Effects of Creative Movement Therapies on Social Communication, Behavioral-Affective, Sensorimotor, Cognitive, and Functional Participation Skills of Individuals With Autism Spectrum Disorder: A Systematic Review

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder affecting multiple developmental domains including social communication, behavioral-affective, sensorimotor, and cognitive systems. There is growing evidence for the use of holistic, whole-body, Creative Movement Therapies (CMT) such as music, dance, yoga, theater, and martial arts in addressing the multisystem impairments in ASD. We conducted a comprehensive quantitative and qualitative review of the evidence to date on the effects of CMT on multiple systems in individuals with ASD. The strongest evidence, both in terms of quantity and quality, exists for music and martial arts-based interventions followed by yoga and theater, with very limited research on dance-based approaches. Our review of 72 studies (N = 1,939 participants) across participants with ASD ranging from 3 to 65 years of age suggests that at present there is consistent evidence from high quality studies for small-to-large sized improvements in social communication skills following music and martial arts therapies and medium-to-large improvements in motor and cognitive skills following yoga and martial arts training, with insufficient evidence to date for gains in affective, sensory, and functional participation domains following CMT. Although promising, our review serves as a call for more rigorous high-quality research to assess the multisystem effects of CMT in ASD. Based on the existing literature, we discuss implications of our findings for autism researchers and also provide evidence-based guidelines for clinicians to incorporate CMT approaches in their plan of care for individuals with ASD.

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder affecting multiple developmental domains including social communication, behavioral-affective, sensorimotor, and cognitive systems. There is growing evidence for the use of holistic, whole-body, Creative Movement Therapies (CMT) such as music, dance, yoga, theater, and martial arts in addressing the multisystem impairments in ASD. We conducted a comprehensive quantitative and qualitative review of the evidence to date on the effects of CMT on multiple systems in individuals with ASD. The strongest evidence, both in terms of quantity and quality, exists for music and martial arts-based interventions followed by yoga and theater, with very limited research on dance-based approaches. Our review of 72 studies (N = 1,939 participants) across participants with ASD ranging from 3 to 65 years of age suggests that at present there is consistent evidence from high quality studies for small-to-large sized improvements in social communication skills following music and martial arts therapies and medium-to-large improvements in motor and cognitive skills following yoga and martial arts training, with insufficient evidence to date for gains in affective, sensory, and functional participation domains following CMT. Although promising, our review serves as a call for more rigorous high-quality research to assess the multisystem effects of CMT in ASD. Based on the existing literature, we discuss implications of our findings for autism researchers and also provide evidence-based guidelines for clinicians to incorporate CMT approaches in their plan of care for individuals with ASD.

INTRODUCTION
Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that affects multiple domains including the social communication, behavioral-affective, sensorimotor, and cognitive systems. Currently, around 1 in every 54 children in the United States qualify for a diagnosis of ASD (1) and boys are almost four times more likely to be diagnosed with ASD than girls of the same age (2). The hallmark impairments in ASD include poor reciprocal social interactions, difficulties with verbal and non-verbal communication, and restricted and repetitive behaviors and interests (3). For instance, children with ASD have difficulties in responding to social stimuli, sharing their play with peers and caregivers, developing and maintaining relationships, as well as understanding body language, gestures, and facial expressions of others (4)(5)(6)(7)(8). In terms of behavioral-affective impairments, children demonstrate repetitive and stereotyped behaviors such as finger flicking and hand flapping, highly circumscribed and restricted interests, insist on sameness relative to daily routines/schedules, demonstrate extreme distress to small changes in daily routines, and difficulties with transitions between activities (2,9). Moreover, children may also have sensory symptoms including hypo-and hyper-sensitivity to sensory input and unusual responses to sensory stimuli in multiple domains including auditory, tactile-proprioceptive, vestibular, olfactory, and visual senses (9)(10)(11)(12). In addition, children may also demonstrate disruptive behaviors such as aggression, tantrums, defiance, and self-injurious behaviors, as well as increased levels of negative affect (10,11,13,14). Moreover, children with ASD also demonstrate cognitive difficulties such as attentional deficits, impaired decisionmaking, and impaired executive functioning (i.e., working memory, cognitive flexibility, self-control, generativity, and planning), with deficits being more pronounced during open-ended compared to structured settings (15)(16)(17).
Besides the diagnostic symptoms, children with ASD also exhibit a variety of other impairments within the sensorimotor domain that may lead to significant challenges in their activities of daily living (18)(19)(20)(21)(22)(23)(24)(25)(26). Although the exact prevalence estimates of motor impairments in ASD vary widely across studies from around 35% to over 85%, there is a growing consensus that children diagnosed with ASD exhibit motor impairments in gross and fine motor skills (e.g., bilateral coordination, gait and postural stability, handwriting, manual dexterity skills, and visuomotor control), as well as socially-embedded motor skills, including imitation, praxis (performance of skilled functional movement sequences/gestures), and interpersonal synchrony (ability to synchronize movements with those of another person) (19,22,(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38). Several studies have documented the association between motor impairments and severity of core autism symptoms in social communication, repetitive behaviors, and cognitive domains (22,(39)(40)(41)(42)(43)(44)(45)(46)(47). Moreover, sensorimotor difficulties could limit children's social participation and affect their activities of daily living including self-care, mobility, and leisure (41,(48)(49)(50)(51). In short, children with ASD have multisystem impairments that need to be addressed through holistic evidencebased interventions (22,24,(52)(53)(54). Current standard interventions for ASD focus primarily on addressing the core social communication and behavioral impairments. Some popular evidence-based approaches include Applied Behavioral Analysis (ABA) (55), Treatment and Education of Autism and related Communication Handicapped Children (TEACHH) (56), Picture Exchange Communication System (PECS) (57), as well as developmental approaches such as Floor time (58), Social Communication, Emotional Regulation and Transactional Support Model (SCERTS) (59), Early Start Denver Model (ESDM) (60), and Pivotal Response Training (PRT) (61,62). ABA-based approaches are considered the gold standard treatment for ASD and use principles of operant conditioning and intensive structured task practice to promote social communication and behavioral skills (55,(63)(64)(65)(66)(67)(68). Similarly, the TEACHH approach uses visual cues to promote learning through picture schedules and also provides guidelines to increase structure and consistency in the environment, supplies-used, and therapists working with children with ASD (56,69). Conventional therapies are usually very structured, adult-driven, and use a more sedentary approach (67,70,71). On the other hand, developmental approaches facilitate ageappropriate developmental skills such as joint attention, play, and imitation using child-preferred, play-based therapeutic activities within naturalistic settings (67,70,72). However, interestingly, both conventional and developmental approaches do not focus on addressing the sensorimotor impairments that are clearly highly prevalent in ASD (22,46). This highlights a dire need to expand therapeutic interventions to address not just the core impairments but also the multiple co-morbidities in ASD.
Over the past several years, there has been a growing interest in exploring the effects of novel, alternative and integrated behavioral treatment approaches in addressing the multisystem impairments in ASD (27,(73)(74)(75)(76)(77)(78)(79)(80)(81)(82)(83)(84). These holistic, whole-body movement-based, multisystem treatment approaches include but are not limited to structured physical activity, music therapies, yoga, martial arts, dance, and theater-based interventions (53,73,(85)(86)(87)(88)(89). For the purpose of this review, we use the term "Creative Movement Therapy (CMT)" as an umbrella term that encompasses alternative behavioral interventions including music, dance, yoga, martial arts, and theater. The rationale for grouping these interventions together is that all these approaches use movement to integrate the social, emotional, cognitive, and physical aspects of the individual. Approaches involving CMT differ from conventional ASD interventions in that they are based in whole-body movement and promote self-expression (e.g., theater), creativity (e.g., innovative ways of moving body and using props in dance and theater), and improvisation (e.g., music making using instruments, moving to the rhythm of music). These interventions typically encourage child-led activities, playful exploration, and are therefore inherently more enjoyable and motivating for children with ASD (53,90). From a theoretical perspective, CMT approaches are grounded in the ecological Dynamical Systems Theory (DST) (91,92) and the Shared Affective Motion Experience (SAME) theory (93). The DST emphasizes that basic perception-action cycles of bodily movement form the basis for higher-order social communication and cognitive skills (82). Similarly, the SAME theory suggests that music-and movement-based experiences are multimodal in nature and activate similar "mirror" networks in the brain of participants, thereby forming the basis for social, emotional, and motoric connectedness between them (83). This is especially crucial for individuals with ASD given their deficits in multimodal integration stemming from long-distance brain under-connectivity (94)(95)(96).
In addition, given their very nature, CMT interventions are known to have multisystem effects on the sensorimotor domains as well as on the social communication, cognitive/attentional, and behavioral/affective domains in individuals with ASD. For instance, practicing simple and complex movement sequences during choreographed dance routines provides opportunities to promote rhythmic synchronization, multi-limb coordination, balance, gait, and postural control in participants (85). On the other hand, music-based group activities provide a medium for children with ASD to connect with social partners, improve communication abilities, and lead to greater positive affect/engagement (4,6,53,93,(97)(98)(99)(100)(101)(102)(103). Similarly, short bouts of exercise that incorporate self-discipline, goal-oriented behavior, multistep action sequences, and sustained focus, as seen with any martial arts-based techniques, could enhance cognitive abilities such as executive functioning in children with ASD (104).
Although the preliminary evidence is promising, currently, it is unclear if CMT approaches can be considered as evidencebased interventions in ASD. Therefore, this review aims to synthesize the literature to date on the effects of CMT on social communication, behavioral-affective, cognitive, sensorimotor, and functional/participation skills of individuals with ASD across the lifespan (note that for the purpose of the review, we excluded studies that focused on structured physical activity, animalassisted therapies, or technology-based interventions given the clear differences in the key intervention components of CMT approaches compared to the above-mentioned approaches). A few previous reviews have assessed the effects of CMT in children with ASD (75,76,78,105). However, most of them have been restricted to examining the effects of a single type of CMT in individuals with ASD. It would be crucial to compile information on different CMT approaches to compare and contrast the differential effects of these approaches on multiple systems in ASD. Moreover, except a couple of reviews by Zou et al. and Geretsegger et al., none of the other reviews conducted a risk of bias analysis for the reviewed studies or calculated effect size (ES) estimates based on data reported in the reviewed literature (75,105). Assessing methodological quality of studies through a risk of bias analysis enables researchers to estimate the level of confidence in study findings and guides interpretation of study results. Similarly, ES estimates from individual studies indicate the magnitude of the treatment effect and are thus crucial to evaluate the clinical utility of specific treatment approaches. We address these gaps in the literature by providing a comprehensive review of empirical reports studying the effects of CMT approaches through August 2021 in children with ASD. Specifically, we (i) summarize the narrative literature and compare the efficacy of different types of CMT in addressing multisystem impairments in individuals with ASD, and (ii) provide quantitative ES estimates for outcome measures addressed using CMT approaches to objectively evaluate the clinical importance of CMT for individuals with ASD.

