjusted for developmental quotients, with no significant between-groups effects (effect sizes of .451 for MEHRIT and .915 for community treatment).

Another fair quality RCT from Asia examined DIR/Floortime (n=15) compared with center-based ABA (n=16).⁹² Groups were stratified based on age (24–47 months, 28–72 months) and ASD severity, based upon CARS scores. Both groups continued to receive treatment-as-usual, including enrollment in preschool programs and community-based services (such as speech or behavioral therapies.) Relative to the center-based group, the DIR/Floortime group showed significant improvement on the Functional Emotional Assessment Scale (p<.05) and ASD symptoms as rated by the CARS (2.9 vs. .08, p<.01). Parents in the DIR/Floortime group also rated their children as showing significant improvements in emotional development (p<.01). A fourth fair quality RCT comparing parent training plus special education preschool to special education preschool alone reported no between-group differences on language development after 12 months of intervention, though language skills within both groups improved over time.⁹⁴

Three poor quality studies, two European prospective cohort studies ^{97, 108} and a crossover RCT from China,⁹⁹ compared parent training to lower intensity supportive interventions. Mean ages ranged from 25.33–33.6 months. Both involved home visits and working with children and parents. A lower intensity treatment model, Autism-1–2–3, compared two groups that received the same series of 10 half-hour child- and parent-training sessions, with one group having a lagged start date and serving as a control. It did not yield group differences on ASD symptoms, language skills, or parent stress scores.⁹⁹ Another lower intensity model, the Barnet Early Autism Model (BEAM), incorporated aspects of ABA, TEACCH, PECS, and other occupational and speech-language interventions. It provided an average of 6.4 hours of home-based intervention per week over ten months. Participants were compared with a care-as-usual group and were not randomly assigned. Relative to the control group, BEAM recipients improved significantly more in adaptive behavior (p<.001) and receptive language (p<.05) but not IQ, with baseline levels of parenting stress negatively related to language and adaptive outcomes. ¹⁰⁸ The Keyhole model incorporated elements of Hanen's More Than Words and the TEACCH programs. It compared 15 to 18 home visits over a 9 month period (n=35) targeting adaptive skills, ASD symptoms, and parent stress to a lower-intensity intervention model (n=26; 5 home visits, no additional services or supports). Compared with the lower-intensity group, children in the Keyhole intervention showed improved adaptive, imitation, and communication skills, based upon parent report. Mothers in in the Keyhole group also reported improved health but not stress.⁹⁷

Studies Addressing Joint Attention Outcomes

A fair quality pilot RCT evaluating a teacher-implemented joint attention intervention randomized child-teacher dyads in public preschools to either intervention (n=9 children, mean child age=46 ±5 months, mean mental age=30.3 ±5.01 months) or wait list control (n=7, mean age=43.01±6 months, mean mental age=33.8±8.74 months).¹⁵⁷ The manualized JASP/ER (Joint Attention and Symbolic Play/Engagement and Regulation) intervention trained teachers in eleven key strategies including setting up the environment, following the child's lead, imitating the child's play action, contingent language, and modeling joint attention. Teachers received a 1-hour training workshop and 1-hour of individual training with the child from a JASP/ER interventionist per week. Interventionists also instructed teachers to use JASP/ER strategies daily. At the 5-week followup, children in the intervention group improved in total initiations of joint attention and in pointing compared with the control arm (p<.005) and in showing (p<.01) in classroom observations (large effect sizes for each measure, 1.85 to 2.02). Groups did not differ on measures of looking or giving. Most scores on the Early Social Communication Scales joint attention measures and frequency of joint attention initiations in videotaped interactions did not differ significantly between groups. Object engagement declined and supported engagement improved in the treatment group compared with control (large effect sizes, d=1.24 to 1.41, p≤.05). Observations of teachers also demonstrated increased use of JASP/ER strategies in the treatment arm.

