Network Meta-Analysis of Physical Activity Interventions On Cognitive Functions In Children With Autism Spectrum Disorder: A Protocol of Randomized Controlled Trials

Background: Autism spectrum disorder (ASD) symptoms are usually observed by the age of 2 years. However, the mechanism of ASD is still encompassed in a block box and no identied cure exists. Based on accumulating evidence, intensive early treatment such as physical activity or exercise can make a signicant difference in the cognitive control and development in children with ASD. This study aims to update the knowledge on extant literature and explore the ecacy of physical activity intervention strategies (PAIS) on cognitive functions in children with ASD. Methods: A systematic review and network meta-analysis will be conducted following the preferred reporting items for systematic review and meta-analysis protocols for Network Meta-Analyses (PRISMA-NMA). Nine bibliographic databases (APA PsycInfo, Cochrane Central Register of Controlled Trials, Dimensions, ERIC, MEDLINE Complete, PubMed, Scopus, SPORTDiscus, Web of Science) will be systematically searched to screen eligible articles based on a series of inclusion and exclusion criteria. A study will be considered for inclusion if the study: (a) is not classied as a systematic review with or without meta-analysis; (b) is published from inception to date; (c) includes children aged 0-12 years with ASD; (d) quantitively measures cognitive outcomes; (e) treatment includes at least one PAIS. The internal validity and quality of evidence will be evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. Statistical analyses will be produced in RStudio 3.6 with the BUGSnet package and Comprehensive Meta-Analysis 3.3. Discussion: This study will provide an updated review of the extant literature by using an appropriate network meta-analytic model and address the questions regarding ecacy of PAIS that signicantly impact cognitive functions in children with ASD with implications for future decision making. Systematic review registration: PROSPERO CRD42021279054.

In prior research, various types of PAIS were investigated individually or through narrow groups. In the current study, a full range of PAIS will be examined. PAIS have included aerobic exercise, mindfulness practices, school physical education programs, and Montessori methods [32,37]. Fang and colleagues [38] concluded favorable results on exergaming interventions on cognitive functions but synthesis results from a follow up meta-analysis are lacking. Horseback riding was found to be bene cial to children with ASD in cognitive functioning, social interaction, and sensory processing improvement [33,35]. Some other physical activities, such as yoga and dance, were also introduced as intervention for children with ASD with positive outcomes in behavioral functioning [39]. Medical/clinical conditions of participants, as well as the session frequencies and lengths of the interventions need more research to be conducted. Hence, there are urgent needs for researchers to explicitly examine what type of PAIS is the most effective to bene t children with ASD.
Systematic review and pairwise meta-analysis are unable to compare treatment effects across interventions and indirect effects. Therefore, the network meta-analysis (NMA) is proposed as a novel analysis technique in the eld of physical activity and health promotion to ll the gap and satisfy the urgent need [40]. To the best of our knowledge, the current study will be the rst NMA applied to systematically review and analyze the e cacy of PAIS on cognitive functions in children with ASD. Therefore, our expectation is to explore PAIS based on selected direct/indirect head-to-head trials to suggest important considerations for decision making in the eld of pediatric health promotion.

Mini Review
In order to gain a comprehensive overview on the existing literature as well as justify the necessity of conducting a network meta-analysis on PAIS for children and adolescents with ASD, we conducted a mini review based on the systematic reviews and meta-analyses in the current literature. PubMed, APA PsycInfo, SPORTDiscus, Web of Science, Cochrane Central Register of Controlled Trials, Education Resources Information Center (ERIC), SocINDEX, MEDLINE Complete, and Dimensions were applied for the literature search. The keywords used for the literature search involved four domains: intervention mode, cognitive performance, population, and autism. Combinations of the keywords regarding each domain were "physical activity OR PA OR exercise OR sport*" AND "cognitive function OR cognition OR executive function OR inhibitory control OR memory OR communication OR stereotyp* behavior*" AND "children OR kid* OR teenager OR students" AND "autism spectrum disorder OR ASD". Eligible studies should meet the following criteria: (1) systematic review or meta-analysis written in English and published between 2001 and 2021; (2) children or adolescents with ASD included; (3) PAIS reviewed; and (4) effects of PAIS on cognitive functions examined.