Search Protocol
We reviewed literature from four different electronic databases related to allied health, psychology, physical therapy/kinesiology, and education, namely, PubMed , PsycINFO (1969PsycINFO ( -2021, Scopus (1966Scopus ( -2021, and CINAHL . The combination of key terms used included, (a) "music, " "dance, " "yoga, " and "play, " (b) "intervention, " "therapy, " and (c) "autism" (please see Appendix 1 for details of search strategy). We also conducted additional hand searches of reference sections of included studies and previous review papers to identify missed literature through August 2021.

Eligibility Criteria
We included studies published in peer-reviewed journals that assessed the effects of creative movement and play-based therapies in individuals with ASD using experimental or quasiexperimental, longitudinal study designs. Studies were excluded based on the following criteria: (a) only included individuals with other developmental disabilities such as Cerebral Palsy, Down's Syndrome, Attention Deficit Hyperactivity Disorder, Intellectual disability, Spina Bifida, Dyslexia, Learning Disability, etc. [note that studies (N = 5) that recruited mixed samples i.e., individuals with ASD and individuals with other developmental diagnoses were included since we wanted the review to be comprehensive and inclusive of all studies that recruited samples of individuals with ASD], (b) review papers, case-studies, qualitative studies, purely narrative reports, observational studies or reports describing the protocol for a future study, (c) interventions directed solely toward parents/primary caregivers of individuals with ASD, (d) studies that used structured physical activity, animal-assisted therapies, or technology-based interventions in ASD, (e) reports in languages other than English, and (f) gray literature including theses and dissertations.

Data Extraction and Evaluation
After screening 2,643 articles using our eligibility criteria [PubMed (1,354), PsycINFO (821), Scopus (267), and CINHAL (201)] and removing duplicates, 72 articles qualified for our review. Two trained research assistants and the last author screened titles and abstracts of the 2,643 articles based on our eligibility criteria. When necessary, full texts of articles were reviewed to assess eligibility of the study (see Figure 1 for details of search process). All three coders agreed in their ratings for 90% of studies. Disagreements between coders for study inclusion were resolved through discussions and consensus scoring.

Risk of Bias Assessment
We employed the Physiotherapy Evidence Database (PEDro) scale and the NIH quality assessment tool (106,107) to assess risk of bias in reviewed studies. The PEDro scale was used to evaluate the internal and external validity of randomized controlled trials (RCTs) and controlled clinical trials (CCT's) included within our review. The PEDro has a total of 11 items which are scored on a dichotomous scale (No = 0, Yes = 1) of which 10 items are scored for each RCT/CCT to obtain a study score out of a maximum possible score of 10 (first item on the PEDro is not included in the total score) (106). Studies with a PEDro score ≥ 6 are classified as having low risk of bias. For single group pre-post designs, we used the NIH quality assessment tool to assess risk of bias (107). The NIH tool comprises 12 items that are scored on a dichotomous scale (No = 0, 1 = Yes) to assess internal validity of reviewed studies. Questions 6, 7, 9, and 10 include multiple questions per item. For these questions, if studies satisfied all criteria listed in the item, we gave them full points (score of 1). However, if studies satisfied some but not all criteria, a partial score of 0.5 was awarded for the item. The original tool recommends raters to categorize studies based on their risk of bias into categories of "good, " "fair, " and "poor" with studies rated as "good" having low risk of bias and studies rated as "poor" having high risk of bias (106). We classified studies with total scores ≥ 9 as "good, " studies with total scores ≤ 6 as "poor, " and all other studies as having "fair" quality. In addition to the above-mentioned tools, we also used the Levels of evidence as outlined by Sackett et al. (108) to classify all the 72 studies. This grading, based on study design, ranges from Levels I-V. We only included studies from Levels I up to III in our review. Level I is the highest level of evidence and includes systematic reviews, meta-analyses, and RCT's with a PEDro score of ≥ 6, Level II includes RCT's with a PEDro score < 6 and all CCT's, whereas Level III includes single group before-after (pre-post) study designs.

Study Coding Procedures
We coded each study in the review for sample and study characteristics, methodological quality, intervention characteristics (FITT: Frequency, Intensity, Time, Type), assessments used, dependent variables, and treatment effects (see Appendix 2 for coding details). In addition to a narrative description of studies, we also report on quantitative ES from reviewed studies along with their confidence intervals to obtain estimates of the true magnitude of treatment effects Frontiers in Psychiatry | www.frontiersin.org following CMT in individuals with ASD. For parametric data, when adequate data were provided in the original report, we calculated ES i.e., standardized mean difference (d) values (109)(110)(111). For papers that reported non-parametric statistics, ES were calculated using U-and z-statistics (112). In studies where the original report did not provide estimates of central tendency and variability of measured outcomes, we calculated ES using parameter estimates (F-and t-values) and p-values. We acknowledge that these estimates are more inaccurate compared to ES estimates calculated using measures of central tendency and spread within the sample (see Tables 4A-C for details); however, we wanted to provide readers with ballpark estimates of ES. We classified ES according to Cohen's conventions as small (0.1-0.3), medium (0.3-0.49), or large (0.5 and above) (113). We also report 95% confidence intervals (CI) around ES estimates to identify robust, statistically significant effects of CMT in ASD (114,115). Specifically, if a CI does not include 0, it implies a truly significant non-zero treatment effect at the 5% significance level. For the purpose of reliability, all authors as well as 2 undergraduate students coded a subset of the 72 studies using a detailed coding form. Intra-rater reliability of over 99% and inter-rater reliability of over 90% were achieved through consensus coding on scores that coders disagreed on. Following reliability, rest of the papers were divided and coded by the first and last authors.