In another good quality pilot RCT of JASP/ER, investigators randomized minimally verbal (<10 spontaneous functional communicative words) preschoolers enrolled in intensive ABA-based interventions for at least 12 months to either JASP/ER or control (standard intensive preschool, n=8, mean age=54.68±10.25, mean mental age=13.91±3.85).¹⁵¹ Treatment group participants (n=7, mean age=48.73±11.68 months, mean mental age=17.21±3.91 months) received 1-hour of JASP/ER training per week in addition to the intensive preschool. At the 3-month followup, the JASP/ER group increased in play types and decreased time unengaged significantly from baseline (p=.04). Changes were not significant for the control group. The JASP/ER group also spent less time disengaged during class observations compared with the control group (effect size=1.63, p=.05), initiated more requesting gestures (effect size=1.51, p=.01) and evidenced more diversity of spontaneous play (effect size=0.81, p=.04). Groups did not differ on Early Social Communication Scales variables related to joint attention.

Another fair quality RCT¹⁵⁵ conducted in 59 Norwegian preschools over 8 weeks evaluated a manualized adaptation of a joint attention intervention reported below.^{140, 141, 148, 149} Children in the intervention group (n=34, mean age=47.6±8.30 months, DQ=53.3±19.2) attended regular or specialized ASD preschools and also received up to 80 sessions (20 minutes twice daily, 5 days/week) of intervention focused on promoting joint attention and engagement within play activities. Children in the control group (n=27, age=50.3±8.3 months, DQ=59.9±19.7) also attended regular or specialized preschools. Groups did not differ in number of preschool hours or 1:1 training or support. The control group had greater expressive language age at baseline compared with the treatment group (mean 24.9±12.8 vs. 18.8±10.5, p=.047). At the 8-week followup, frequency of joint attention skills during teacher-child play were significantly better in the treatment group compared with control (effect size=0.44) but the duration of joint engagement did not differ between groups. Duration of joint engagement was greater in mother-child play in the treatment group vs. control (mean 12.2% longer duration of joint engagement, effect size=0.67). Although initiation of joint attention skills increased in the treatment group, group differences were not significant, thus effects on joint attention seen with teachers did not generalize. Frequency of joint attention initiation as measured on the Early Social Communication Scales did not differ between groups. Adjusting analyses to account for expressive language differences did not change results. Further, investigators found no putative moderators (age, DQ, language age, program philosophy) to be significant, suggesting that the intervention may be applicable across developmental levels.¹⁵⁵

Another fair quality RCT comparing joint attention and symbolic play interventions delivered via an interventionist included 58 children with ASD between 3 and 4 years of age. Investigators assessed language development, joint attention and play skills, and mother-child interactions at pre- and post-intervention and 6 and 12 months after the end of the 5 to 6 week intervention.^{140, 141, 148, 149} Children in both groups showed significantly greater growth in expressive language, initiation of joint attention, and duration of child-initiated joint attention over time than did participants in the control group (p<.01 to <.05, moderate to large effect sizes). Growth in receptive language was not significantly affected by the intervention from pre-intervention to 12 months post-intervention. Children in the symbolic play group also showed significantly more growth in play level than did children in either the joint attention (p<.01) or control (p<.001) groups.

In a subsequent report on 52 of the 58 RCT participants assessing joint attention quality, both the joint attention and symbolic play groups improved in shared positive affect during joint attention and in shared positive affect with utterances during joint attention at 6 and 12 months post-intervention (p<.05) but not at intervention exit.¹⁴⁹ Differences between groups at the 6 and 12 month time points were not significant. The control group generally declined in instances of shared affect over the followup time points. Forty of the 58 participants in the RCT also participated in followup 5 years post-intervention.¹⁴⁸ Fifteen of 20 children in the joint attention group, 14 of 21 in the symbolic play group, and 11 of 17 in the control returned at 5 years; mean age across groups was 8 years and 8 months. Of the 40 participants, five were enrolled in regular education, 17 in regular education with some special education support, and 18 were in special education classrooms; placement did not differ among groups. At followup, 5/15 participants in the joint attention group, 1/14 in the symbolic play group, and 2/11 in the control arm were considered non-spectrum. Thirty-two of the 40 participants achieved valid scores on language assessments at followup. Ability to use spoken language at followup (“passing” the language assessments) was predicted by children's average play level at baseline (p<.01). Number of functional play types at baseline predicted greater cognitive skills. Age at baseline, initiation of joint attention, play level and treatment group assignment predicted subsequent vocabulary ability (all p<.03); these factors together explained 64 percent of spoken language variability.

In a fair quality RCT of a joint attention intervention adapted from this study^{140, 141, 148, 149}investigators randomized 38 caregiver/child dyads to either immediate, parent-mediated treatment (n=19) or a wait list control group (n=19).¹⁵⁰ The 8-week treatment included individualized, developmentally appropriate play routines to promote parents' following of their children's interests and activities. Children in both groups ranged in age from 21 to 36 months (mean=30.82 months, mean mental age=19.2 months). At the end of intervention, children in the treatment group demonstrated less object-focused play, more responsiveness to joint attention, more functional play acts, and more joint engagement than children in the control group (p<.05). Groups did not differ in initiations of joint attention, diversity of symbolic play, or unengaged actions. At followup of the treatment group 12 months after the end of intervention, results suggested maintenance of gains in joint engagement, response to joint attention, and reduction of object engagement, but changes in scores were not significant. Types of functional play acts improved in the treatment group at the 12-month followup (p<.01). In analyses of potential predictors of outcome, greater caregiver quality of involvement (rated by investigators) predicted increased joint engagement (p<.05) but not other play skills or engagement outcomes. Parent-rated adherence or competence did not predict changes in any outcome. Number of hours of other intervention similarly did not predict any outcomes.

A fair quality RCT of a classroom-based joint attention or symbolic play intervention based on the manualized approaches in other studies reported above^{140, 141} randomized 14 special education teachers to either a symbolic play followed by a joint attention intervention (n=10 children, mean age=54.50±5.06 months, mean mental age=25.29±15.77 months), the joint attention intervention followed by symbolic play (n children=14, mean age=56.21±10.42, mean mental age=36.25±11 months), or a waitlist control (n children=9, mean age=59.67±10.61, mean mental age=30.38±13.19).¹⁵⁸ Treatment occurred in eight weekly sessions over 8 weeks (4 weeks on either joint attention or symbolic play followed by 4 weeks on the other approach), groups did not differ on play or joint attention behaviors in classroom observations at followup. Children randomized to either treatment arm spent more time in a joint engagement state compared with the control arm (effect size=0.63). In analyses combining the treatment groups, joint engagement time, joint attention responses/minute, joint attention initiations/minute, symbolic play acts/minute, all assessed via classroom observations, increased significantly from baseline to post-intervention (effect sizes of 0.41, 0.43, 0.21, and 0.51 respectively). In investigator-mediated rating of early social communication, the number of joint attention responses increased from baseline (effect size=0.23); children were able to generalize increases in responding to joint attention to a novel individual. Initiation of joint attention did not increase significantly nor did functional play or level of structured play. No potential modifiers (age, ASD severity, mental age) were significantly associated with treatment outcomes.

Another fair quality RCT conducted in Belgium included 36 children (18 in each arm) receiving either standard care in low-intensity rehabilitation centers for children with ASD (focus on communication, social skills, play, and motor skills for 3 to 5 hours/week) or standard care + a joint attention- and imitation-focused intervention delivered for 1 hour/week (two 30-minute sessions for a total of 24 sessions).¹⁵⁹ The joint attention/imitation intervention included games and activities to promote following and initiating requests; gaze following; pointing; initiating joint attention; and gestural, vocal, symbolic, or social imitation. Children ranged in age from 4.07 to 6.92 years, with IQs in the average to mild intellectual disability range (50–105 full scale IQ). After 12 weeks of intervention, the joint attention/imitation group had higher total joint attention scores, improved gaze following, and greater request initiations than the treatment as usual group (effect sizes 0.11 to 0.22, all p values ≤.05). The number of elicited joint attention acts increased from 6.53 to 8.41 and the number of spontaneous declarative joint attention acts increased from .89 to 1.72 for the treatment group from baseline to followup; correct imitations increased from 34.11 to 41.12. Initiating declarative joint attention decreased significantly for both groups from baseline to followup (p<.05). Scores on measures of imitation did not differ between groups, though both groups improved over time. Higher baseline verbal IQ was associated with gains in imitation in the treatment group (p<.05), but no other variables tested (age, mental age, full scale IQ, performance IQ, baseline imitation and joint attention skills) were statistically significant. Children in the treatment group improved equally regardless of age or IQ level.