Description of Studies
All 72 studies reviewed were published between 1994 and 2021 although only 25% of the studies specifically mentioned the year of data collection in the published report. Out of the 72 studies that we reviewed, 25 used music therapy approaches, 11 studies employed yoga-based interventions, 16 studies assessed the efficacy of martial arts-based interventions, 12 studies employed theater-based interventions, 7 studies assessed the effects of dance, and lastly, 1 study employed a combination of music and dance therapies. Of these studies, 30 were conducted in the US, 8 studies were from Iran, 6 from India, 4 from Germany, 3 each from UK and South Korea, 2 each from Hong Kong, Italy, Australia, and Brazil, 1 study each conducted in the Netherlands, Portugal, Greece, Spain, Portugal and Spain, France and Canada, and finally three studies that were subsets of the same larger study (98,116,117) (4,6,27,86,101,104,(118)(119)(120)(121)(122)(123). Furthermore, 3 music therapy papers (98,116,117) reported on samples collected as part of the same, large-scale international study, 3 theater-papers reported on data collected across multiple cohorts by Corbett et al. (124)(125)(126) and 2 more martial arts studies had an overlap in reported samples (127,128).

Sample Characteristics
The 72 studies had a total sample size of 1,939 participants with ASD. Among the studies that did provide gender-related data (total N = 1,573), there were 1,338 males and 235 females. Sixtysix studies were conducted in children between 3 and 21 years, 5 studies included both children and adults, and only 1 study was conducted purely in adults with ASD (see Table 1). Specifically, the ages of participants across CMT approaches were as follows: music (3-38 years), yoga (3-23 years), martial arts (5-17 years), theater (6-21 years), and dance (8-65 years), indicating that within the studies that met our inclusion criteria, music, yoga and dance approaches were the three types of CMT approaches that have been implemented in adults with ASD. All studies provided interventions to individuals with ASD only, except one study that provided training to both individuals with ASD and their caregivers (146). All studies reported that participants did not have prior exposure to CMT.
Sixty-seven studies recruited only individuals with ASD and the remaining 5 studies included children with ASD as well as children with other diagnoses including ADHD, anxiety disorder, learning disability, sensory processing disorder, and emotional and behavioral disorder. Across studies, the diagnosis of ASD was confirmed using multiple measures including standardized tests, physician report and parent-report questionnaires (see Table 1). Specifically, 41 studies employed gold standard measures such as the Autism Diagnostic Observation Schedule (ADOS), Autism Diagnostic Interview-Revised (ADI-R), Gilliam Autism Rating Scale (GARS), and Childhood Autism Rating Scale (CARS) to confirm ASD diagnosis, 19 studies relied on physician diagnosis made using criteria listed in the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classification of Disease (ICD), and 12 studies did not provide details of methods used to confirm participants' ASD diagnosis (see Table 1). In terms of intellectual functioning of participants, only 27 of the 72 studies reported on assessing Intellectual Quotient (IQ) scores using various scales such as Wechsler Abbreviated Scale of Intelligence-2nd edition (WASI-2) or the PsychoEducational Profile (PEP). Overall, only five studies included children with mild intellectual disability in their sample, with the remaining studies including participants without any accompanying intellectual disability (101,117,119,150,170). Although a vast majority of included studies did not report on socioeconomic status, the remaining studies primarily included participants from middle and upper-middle class families. Study sample sizes across the different CMT interventions were as follows: 764 participants in music therapy interventions (455 received experimental intervention and 309 received control interventions), 317 in yoga therapy (184 in experimental group and 133 controls), 326 in martial arts (176 in experimental group and 150 control group participants), 246 participants in theater training (162 received experimental intervention and 84 were in control group; 1 theater study (N = 8) did not provide the distribution of the sample into the intervention groups), 262 participated in dance-based studies (139 received experimental group intervention and 123 received a control intervention), and 16 participated in combined music and dance intervention (8 in experimental and 8 in control group). There was great

Study Characteristics
Out of the 72 studies, 34 studies were RCTs, 17 were CCTs, 18 were pre-post designs, 2 studies were cross-over designs, and 1 study employed a reversal design. We scored the PEDro scale for the 52 clinical trials (16 music, 7 yoga, 13 martial arts, 8 theater, 7 dance, and 1 combined dance and music intervention) and the NIH quality assessment tool for the 20 single group prepost design studies (9 music, 4 yoga, 3 martial arts, 4 theater) reviewed to assess risk of bias (see Tables 2A,B). The clinical trials included in the review employed the following control groups: waitlist control, ABA therapy, routine or enhanced standardof-care, seated play, school-based educational programming, social skills training, physical education training, robotic therapy, equine-assisted therapy, or no intervention. In terms of fidelity of implementation of training procedures, of the total 72 studies, around 25% (N = 18) used and provided details of specific checklists employed to monitor the consistency of treatment implementation, another 22% (N = 16) provided brief details of some form of fidelity checks, and 51% (N = 38) of studies did not provide any information on intervention fidelity. In terms of intervention implementation, music therapy and yoga therapybased studies were almost equally split between an individualized vs. group-based format (Music: 13 out of 25 studies and yoga: 5 out of 11 studies provided individualized intervention), whereas martial arts (12 out of 16 studies), theater (10 out of 12 studies), dance (6 out of 7 studies), and combination-based approaches frequently employed group-based implementation with group sizes varying between 3 and 12 participants (see Table 1).

Controlled Intervention Studies
Out of the 52 clinical trials, 50% studies (N = 26) had a high risk of bias (PEDro scores < 6). No study satisfied all the 11 criteria (see Table 2A). Among factors contributing to risk of bias, few studies concealed allocation of subjects to intervention groups (N = 8), and blinding of subjects (N = 1), therapists (N = 9), and assessors (N = 20) were ensured to a varying extent by reviewed studies. Although not as frequent, other factors associated with risk of bias included random subject allocation (N = 36 satisfied the criterion) and baseline similarity of groups on key prognostic measures (N = 33 satisfied the criterion) (see Table 2A).

Single Group Pre-post Designs
Based on the NIH quality assessment tool rating used for assessing the 20 pre-post designs, 1 study was rated as "poor" indicating high risk of bias, 12 studies were "fair" indicating moderate risk of bias, and 7 studies were rated as "good" indicating low risk of bias. Specifically, none of the studies measured outcomes multiple times at pretest and posttest to get stable estimates of child performance, and all but three studies did not discuss power analyses to justify the choice of sample sizes. Another area of concern included blinding of assessors which was ensured in only 3 studies. Finally, 50% studies (N = 10) did not report on validity and reliability of assessed outcome measures, with 30% of the remaining studies (N = 6) reporting on only one but not both these measures (see Table 2B).

Intervention Characteristics
The mean duration of studies that provided music-based interventions was 12 weeks (SD = 9.73, Range = 4-52 weeks), with a mean frequency of around 2 sessions/week (SD = 1.3, Range = 2-5 times), and each session lasting around 40 min (SD = 14.9, Range = 30-90 min, see Table 1 for details). Studies that provided yoga therapy had the longest mean intervention In terms of intervention providers, most of the CMT approaches were provided by either licensed clinicians or specialized instructors trained in the CMT approach (N = 62). Fifteen studies (6 music, 3 yoga, 2 martial arts, 3 theater, 1 music and dance combined) asked teachers, caregivers, support staff, models, students or peers, etc. to assist in the intervention delivery process (see Table 1). Only 2 theater studies by Corbett et al. had teaching staff and peers deliver the intervention independently of clinicians after conducting a 2-day intensive training seminar (87,165). Several papers mentioned using conventional ASD treatment strategies while providing CMT interventions to children with ASD. Common training strategies were based on principles of conventional ASD treatments such as ABA, TEACHH, and PECS and specifically included the use of picture schedules/visual cues, incremental prompting (verbal, gestural, modeling, handon-hand assistance), reinforcement schedules, structured and predictable training routines, motivational strategies, activities designed keeping in mind the participant's sensory needs, and the use of non-competitive, goal-directed, and child-led activities to ensure child compliance. Although a total of 15 studies (2 music, 2 yoga, 5 martial arts, 4 theater, 1 dance, 1 music and dance) mentioned progression in training across intervention weeks,    (104)

Gray shaded cells indicate that the criterion has been fully satisfied (receives a score of 1 point for that item), cells with diagonal stripes indicate that the criterion was partially satisfied (receives a score of 0.5) and blank cells indicate that the criterion has not been satisfied (receives a score of 0 points for that item). Total scores are calculated by summing the number of gray shaded cells (except criterion 1) for each individual study. ROB, Risk of bias; H, High risk of bias; M, Moderate risk of bias; L, Low risk of bias.
Risk of bias rating based on scoring on ratings on NIH quality assessment tool for pre-post designs (i.e., Low risk = scores ≥ 9, High risk = scores ≤ 6, Moderate risk = scores of 6.1-8.9).
Frontiers in Psychiatry | www.frontiersin.org only four of these studies (1 music, 1 yoga, 1 martial art, 1 theater) discussed specific principles of treatment progression over the course of the program. The remaining 57 studies provided no information on treatment principles and progression. Common music therapy training approaches evaluated included Improvisational Music Therapy (IMT) and Relational Music Therapy (RMT). Similarly, yoga-based training approaches included Mandala Yoga and Mindfulness training, Relaxation Response-based training, and ABA basedintegrated Yoga training. Common martial art approaches included Kata, Judo, Karate, Tai chi, and Taekwondo. Theater-based studies used programs such as the Social Emotional NeuroScience Endocrinology (SENSE) Theater, Social Competence Intervention Program (SCIP), and Sociodramatic Affective Relational Intervention (SDARI). Lastly, Dance Movement Therapy (DMT) and traditional Greek dance were some of the approaches used in dance-based studies (see Table 1).
In terms of the location of intervention delivery, 5 studies (all music) delivered interventions at the child's home, 13 studies (4 music, 4 yoga, 2 martial arts, 1 theater, 1 dance, 1 music and dance) delivered interventions at the child's school, 30 studies (10 music, 1 yoga, 8 martial arts, 7 theater, 4 dance) provided intervention at other indoor settings such as a community center, YMCA, etc., 3 studies (1 music, 1 martial art, 1 theater) provided intervention either at the child's school or a community center, 2 studies (both yoga-based involving the same sample) conducted their intervention in a calm and open outdoor setting, and 19 of the remaining studies (5 music, 4 yoga, 5 martial arts, 3 theater, 2 dance) did not provide any specific information on where the intervention was provided.

Outcomes Measures and Treatment Effects
Of the 72 studies reviewed, 31 reported within-group changes, 31 reported between-group differences, and 10 studies reported both between-and within-group changes. Studies used a combination of tests and measures including standardized tests, self-/parent-/teacher-reported questionnaires, video coding, and observational measures to assess the impact of CMT on multiple domains including social communication, behavioral-affective, sensorimotor, cognitive, functional skills, and quality of life (see Tables 3A-C, 6). Twenty-eight studies (16 music, 3 yoga, 1 martial arts, 6 theater, 1 dance, 1 combined music and dance) reported on the inter-and intra-reliability of the assessments employed. In terms of reporting treatment effects, 28 studies reported ES for the assessed outcomes (see Tables 4A-C). We were able to use data from the original papers to calculate ES in 64 out of the total 72 studies (i.e., 89% studies). Tables 4A-C provides a comparison between ES we calculated based on data provided in the paper and ES estimates reported in the original paper, and also discusses the level of agreement between both sets of estimates. Table 6 displays our results to indicate the number of studies stratified by CMT approach where calculated ES were statistically significant (i.e., CI did not include 0). Below, we provide a narrative description of the types of measures employed and summarize the salient treatment effects from the studies included in our review across the developmental domains. Please note that each section discusses the results reported in the original papers assessing those domains followed by a summary of the results from our own ES calculations for the specific domain.

Social Communication
A total of 47 studies i.e., 17 Level I (7 music, 1 yoga, 3 martial arts, 4 theater, 2 dance), 15 Level II studies (6 music, 1 yoga, 2 martial arts, 4 theater, 1 dance, 1 music and dance combined) and 15 Level III studies (7 music, 2 yoga, 2 martial arts, 4 theater) assessed changes in social communication skills following CMT (refer to section Risk of Bias Assessment for definition of levels; see Table 6). The social communication outcome measures employed included standardized tests such as the ADOS, CARS, and GARS, observational measures such as the ESCS and JTAT, parent/teacher-report questionnaires such as the ATEC, SRS, VABS, and SSRS, as well as video-based coding measures of joint attention, verbalization, and turn taking (see Tables 3A-C). All except one Level II study (121) reported quantitative data on social communication outcomes following CMT interventions. Using data from original reports, we were able to calculate a total of 91 ES, specifically, 38 ES from 12 out of the total 17 level I studies, 21 ES from 11 out of total 15 level II studies, and 32 ES from 9 out of the total 15 level III studies.
Of the 17 Level I studies, three studies reported no significant changes (1 martial arts, 1 theater, 1 dance) and 14 studies (7 music, 1 yoga, 2 martial arts, 3 theater, 1 dance) reported positive effects with small to large effect sizes (ES: 0.09-4.06) within their original report. Out of the 38 ES we calculated, CI for 14 ES from 9 studies did not include 0 (5 Music, 1 yoga, 2 martial arts, 1 theater). The largest multi-site study in our review that included 364 children from 9 countries was the only study that reported only small improvements on the social affect subdomain of the ADOS (ES: 0.03-0.2) and the social motivation and autistic mannerisms subscales of the SRS (ES: 0.04-0.02) in the experimental group following a 20-week improvisational music therapy program compared to a comparison group that received a standard-of-care intervention; however, these findings were not statistically significant at the between-group level (116).
Among all 15 Level II studies, improvements of medium sizes (ES: 0.22-0.79) in social communication outcomes were reported in the original papers by music (6), yoga (1), martial arts (2), theater (4), dance (1) and combined music and dance (1) interventions, but out of the 21 ES we calculated, only 6 ES from 3 studies (2 martial arts, 1 theater) were statistically significant (CI did not include 0). In terms of the level III studies, only 3 of the 15 studies reported small to large ES on the ADOS and SRS following theater (2) and yoga (1) interventions (see Tables 4B,C). Similarly, despite large mean ES (0.88-3.04) estimates, only 6 out of the 32 ES we calculated from the 3 Level III studies did not include 0 (2 music and 1 theater).
Overall, out of the 91 calculated ES across 47 studies, 26 ES from 15 studies (∼32% studies) were statistically significant (CI did not include 0) and indicated effects that were varying in magnitude from small to large (see Tables 4A-C, 6). Specifically, there is moderately strong evidence for beneficial effects of music (5 Level I and 2 Level III studies) followed by martial arts                      Observed behaviors (2) I, 1 Level II, and 1 Level III study) interventions in promoting social communication outcomes (see Table 6). One salient finding from our ES calculations is that although all 12 theater studies assessed social communication skills, only 3 of these studies (25%) showed significant improvement (CI of ES did not include 0) in social communication despite the heavy emphasis on peer-mediated social skill training in theater-based interventions (see Table 6). Taken altogether, our review suggests that there is moderate evidence from multiple Level I studies for small-to-large-sized improvements in social communication skills following mainly music and martial arts-based approaches.