Finally, a fair quality RCT of a joint attention intervention assessed the effects of a roughly 7-month home-based parent training approach targeting focusing on faces, reciprocal communication/turn-taking, and joint attention compared with community-based treatment as usual.¹⁶⁰ The 11 participants in the experimental arm had a mean age of 24.6±4 months and mean Mullen expressive language score of 24.6±6.7 (control group: mean age=27.5±3.4, mean expressive language=24.8±6.9). Reported weekly hours, including the joint intervention sessions for experimental group participants ranged from 2.98±1.25 to 17.88±9.06. Hours/week ranged from 6.25±6.49 to 21.35±11.51 in the control arm. At followup 4 weeks post-intervention focusing on faces and responding to joint attention were significantly improved in the treatment group compared with control (p<.001); scores for the treatment arm remained significantly improved vs. the control group from the 4 week to the 8 week followup and from baseline to the 8 week followup. The effect size for between group differences at the 8 week followup on the focusing on faces outcome was 0.84 and 1.18 for responding to joint attention. Effect sizes for initiations of joint attention were not significant. Language outcomes were significantly improved for the treatment group compared with control. While both groups improved over time, Mullen receptive language and Vineland communication scores were significantly better in the treatment vs. control arm (p values <.05). Effect sizes for differences at the 4-week followup were 0.59 (Vineland) and 0.34 (Mullen); scores for the 8-week followup were not reported.

Studies Addressing Pretend Play

One poor quality nonrandomized, crossover study conducted in a private preschool included 12 high functioning children with ASD (age range 55–75 months).¹⁶¹ Intervention group participants received the Picture Me Playing intervention, which included scripted stories built around specific toys to model and encourage pretend play. Instances of play dialogue increased significantly following intervention for the treatment group compared with control (3.6 times more utterances over baseline vs. 1.79 times, p<.05), though frequency of play utterances in both groups improved from baseline. Gains in pretend play for both groups also generalized to a toy not used in the intervention and without scripted utterances.

Studies Addressing Imitation

A good quality pilot RCT of Reciprocal Imitation Training, which uses naturalistic approaches to promote imitation and social interaction, allocated 27 children to either Reciprocal Imitation Training (n=14, mean age=39.3±7.3 months, mental age=20.8±6.6) for 3 hours/week for 10 weeks or control/treatment as usual (n=13, mean age=36.5±8.00, mental age=17.9±7.5).^{152, 153} The interventionist-led imitation training included modeling of play and gestures and contingent imitation of children's responses and actions with toys. Children in both arms continued to receive between .25 and 25.5 hours of additional intervention per week. Data for 21 of the children was also reported in an earlier pilot,¹⁵³ which reported gains in imitation for the treatment group compared with control (p<.05). Gains in imitation were associated with the number of spontaneous play acts at baseline. In the followup RCT,¹⁵² the intervention group made more joint attention initiations compared with control (p<.05). Intervention participants also improved on the Social-Emotional Scale compared with the control arm (p=.02). Changes in imitation were not shown to be associated with gains in social functioning.

Studies Addressing Parent/Child Communication

In a fair quality randomized trial of a focused play intervention, investigators allocated children to either the play intervention (n=36, mean age=58.3±12.7 months) or a control group (n=34, mean age=55.9±11.9 months).¹⁵⁶ Parents of children in the treatment and the control groups could participate in a parent education program focused on advocacy for their children. Parents in the treatment group also participated in a manualized play time intervention, which used home-based sessions (90 minutes/week for 12 weeks) to promote parental engagement and encouragement of child communication. Children in both groups continued to receive a mean of 14 hours (± 5–8 hours) of school programming and individual services such as ABA-based approaches for a mean of 12±10 to 12 hours/week during the treatment phase. Children also received a mean of >12 hours of school or individual services during the 12-month followup period.