Behavioral-Affective
Twenty-one papers in our review assessed behavioral outcomes i.e., 8 Level I (2 music, 1 yoga, 1 martial arts, 3 theater, 1 dance), 9 Level II studies (3 music, 2 yoga, 1 martial art, 1 theater, 1 dance, and 1 combined dance and music), and 4 Level III studies (1 yoga, 1 martial arts, and 2 theater). Behavioral states were assessed using standardized scales such as the GARS, Scale for Assessment of Negative Symptoms (SANS), and Pervasive Developmental Disorder Behavioral Inventory-Children (PDDBI-C), video-based coding of on-and off-task behaviors and the amount of redirection required during training sessions, as well as using questionnaire-based measures such as the Child Behavior Checklist (CBCL), Aberrant Behavior Checklist (ABC-C), and Autism Treatment and Evaluation Checklist (ATEC) (see Tables 3A-C). A total of 18 studies reported significant effects of CMT on behavioral skills, with the remaining 3 theater-based studies (87,124,166) reporting nonsignificant effects. Only 6 of the 18 studies that reported positive effects of CMT provided ES estimates in their original report ranging from small to large in magnitude. Furthermore, we were able to calculate a total of 39 ES from 14 studies, specifically, 8 ES from 5 out of the total 8 Level I studies, 20 ES from 6 out of the total 9 Level II studies, and 11 EFs from 3 out of the total 4 Level III studies. Our own calculations based on these papers suggested mostly medium effects for CMT (note that we obtained large ES estimates for 2 studies where F-values were used to calculate ES; however, these measures are more imprecise compared to ES calculations using means and SD/SE values).
As an example of positive intervention effects, Hartshorn et al. (130) reported significant, large improvements in ontask behaviors (ES: 1.28-4.11) from an early to a late training session following their Level II 8-week music and movement intervention compared to a no-intervention control group (130). Similarly, following a Level II, 14-week, Kata martial art intervention, the experimental group showed significant, medium-sized improvements on the stereotypy subscale (ES: 0.47-0.66) of the GARS compared to a control group, with gains retained at 1 month follow-up (122). Hildebrandt et al. conducted a Level II RCT to assess the effectiveness of a 10-week manualized dance movement therapy intervention on negative symptoms using the standardized, clinician-rated SANS scale in 78 individuals with ASD. The authors concluded that although the results did not reach statistical significance at the betweengroup level, they found promising trends for greater symptom reduction (ES: 0.008-0.47) in the experimental group compared to the waitlist control group in overall negative symptoms as well as most subscales of the SANS (177). Overall, out of the 39 total calculated ES across 14 studies, 17 estimates calculated from 7 studies (∼50% studies), specifically, 2 music (Level II), 3 yoga (1 each of Levels I, II, and III), 1 martial art (Level III), and 1 theater-based (Level I) intervention were statistically significant (CI did not include 0). Although the present state of the literature is insufficient to systematically evaluate the differing effects of various types of CMT, there is currently some consistent evidence for medium-sized, positive effects of CMT in reducing behavioral symptoms in individuals with ASD.
A total of 20 studies, i.e., 5 Level I (1 music, 3 theater, 1 dance), 6 Level II (1 music, 1 yoga, 1 theater, 2 dance, 1 music and dance combined), and 9 Level III (3 music, 2 yoga, 4 theater) studies assessed affective outcomes using questionnaires such as the Empathy/Systemizing Quotient (EQ/SQ), Brief Psychiatric Rating Scale (BPRS) Positive and Negative Affect Schedule (PANAS-C), and a computerized test such as the Multifaceted Empathy Test (MET) (see Tables 3A-C). We found that only 4 out of the 20 studies (1 yoga, 2 theater, and 1 dance) reported non-significant effects, with majority of the remaining studies suggesting small to medium-sized improvements. From the studies that reported training-related affective improvements, we were able to calculate a total of 57 ES, with 5 ES calculated from 4 out of the total 5 level I studies, 34 ES obtained from 3 out of the total 6 level II studies, and 18 ES from 4 out of the total 9 level III studies assessing changes in affective states following CMT.
Among the 5 Level I studies that indicated positive effects following CMT (1 music, 1 dance, 3 theater), largest ES estimates were reported by the 2 studies by Corbett et al. that found improvements in affect/emotion recognition and a reduction in anxiety (ES: 0.49-0.97) following a 10-week Social Emotional NeuroScience Endocrinology (SENSE) theater intervention in 8-14-year-old children with ASD (87,165). While the Level I dance study by Koehne et al. (171) found significant improvements in emotion interference and empathy (ES: 0.31) on a computerized Multifaceted Empathy Test (MET) following a 10-week imitation-and synchronization-based group DMT in youth and adults with ASD, the single Level I music-based study by Srinivasan et al. (118) reported only within-group improvements (ES: 0.32-0.43) in levels of negative and interested affect in the group receiving rhythm therapy.
Of the 6 Level II studies (1 music, 1 yoga, 1 theater, 2 dance, and 1 combined music and dance), a majority of the studies reported small to medium positive effects on empathy, emotional synchronicity, joy, and overall psychological well-being inclusive of anxiety, depressed affect, tension, and vitality (ES: 0.31-0.68). For example, Kim et al. used a within-subject comparison crossover design for improvisational music therapy and toy play sessions in 10 children with ASD and found that children showed greater frequencies of joyful events and mirrored emotional synchronicity (ES: 0.47-0.55) with the therapist during music therapy sessions compared to toy play sessions (101). The 9 Level III pre-post designs provided the largest variations in ES estimates ranging from small to large (0.3-1.12) across multiple studies for multiple outcomes related to anxiety, self-esteem, empathy, resiliency, emotion recognition, and enjoyment during sessions. However, out of the total 57 calculated ES across all affective outcomes from 11 studies, only 3 ES-from one Level III music and one Level I theater-based intervention (∼18% studies) had a CI that did not include 0. Moreover, similar to social communication outcomes, although a majority of theater-based studies assessed affective outcomes, our calculations suggest that only 1 study found significant, non-zero effects of the intervention on affective outcomes. Overall, our review suggests that although individual studies concluded small-to-large-sized positive effects, there is at present insufficient evidence supporting the beneficial effects of CMT on affective outcomes in ASD.
Taken altogether, we found only limited evidence from 26% studies (8 out of 31 studies) that were mainly Level II and Level III studies with high risk-of-bias for beneficial effects of yoga, music, martial arts, theater, and dance on behavioral-affective outcomes in ASD (see Table 6).

Sensorimotor
Of the three studies that assessed sensory skills using either questionnaires (Short sensory Profile, Questionnaire of movement therapy) or video coding-based measures, 2 Level II studies reported moderate-sized positive effects following dance and music interventions on children's body awareness and their negative response to touch (77,130), with the 3rd theater-based study (Level III) reporting non-significant effects (124). We could calculate only 2 ES from 2 out of 2 Level II studies of which only 1 ES from a Level II music study did not include 0. Specifically, an 8-week intervention of music-based movement therapy led to improvements in children's negative response to touch (ES: 0.59) during training sessions compared to a waitlist control group (130). Although Koch et al., reported moderate improvements in awareness of body movement (ES: 0.62) after a 7-week long manualized DMT intervention, the CI of the calculated ES included 0 [(77); see Tables 4A-C]. Given the few studies that have assessed effects of CMT on sensory outcomes, at present, there is insufficient evidence to make definitive conclusions on the effects of CMT approaches on sensory outcomes in ASD.
Fifteen studies that assessed motor outcomes, i.e., 5 Level I (1 music, 2 yoga, 2 martial arts), 6 Level II studies (2 music, 1 yoga, 1 martial arts, 1 dance, 1 music and dance combined), and 4 Level III studies (2 music, 1 yoga, 1 martial arts) used standardized tests such as the Bruininks-Oseretsky Test of Motor Performance-2nd Edition (BOT-2), Movement Assessment Battery-2nd Edition (MABC), and the Korperkoordinationstest fur Kinder (KTK), questionnaires such as the ATEC and imitation test battery, as well as observation-based quantitative measures such as posturography, and video-coding for imitation and interpersonal synchrony. Out of the 8 studies that reported ES estimates, five studies (1 music, 1 yoga, 2 martial arts, 1 music and dance combined) reported medium-to-large positive effects of CMT and 3 music-based studies (1 Level I, 1 Level II, and 1 Level III;  see Tables 2A,B) reported non-significant effects on motor skills following intervention (120,138,140). However, even among the clinical trials, 3 studies (27,150,157) reported only withingroup effects suggesting that the positive effects were perhaps not robust enough to attain statistical significance at the betweengroup level. We were able to calculate a total of 18 ES, i.e., 10 ES from 4 out of the total 5 Level I studies, 5 ES from 1 out of the total 6 Level II studies and 3 ES from 2 out of the total 4 Level III studies assessing motor outcomes.
Large effects were obtained from the Level I study by Sarabzadeh et al. (158) on ball skills and balance subscales (ES: 3.14-3.16) assessed using the standardized MABC-2 test after a 6-week Tai Chi Chuan martial arts-based intervention. Similarly, the single Level II dance intervention that assessed motor outcomes reported improvements in the Korperkoordinations test fur Kinder (KTK) test (ES: 0.88-2.63), a measure of neuromuscular coordination including balance and agility, following an 8-week traditional Greek dance program in 5 children with ASD (85). Of the 3 studies that employed quantitative measures to assess motor outcomes, Garcia et al. in a Level III study reported large improvements in moderateto-vigorous physical activity levels (MVPA) (ES: 0.97) using Actigraph GT9X accelerometers following a judo intervention in children with ASD (127). The other Level I and II studies reported improved static and dynamic balance and reduced postural sway during an eyes closed single leg balance task, respectively (ES: 0.5-5.34), following a 10-week Kata and an 8-week Taekwondo intervention in children with ASD, respectively (157,159). Overall, out of the 18 ES calculated from 7 studies, CI for 8 ES calculated from 4 Level I studies (2 yoga, 2 martial arts) and 1 Level II study (music) did not include 0 (∼71% studies). Therefore, there seems to be limited, yet very promising evidence from mainly Level I studies for medium-to-large sized improvements in motor outcomes following martial arts and yoga-based interventions in ASD.
Therefore, altogether across the sensorimotor domain, there is limited evidence from around 28% studies (5 out of 18 studies, i.e., 4 Level I and 1 Level II) which showed improvements in assessed outcomes following predominantly yoga and martial arts interventions (see Table 6).