In analyses at the end of intervention, maternal synchronization (maternal direction of child attention or utterances in line with toys/actions in which child was already engaged vs. redirecting or not synchronized with child's actions) was significantly greater in the treatment group compared with control (effect size=0.08, p<.05). Maternal synchronization was moderated by baseline maternal insightfulness (p<.05) and synchronization was greater in those mothers rated as insightful compared with non-insightful (effect size=0.31, p<.05). Expressive language scores did not differ between groups at the end of intervention or at followup 12 months post-intervention (effect size for baseline to followup change=0.03, p=ns). Children with baseline expressive language abilities below 11.3 months showed greater gains in language in the intervention group vs. control (effect size=0.25 for 24 children with low language skills). The link between short-term gain in maternal synchronization and long-term language (12 months post-treatment) gains was not moderated by maternal insightfulness, nor did initial language skills moderate the link between gains in maternal synchronization after 12 weeks and long term gains in expressive language.¹⁵⁶

Another fair quality RCT included 14 participants (age range 28 to 68 months, mean 41.14) randomized to either an adapted More Than Words curriculum focused on teaching parents to understand child communication and promote verbal responsiveness or to a waiting list.¹⁵⁴ Treatment group parents received approximately 12 hours of training and 14 small-group parent-child coaching sessions. Overall, children had mean auditory language age of 14.79 months and expressive age of 20.21 months with greater baseline language abilities in the waitlist group compared with the treatment group. At followup, treatment group parents improved significantly compared with the control group in measures of verbal engagement with their children (p values ≤.03). Children in the treatment group increased in prompted communication acts compared with control (p<.03), but spontaneous verbal and nonverbal communication acts did not differ between groups.

Age

As in the 2011 review, several studies reported associations between age at intake and improved outcomes. In one RCT of an approach incorporating parent training, younger age was associated with greater improvements: greater language gains were seen in children who were younger with lower functioning levels at baseline.^{72, 105}

Age effects were not consistent, however, and may reflect characteristics of subgroups and treatment characteristics that need further elucidation. For example, one study comparing preschool-delivered intensive early intervention and treatment as usual reported larger adaptive behavior gains for older children in the early intervention group.⁷⁷ Another RCT compared early intensive treatment delivered by parents and by specialized center staff with eclectic treatment and identified predictors of progress: in the parent training group, older children achieved better adaptive behavior outcomes; younger children made more gains in early language comprehension and production.

In a retrospective cohort study of a community-based early intervention program, outcomes were related to age at enrollment, treatment duration, and higher baseline adaptive scores. A significant interaction emerged between age at enrollment and group membership, with younger starting age influencing outcomes for the treatment group but not the waitlist control.⁸⁶ In contrast to the early intervention studies, in an RCT assessing emotion recognition, older age was correlated with improved identification of fear expressions, affect recognition, and the mind reading desire based task.¹²³ Another RCT of a preschool-based joint attention intervention compared an 8-week treatment program focused on increasing initiating, giving, and sharing joint attention skills plus preschool to preschool alone in 61 children with ASD.¹⁵⁵ In exploratory analyses, investigators found no putative moderators (age, developmental quotient, language age, program philosophy) to be significant, suggesting that the intervention may be applicable across developmental levels.

IQ/Cognition

Associations of outcome and IQ or measures of cognition were mixed across studies. Intervention efficacy was associated with baseline cognitive scores in one early intervention study comparing preschool models classrooms,⁸³ with higher baseline cognitive scores associated with less improvement in children in TEACCH model classrooms. In an early intervention prospective cohort study, baseline IQ was positively correlated with socialization, communication, daily living, and composite score gains on the Vineland in the treatment group; however, baseline IQ did not correlate with IQ at followup.⁷⁷ In another early intervention study, higher pre-treatment mental development state and early language skills predicted better outcome on parent-reported adaptive behaviors in the eclectic treatment group.^{101, 102} In a study assessing emotion recognition, higher verbal IQ was associated with some short term improvements in fear recognition and mind reading tasks,¹²³ while in another emotion recognition RCT, IQ was not correlated with improved outcomes in either the treatment or control groups.¹³² In another RCT of a group-based social skills approach, IQ was not associated with response status;¹²⁴ similarly, treatment gains were not associated with IQ in an RCT comparing parent training plus risperidone to risperidone alone.^{167, 179-181} In one study of CBT focused on executive function outcomes, higher baseline scores predicted greater improvements in flexible thinking, social tasks, parent- and teacher-rated executive function shift and planning/organization measure, parent-rated Social Responsiveness Scale total score (p values <0.05). Higher IQ predicted greater improvements in flexible thinking and the challenge task plan measure. Younger age predicted greater improvement on the challenge task and parent-rated executive function measures of shift and planning/organization (p<0.05). Female sex predicted greater improvement on the parent –rated Social Responsiveness Scale total score (p values <0.05). ¹⁸³ In a play-interaction study targeting imitation and joint attention, higher baseline verbal IQ was associated with gains in imitation in the treatment group (p<.05), but no other variables tested (age, mental age, full scale IQ, performance IQ, baseline imitation and joint attention skills) were statistically significant. Children in the treatment group improved equally regardless of age or IQ level.¹⁵⁹

ASD Severity/Symptom Severity and Diagnoses

In some studies, children with lower symptom severity or less severe diagnoses improved more than participants with greater impairments. In an RCT assessing ABA-based early intervention, lower baseline ASD severity was associated with parent-reported cognitive and adaptive growth for children who received eclectic vs. ABA intervention, but not with improvements in standardized cognitive test scores.^{81, 82} A prospective cohort study of preschool-based early intensive intervention reported that children in the early intervention group with PDD-NOS or Asperger diagnoses (but not autism) had greater gains in overall adaptive behavior, communication, and daily living skills.⁷⁷ A prospective cohort study comparing four early intervention approaches (home-based 1:1 ABA intervention, low intensity home-based programming for children with special needs [portage], home-based, local health authority-developed intervention incorporating parent training, and special education nursery/preschool) evaluated relationships between ASD severity, time in intervention, and effectiveness of intervention.⁹⁵ Hours of intervention ranged from 2 to 40 across groups, with the home-based ABA group receiving the most (mean 30.4/week) and the Portage group the least (mean 8.5/week). Baseline ASD severity and total intervention hours modified effects of treatment significantly. First, baseline ASD severity was inversely related to composite change scores for all but the home-based ABA group and was positively related in that group. That is, children with more severe ASD symptoms made more progress in ABA and less in the other intervention groups. Second, more intervention time was negatively related to composite change scores for children in ABA but not in the other groups. More hours of ABA were associated with less progress relative to more hours of school enrollment or other home-based interventions.

Two reports^{88, 90} including participants in a retrospective cohort study evaluating an early intervention approach⁸⁶ assessed potential outcome predictors including baseline age, Vineland scores, IQ, and ASD severity (CARS). Younger age at intake, higher initial developmental levels⁹⁰ and treatment intensity^{88, 90} were related to better treatment outcomes. Vineland standard scores and IQ and mental age were higher for the 32 children whose followup standard scores on cognitive and/or adaptive behavior were in the low average range or better (>85) and whose CARS scores were in or very close to the non-ASD range (<30), Similarly, these “average outcome” children had significantly lower intake CARS severity scores, began intervention earlier (mean 42 months vs. 55 for rest of sample), and received intervention for a longer duration. More of these children also had diagnoses of PDD-NOS. Children who had poor outcomes at followup (n=75) had statistically significantly lower baseline IQ , mental age, rate of development, and Vineland scores (except for the socialization domain), with p values ranging from .01 to <.001. Differences likely were not clinically significant, however, and diagnostic category, severity, age at entry, and duration of therapy were not significantly different in the poor outcome group compared with the rest of the sample.