Cognitive
The 6 Level I studies (2 yoga, 4 martial arts) that assessed cognition used EEG measures to record neural activity, computerized tests such as the Go-No-Go and the Hearts and flowers test, as well as questionnaires such as the ATEC to report medium to large ES for improvements in executive functioning, visual memory, cognitive awareness, and brain activation patterns following CMT (104,148,152,154,155,161) (see Tables 3A,B, 6). For instance, Chan et al. reported large improvements in self-control (ES: 0.84), indicated by a reduction in the # of rule violations during a Tower of London task following a Nei Yang Gong martial arts intervention compared to a control group that received progressive muscle relaxation (154). Similarly, following a 13-week mixed martial arts intervention, Phung and Goldberg reported improvements in accuracy (ES: 0.83-1.01) on the computerized Hearts and Flowers executive functioning test (104). Of the 22 ES we could calculate from the 6 studies, the CI of 14 ES from 4 Level I studies (2 yoga, 2 martial arts) did not include zero (∼66% studies) (see Table 6). We found some disagreements between reported and calculated ES (see Tables 4A-C for details) (104,154); however, our overall assessment suggests limited promising evidence for medium-tolarge sized improvements in cognitive skills following martial arts and yoga-based interventions.

Functional Skills and Quality of Life
Three studies assessed activities of daily living and QOL using self-and family-report questionnaires such as the World Health Organization-Five Well-being Index (WHO-5), the WHO Disability Assessment Scale (WHODAS) and the Functional Independence Measure (FIM). Of these, 1 study each of Level I and II, respectively (1 music and 1 dance + EAT) [(116, 174); see Tables 3A-C, 6] reported non-significant between-group effects on QOL and functional participation, whereas a single Level III study found medium-sized within-group improvements in QOL following a yoga-based intervention (146). Specifically, de Bruin et al., reported medium-sized improvements in QOL in adolescents with ASD measured on the WHO-5 well-being index (ES: 0.55-0.63) following a 9-week mindfulness training intervention (146). Although Souza-Santos et al., reported non-significant between-group differences, they found withingroup improvements in the dance and combined dance and equine-assisted therapy intervention groups on the Functional Independence Measure (FIM) and WHO Disability Assessment Scale (WHODAS) (174). Based on reported data, we could only calculate 6 within-group ES from a single Level II study. Our calculations confirmed the findings from Souza-Santos et al. (174) with within-group ES estimates ranging from 0.64 to 0.73 and their CI not inclusive of 0. However, overall, at present, there is insufficient evidence to indicate any benefits from CMT on functional participation and QOL of individuals with ASD.

Other Domains
Sixteen (4 Level I, 6 Level II, 6 Level III) out of the 72 papers assessed effects of CMT on other domains including, (1) physiological parameters such as sleep, gastrointestinal (GI) problems, heart rate variability, and cortisol levels, (2) trainingspecific skills such as musical abilities and mindful awareness, and (3) cost effectiveness of provided interventions and parentreported adverse effects (see Tables 3A-C, 6). A combination of quantitative measures such as ECG recordings and salivary cortisol levels, standardized tests, and patient/caregiver-report questionnaires were used to assess these miscellaneous outcomes. For instance, Corbett et al. assessed the effects of theaterbased interventions on salivary cortisol levels, a marker for physiological stress, in 3 separate studies, 2 of which were Level III pre-post designs and one was a Level I RCT (87,124,126). While theater interventions had large within-group effects (ES: 0.73-2.55) for reducing cortisol levels in participants during and after the intervention (124,126), the effects were not strong enough to attain significance at the between-group level (87). A single Level II study also reported significant medium-sized effects (ES: 0.3-0.64) on a questionnaire-based assessment of sleep and GI problems in children with ASD following a 90-day yoga training program (147). Two Level III studies that assessed training-specific musical skills reported mostly large effects (ES: 1.13-2.67) on musical vocal behaviors, rhythmic imitation of musical patterns, turn taking within musical contexts, instrument playing, and singing following music therapy sessions (129,131). No statistically significant effects were demonstrated on mindful awareness and heart rate variability following CMT (151). Three studies reported positive trends in qualitative data on parent/teacher and participant satisfaction, feasibility of implementation, and social validity of CMT (116,140,166).
We were able to calculate a total of 66 ES from 7 studies, i.e., 9 ES from 2 out of the total 4 Level I studies, 31 ES from 1 out of the total 5 Level II studies and 26 ES from 4 out of the total 6 Level III studies. Out of these 66 calculated ES, 8 ES from 2 Level I studies (1 yoga, 1 theater) and 12 ES from 4 Level III studies (2 music, 2 theater) suggested mostly medium-to-large-sized effects on assessed outcomes (see Tables 4A-C, 6). Overall, there is preliminary evidence from mostly within-group designs supporting the effectiveness of music therapies in enhancing children's musical skills and theater-based interventions in improving salivary cortisol levels and reducing stress in individuals with ASD.

Short-and Long-Term Effects of CMT
The efficacy and utility of any therapy depends not only on the effects assessed during and immediately following the intervention, but more importantly on the carryover of training effects into real-world settings beyond the duration of the training. Only 17 (7 Level I, 7 Level II, 3 Level III studies) out of the 72 studies assessed the short-and longterm effects of CMT through follow-up (FU) testing that was conducted between 2 weeks and 12 months post-intervention. Of the 17 papers, only 9 (2 music, 1 yoga, 3 martial arts, 3 theater) studies found sustained improvements in outcomes at FU. The music studies (116,140) that compared rock drumming and improvisational music therapy, respectively, with standardized care found improvements in social and motor outcomes at 2 weeks and 12 months FU, respectively (see Table 3A). Three martial arts-based studies from a single group of authors found retention of positive improvements on behavioral and social communication outcomes at 1-month FU (see Table 3B) (86,122,153). Sustained improvements were also seen following the SENSE theater and SDARI interventions in social communication and behavioral domains (see Table 3C) at 1.5-and 2-months post-intervention respectively (163,165). The only study in the review that conducted multiple FU sessions (at 3, 6, and 12 months) assessed the effects of a dramabased intervention on social communication and behavioral outcomes (170). However, the study only reported outcomes at the final FU visit and suggested sustained improvements in autism severity and emotion recognition at 12 months postintervention compared to baseline values (170). Lastly, 1 yogabased intervention study found sustained improvements in executive functioning at 1.5 months post intervention compared to the posttest and baseline measures (152). On the other hand, 8 studies (3 music, 1 yoga, 1, martial arts, 1 theater, 2 dance) found that the immediate beneficial effects of CMT were not sustained at FU (87,134,136,137,146,172,173). Overall, although 9 of the 17 papers claimed sustained beneficial effects of CMT in individuals with ASD, our calculations suggested that only 4 out of the 22 calculated ES (ES: −0.79-0.71) from a Level II yoga and a Level III theater-based intervention study (152,170) were statistically significant (i.e., CI did not include 0). Thus, there is currently insufficient evidence for short-to-long-term sustained benefits following CMT in individuals with ASD.