In an RCT evaluating an emotion recognition intervention, long term improvements in identification of happiness expressions were associated with greater ADOS severity, as was matching of emotions overall and of sadness specifically.¹²³ In an RCT of a theory of mind training program, children with PDD-NOS improved on most measures of emotion recognition while children with Asperger syndrome improved only in understanding of complex emotions.¹²⁸ In another RCT of a group-based social skills approach, children with Asperger syndrome were more likely to be responders compared with children with PDD-NOS (p=.03).¹²⁴

Finally, an RCT assessing a parent training approach targeting challenging behaviors examined 21 candidate predictors and moderators of outcome scores on the Home Situations Questionnaire (HSQ) and the Aberrant Behavior Checklist, Hyperactivity/Noncompliance (ABC-H) scale.^{167, 179-181} Children received either parent training plus risperidone (n=75, mean age=7.4) or risperidone alone (n=49, mean age=7.5); thus, potential moderation of effect reflects the combination of parent training and risperidone while predictors of effects reflect the impact of risperidone with or without parent training. Investigators examined variables including parent training adherence, age, IQ, family income, maternal education level, parent stress, and child baseline ratings on measures including the Vineland and ABC. Only higher baseline scores on the HSQ (greater noncompliance) predicted greater improvement in either treatment condition (p=.007), with the lower HSQ group demonstrating less mean improvement than those with higher baseline HSQ scores. Though not significant, older children had slightly more improvement than younger children. No variables predicted ABC-H outcomes, though children with higher baseline Vineland composite and communication subscale scores had greater improvement on the ABC-H. While not a significant predictor of outcomes, greater parent adherence to the training program was correlated with better HSQ outcomes (p=.006), but adherence did not correlate with ABC-H scores. No candidate variables were found to moderate the relationship between parent training and HSQ or ABC-H outcomes, which may suggest that parent training is appropriate for the broader range of children with ASD.

Adaptive Behavior

Studies reported mixed findings related to outcomes associated with baseline adaptive behavior. In one retrospective cohort, positive outcomes in both the early intervention and the waitlist control groups were related to higher baseline adaptive scores.⁸⁶ In one early intervention study, initial higher adaptive behaviors predicted better post-treatment early language comprehension.^{101, 102} In an RCT comparing risperidone alone and risperidone plus parent training, treatment gains were not associated with adaptive or maladaptive behaviors.^{167, 179-181}

Language/Communication

The impact of language skills and attention to objects (vs. people) were assessed in three studies. In one RCT of the More Than Words program, the treatment group showed differential effects on child communication depending on children's baseline object interest; children with lower levels of baseline object interest had greater growth in communication skills, whereas children with higher levels of object interest showed attenuated growth.⁹³ In another study of play-focused intervention, children with baseline expressive language abilities below 11.3 months showed greater gains in language in the intervention group vs. control (effect size=0.25 for 24 children with low language skills).¹⁵⁶ In another early intervention study, children who gained more language comprehension had higher adaptive behavior scores pre-treatment. Pre-treatment language comprehension also predicted post-treatment language production.^{101, 102}

An RCT evaluating an imitation-based approach to affect social functioning^{152, 153}assessed whether changes in social functioning were tied to changes in participants' imitation skills. Gains in imitation were associated with the number of spontaneous play acts at baseline; however, changes in imitation were not shown to be associated with gains in social functioning. This finding could be because the study had too few participants (n=27) to detect such an effect.

Other Factors

One RCT compared the effects of a 6-week joint attention or symbolic play intervention with a control arm in participants receiving 30 hours of early intervention; at the 5 year followup, investigators assessed diagnoses and language skills for 40 of the 58 original participants.^{140, 141, 148, 149}Investigators also identified potential predictors of vocabulary and cognitive changes via regression analyses. Potential predictors included child age, sex, maternal education, play levels and types, and joint attention responses. Ability to use spoken language at followup (“passing” the language assessments) was predicted by children's average play level at baseline (p<.01). Number of functional play types at baseline predicted greater cognitive skills. Younger age at baseline, initiation of joint attention, play level and treatment group assignment (either joint attention or symbolic play) predicted subsequent vocabulary ability (all p<.03); these factors together explained 64 percent of spoken language variability. Importantly, this study is limited in that children were often receiving intensive levels of intervention outside of the intervention setting, making impact of prescribed intervention hard to determine.