Summary of Results
Creative movement therapies have been an ongoing topic of study over the past 3 decades. Within CMT approaches, the effects of music-and yoga-based therapies have been studied since the 1990s and 2000s, whereas dance, theater, and martial arts have been studied only more recently over the past decade. Given that this area of study is still in its infancy, there is presently lack of rigorous, definitive evidence supporting the use of CMT approaches within the standard-of-care clinical practice in ASD. There have been a few reviews in the past assessing the individual effectiveness of music, yoga, martial arts, theater, and dance approaches in individuals with ASD. Despite the common underlying theoretical framework and the key intervention ingredients across these different approaches, to date, there has been no umbrella review that has systematically compiled evidence across different types of CMT approaches in individuals with ASD. Our paper addresses this critical gap by providing a comprehensive review of the literature through August 2021, supplemented with a critical risk of bias assessment on different CMT approaches as applied to individuals with ASD. By conducting both a narrative literature synthesis and a quantitative review through calculation of ES estimates of treatment effects, we are able to systematically compare and contrast the efficacy of different types of CMT approaches in individuals with ASD.
Of the total 72 studies, we were able to calculate within-and/or between-group ES estimates for around 89% studies (23 music, 7 yoga, 16 martial arts, 10 theater, and 7 dance, 1 music and dance) of which around 45% studies (N = 29) showed statistically significant, non-zero effects of CMT on assessed outcomes across domains. Specifically, we found evidence for (1) medium-to large-sized improvements in social communication skills from over 30% of the studies (mostly Levels I and II) that assessed these outcomes, (2) medium-sized improvements in the behavioral domain from around 33% studies (mostly Levels II and III) that assessed these skills, (3) medium-to-large improvements in motor outcomes from around 33% studies (mostly Level 1) that assessed movement performance, and (4) medium-to-large improvements in cognitive skills from over 65% (all Level I) of the studies that assessed this domain. In comparison, we found limited evidence to date for the positive impact of CMT on sensory, affective, and functional participation domains. In terms of CMT types, our review suggests that there is presently strongest evidence for the beneficial effects of music-based therapies in promoting social communication skills (5 out of 7 Level I music studies), followed by limited, yet positive evidence for both martial arts and yoga in promoting motor and cognitive skills (2 Level I studies for each approach for each domain), and for martial arts in promoting social communication skills (2 out of 3 Level I studies). Below we summarize the potential mechanisms of change for individual CMT approaches.

Music Therapy Interventions
Our literature search revealed the largest number of studies for music-based interventions compared to all other CMT approaches, with around 40% studies reporting significant improvements (ES: 0.02-4.11) in measured outcomes. Specifically, 35% studies (N = 7) showed improvements of varying sizes in social communication skills (ES: 0.02-4.06) and around 38% (N = 3) studies suggested large improvements in behavioral-affective outcomes (ES: 1. 28-4.11).
Previous literature in the field of music and autism suggests that children with ASD particularly enjoy musical experiences, and in fact have enhanced musical perception skills (53,178). From a brain imaging standpoint, there is substantial evidence that musical practice promotes multimodal integration by activating long range connections that simultaneously engage the auditory, visual, somatosensory, motor, and premotor areas as well as brain networks such as the mirror neuron system that are especially dysfunctional in ASD (52,101,106). Given the considerable overlap between brain substrates underlying speech and music, and the overall structural similarity between music and language, it has also been argued that musical training can in fact lead to enhanced speech processing in individuals with ASD (52). Overall, there is considerable behavioral and neuroimaging evidence to support the mechanisms for beneficial effects of music-based training in ASD (179).
Our review findings of consistent improvements in social communication skills following music therapies is not surprising given that such activities are based in rhythm, melody, and harmony, and involve components of singing, listening, music making with instruments, moving to the beat of the music, and spontaneous improvisation, all of which provide abundant opportunities for practice of social communication skills such as turn taking, joint attention, imitation, and verbal communication (53). Several of the reviewed studies also provided opportunities for flexibility and child-led activities during training that probably fostered children's engagement and led to better outcomes (4,6,27,118,119,135). Although authors also hypothesized that musical training provides a non-intimidating context that may contribute to a reduction in off-task behaviors, stereotypes, and other repetitive behaviors in participants (130,131), there is presently a need for more rigorous, high-quality research to support the use of music-based approaches in improving behavioral impairments in ASD. Moreover, music making using different types of instruments challenges the fine motor and cognitive systems as it typically involves complex, sequential, and precise finger and hand movements that require intricate motor planning and execution (53). Although the current evidence on the effects of music-based interventions on sensorimotor and cognitive skills is scant, this is definitely an area that deserves further attention.
Taken altogether, although music therapy approaches have the strongest evidence among other CMT approaches, there is a clear need for more research to assess the multisystem effects of these interventions on primary and secondary comorbidities in ASD.

Yoga Therapy Interventions
Yoga-based interventions fall under the category of holistic mind-body therapies and are based on principles and techniques of yogic practice that date back several millennia to ancient India. Yoga and mindfulness practice has been postulated to have physical, mental, and spiritual effects (180, 181). Studies included in our review evaluated the effects of yoga practice on behavioral, social communication, and motor skills as well as on physiological parameters, with very few studies assessing effects on affective control, cognitive, and functional skills. We were able to calculate ES estimates from around 55% of all yoga studies; nevertheless, a majority of the studies indicated mediumto-large improvements (ES: 0.43-2.66) in measured outcomes. Yoga-based programs emphasize the practice of breathing control and mindful body awareness that may help individuals with ASD manage their behavioral and mood/affective symptoms (88,145,148,150). Moreover, the practice of static and dynamic postures that focus on improving balance, core muscle strength, flexibility, and body awareness may impact the sensorimotor systems (150) and also have physiological effects on digestion, sleep, and HR variability (88,145,148); see Table 5 for components). Yoga and mindful awareness have also been found to help with attention regulation, memory, and executive control (152,154,155). Additionally, the only study that provided mindfulness training to adolescents with ASD and their caregivers found improved quality of parent-child interactions as evidenced by decreased parental stress and improved behavioral regulation (146). Despite extensive evidence for the wholebody effects of yoga in healthy individuals (182)(183)(184), our review suggests that the use of yoga-based interventions in ASD is presently an under-researched topic that deserves greater systematic investigation.

Martial-Arts Interventions
Around 56% of reviewed martial arts-based studies reported significant small-to-large improvements in measured outcomes (ES: 0.29-5.34), specifically in social communication (N = 4, ES: 1.13-1.15), cognitive (N = 2, ES: 0.42-1.19), and motor (N = 2; ES: 3.14-5.34) domains. Moreover, improvements in the cognitive domain i.e., in executive functioning, which includes working memory, flexible thinking, and inhibitory control are supported by high-quality Level I studies [ (104,154,161); see Table 6]. This is not surprising given the heavy emphasis in martial arts training on discipline, structured practice of multistep action sequences, and movement precision, all of which require focused attention, motor planning, task switching, and working memory (75,104,123,154,159).
The reviewed studies also suggested the potential for martial arts training to impact socialization, behavior, and motor skills. For instance, martial art training led to improved synthesis  and metabolism of neurotransmitters oxytocin, serotonin, and dopamine (75,122). In fact, disturbed metabolism in these very neurotransmitter systems is thought to underlie the social dysfunction and stereotypies commonly seen in individuals with ASD (75,122,153). Similarly, high-energy, dynamic, martial art movement routines are thought to physically resemble stereotypical movements characteristic in ASD, perhaps serving as a functional "substitute" for repetitive behaviors, while still providing the same level of sensory input and reinforcement (75,122). Although there is presently evidence from only 2 Level I studies, it seems plausible that martial arts training may also have effects on the sensorimotor system through practice of poses and action sequences that require good postural control, balance, multi-limb coordination, strength, agility, and optimal processing in the vestibular-and tactile-proprioceptive systems (104,158). Our review of the existing literature suggests that among all CMT approaches, martial arts-based therapies seem to have the strongest evidence at present for promoting multisystem development in social communication, behavior, motor, and cognitive domains.

Theater-Based Interventions
Although a majority of theater-based studies assessed social communication and behavioral-affective outcomes, only around 42% (i.e., 2 Level I, 1 Level 1 and 2 Level III studies) of reviewed studies reported medium-to-large improvements (ES: 0.56-2.55) across these domains as well as a reduction in cortisol levels (ES: 0.73-2.55) following theater training. In fact, 2 of these studies were conducted by the same research group (126,165). The reviewed theater studies typically provided training in a group format emphasizing interactions with peer models, specialized instructors, teachers, and other staff (87). Such a socially embedded and interactive context may provide plenty of opportunities for individuals with ASD to practice critical social communication skills such as joint attention, turn taking, perspective taking, and dialogue delivery, while also learning to recognize and express subtle socio-emotional cues related to facial expressions, voice intonation, and body language. It is therefore surprising to see a lack of statistically significant effects in support of enhanced social communication and behavioralaffective skills following theater training. A salient difference between theater and other CMT approaches is the average session duration, with theater interventions on an average lasting for much longer, i.e., around 2.9 h/session (see Table 5). It remains to be seen if the long duration of intervention sessions impacted abilities of individuals with ASD to sustain engagement during the training program. Overall, despite the highly interactive nature of theater, at present, there is insufficient evidence to support the use of theater-based training to facilitate social communication and behavioral-affective skills in individuals with ASD.

Dance Therapy and Combined Interventions
Although all individual reports (N = 8; 2 Level I, 6 Level II) concluded positive effects of dance-based therapies, our ES calculations from data reported in 6 papers suggested that none of the ES were statistically significant. This was the singular approach where studies recruited participants across the lifespan from 14 to 65 years. The wide age-range might have added to the variability of data, undermining the effects reported in these studies. Dance is an embodied experience incorporating elements of complex coordination, motor planning, and balance that may provide individuals with ASD opportunities to express their emotions through fluid bodily movements and to engage in mirrored practice during group choreography (172,173). Despite the potential for promising effects on multiple systems through the very embodied nature of the experience, the current evidence on dance therapy in ASD is very limited. We call for future research to fully explore the use of dance-based interventions in individuals with ASD. We specifically recommend that future studies assess the effects of dance approaches on participants within a narrower age range.

Implications for Clinical Practice
Based on the studies reviewed, we suggest intervention guidelines for clinicians working with individuals with ASD in terms of assessments and interventions pertaining to CMT approaches.  In terms of assessment measures, we recommend that clinicians use a combination of domain-specific standardized tests, observational measures, parent report questionnaires/interviews, and video coding to assess the impact of CMT approaches on multiple systems (see Tables 3A-C). Ultimately, the choice of assessment tools will depend upon multiple factors including the domains assessed as well as participant characteristics such as age, autism severity, functional level, receptive and expressive communication, and intellectual abilities. From our own experience, we recommend that objective clinician-based assessment tools be supplemented with parent reports to allow assessment of the individual's skills across a variety of structured and naturalistic activities and environments. We also recommend that researchers collect video data of testing and training sessions that can be scored at a later time by unbiased coders, thereby again allowing an evaluation of multiple snapshots of target behaviors across a variety of settings such as the lab, home, school, etc. In terms of treatment, Table 5 provides a summary of our suggested intervention guidelines for the different CMT training approaches in terms of FITT principles (Frequency, Intensity, Time, Type). Note that our guidelines are based on the reviewed literature, specifically, training programs that led to appreciable improvements in assessed treatment outcomes. The choice of CMT approach should ultimately depend on the preferences of the individual with ASD/their family. Clinicians should choose the approach that their client is most excited about and that they are comfortable delivering. Moreover, based on the evidence from this review, we recommend that there is at present, most consistent evidence from high-quality studies with low risk of bias for enhancements in social communication skills following music and martial arts interventions, and in motor and cognitive skills following yoga and martial arts practice. There is need for more systematic research to support the use of theater and dance-based approaches in the plan of care of individuals with ASD.
To administer CMT interventions to their clients with ASD, allied health professionals may need to consult with certified instructors and work collaboratively with them to tailor interventions to their client/family's needs. Moreover, several studies reported using common training strategies derived from conventional ASD treatments such as ABA, PECS, TEACHH, etc. while delivering CMT approaches with individuals with ASD. While structured practice will be an integral part of every CMT-based session, we strongly recommend that clinicians reserve time during sessions for free play and improvisation that will afford individuals with ASD opportunities for creative movement exploration and self-expression. Although there is a need for more rigorous research in this field, our review certainly suggests that CMT approaches involve embodied experiences that engage multiple systems/domains, are fun and engaging, and may provide individuals with ASD a variety of activity options fostering lifelong learning and creative expression.

Implications for Research
Around 75% of the reviewed studies employed between-group designs; however, <50% of the total studies were RCTs which are considered the gold standard for intervention efficacy research. There is a need for greater methodological rigor of clinical trials to reduce risk of bias by ensuring random and concealed assignment of participants to intervention and control groups, blinding of therapists and assessors, ensuring baseline similarity between groups prior to group assignment, and employing intention-to-treat analyses when possible. Since RCTs require significant amount of financial and personnel-related resources and are not always feasible to conduct in clinical settings, several studies in our review used pre-post designs. Our risk of bias assessment for pre-post designs indicated a need for better justification of sample sizes using power analyses, administration of tests at multiple times to obtain stable estimates of the child's behavior at baseline and post-intervention, better assessment and reporting of validity and reliability of selected outcome measures, and blinding of assessors to ensure unbiased estimates of participant performance.
Overall, across all study designs, we recommend that future studies provide more information on sample characteristics within the original report including measures of autism severity, IQ levels, as well as functional skills assessed using parent questionnaires such as the VABS. This is crucial since, the effects of CMT approaches might differ across participants based on these above-mentioned characteristics. Interestingly, a very small proportion of the reviewed studies included participants with intellectual disability and similarly even fewer studies recruited youth and adults with ASD, suggesting a need for more research with these subject demographics. In terms of study quality, future studies should report on steps taken to assess and ensure treatment fidelity during intervention delivery.
Among reviewed CMT approaches, the greatest quantity of evidence is for music-based interventions; there is therefore, a need for more rigorous research on other CMT approaches as well as efforts directed toward replication of the effects of music-based therapies on multiple systems in ASD using large sample size studies. Specifically, our review suggests that yogaand martial arts-based therapies may be promising to promote multisystem development in individuals with ASD. Although reviewed studies assessed a variety of outcomes, the most frequently assessed domains included social communication and behavioral skills. Given the embodied nature of CMT approaches and their proposed mechanism of action on multiple systems, it would be important for future studies to holistically assess other developmental domains including sensorimotor, affective, and cognitive systems that also present as significant challenges for individuals with ASD. Moreover, researchers should go beyond the impairment domain and begin assessing the impact of CMT approaches on function and participation of individuals with ASD. In order to understand the carryover effects of CMT approaches, studies will need to assess treatment effects both in terms of short-term effects i.e., immediately following intervention completion, as well as the long term maintenance of training-related gains at follow-up. Finally, we urge that authors include their data within the original reports to enable calculation of ES estimates for measured outcomes and metaanalytic analyses.

Limitations
Although we used a comprehensive search strategy (see Appendix 1) to identify eligible studies, it is possible that we may have missed relevant research. For the purposes of our review, we only included clinical trials and pre-post study designs with the exclusion of case studies, narrative reports, and other types of qualitative reports. We also excluded conference proceedings and unpublished theses and dissertations from this review. Finally, we limited our review to only articles published in English. Although our original intention was to restrict our review to studies that used CMT approaches in individuals with ASD, there were a few papers (N = 5) in our review that recruited mixed samples, thereby adding to the heterogeneity of the sample. In terms of intervention components, several studies reported using training strategies derived from conventional, evidence based ASD treatments while delivering CMT to participants. Therefore, at present, the literature does not allow us to tease apart the true effects of the key ingredients of the CMT approaches themselves vs. those of the training strategies used in conjunction with the CMT approaches. In a related vein, a majority of the studies did not report on acceptability, implementation feasibility, cost-effectiveness, etc. of interventions. The extent of the literature does not presently allow the development of clinical practice guidelines for the ASD population; instead, the suggested treatment guidelines reported in our paper are based only on the reviewed studies and the dosing parameters that led to gains in measured outcomes in the reviewed studies.

CONCLUSIONS
Our systematic review aimed at providing a comprehensive summary of the literature through August 2021 on the effects of various types of CMT approaches for individuals with ASD. Our search identified a total of 72 articles that used music, yoga, martial arts, theater, and dance-based intervention programs in 1,939 number of individuals with ASD between 3 and 65 years. Our quantitative synthesis of the published literature suggested strong and consistent evidence for smallto-large improvements in social communication skills following music and martial arts training as well as medium-to-large sized improvements in motor and cognitive skills following martial arts and yoga training. Presently, there is limited evidence in support of theater and dance-based approaches as well as the utility of all CMT approaches in improving affective, sensorimotor, and functional participation skills in individuals with ASD. Our review offers future directions for research examining the effects of CMT approaches in ASD as well as provides intervention guidelines for clinicians to incorporate CMT approaches in their plan of care for their clients with ASD.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author. entry into Excel. The authors would like to thank librarian Jill Livingston who was at the University of Connecticut at the time this review was conducted for her help in designing the search strategy and conducting the comprehensive search across multiple databases.