Literature Search
Two authors (LL and AW) independently conducted the search and screening process. The initial search resulted in 424 studies. After the removal of 362 repetitive and irrelevant items, 62 items were left for the next phase of screening. There were 31 articles excluded based on abstract examination, leading to 31 articles for full-text review. The nal step of screening removed 12 articles due to the following reasons: no cognitive outcome measured (n = 10), and populations other than children or adolescents (n = 2). Therefore, a total of 19 systematic reviews with or without meta-analysis met the inclusion criteria. The process of study selection was presented in Figure 1.
The included studies investigated effects of physical activity intervention on cognitive functions, with small to moderate effect sizes on improved global EF [32,50]. Speci cally, the included reviews reported enhanced performance in working memory, attention, task switching, and inhibitory control [34,50]. The evidence also indicated bene cial effects of social functioning and communication skills, which can be attributed to the group-based activity during the interventions [42,44]. Learning and memory outcomes due to PAIS were not fully explored within selected studies. The other limitation of current systematic review and meta-analysis studies was the combination of children and adolescent groups. However, age was recognized as a signi cant confounder in ASD research. Although common clinical treatments including educational, family, speech, and occupational therapy showed favorable effects, widely supportive evidence showed that physical activity was bene cial for children's lifelong cognitive and motor development. Therefore, it is necessary to explicit the PAIS treatment effects on children's cognitive outcomes from a perspective of physical activity and health promotion. Results are encouraging that various types of evidence showed PAIS were effective to enhance cognitive function in children with ASD; the foundation for the future in-depth data analysis has been established.

The Present NMA Protocol
An insight into directions of future research was provided based on the preliminary ndings of the mini review. Previous outcomes suggested that varying types of PAIS were favored in improving cognitive functions and motor performance, reducing frequency of the symptoms, and enhancing social and communication skill development in children with ASD and other cognitive impairments. The quality of evidence is concerning due to the limited number of meta-analyses on RCTs [53]; we expected to simply include true experimental designs (e.g., RCTs) which might ensure robust synthesis results and increase power to detect effects of interventions [51,57]. The age groups were heterogeneous in nature, whereas we expected to focus on children particularly and explore the e cacy of PAIS on their cognitive performances.
However, the limitation of the pairwise meta-analysis procedure was only appropriate for discerning outcome effects "between a selected intervention strategy and a selected comparison or control condition" [40] (p. 511). Considering pairwise meta-analyses were only able to compare multiple outcomes from a single treatment, NMA would make comparing multiple PAIS on multiple outcomes achievable [58]. Concomitantly, there is a need to conduct a NMA to obtain a more accurate estimation of the e cacy of PAIS in the population. We are aiming to provide an updated review of the extant literature in the eld of physical activity and health promotion using an appropriate network meta-analytic model. Therefore, the following protocol was developed upon the ndings and limitations identi ed in the mini review.

Methods
This review protocol has been registered in the PROSPERO international prospective register of systematic reviews; the registration number is: CRD42021279054. The NMA will comply with the Preferred Reporting Items for Systematic Review and Meta Analyses for Network Meta-Analysis (PRISMA-NMA) statement [59,60]. A component-based Bayesian framework NMA will be conducted following the procedures below [61].

Search Strategy
Nine bibliography databases will be searched: APA PsycInfo, Cochrane Central Register of Controlled Trials, Dimensions, ERIC, MEDLINE Complete, PubMed, Scopus, SPORTDiscus, and Web of Science. In general, the search will be limited RCTs published in English ranged from inception to date. Keywords combinations and MeSH terms applied in the search are displayed in Table 2.

Inclusion and Exclusion Criteria
The following explicated the PICOS components regarding population, intervention, comparator, outcomes, and study design.
3.2.1. Population. Children aged between 0-12 years old with a diagnosis of ASD (Asperger's syndrome, autism, or unspeci ed developmental disorder) with varying severity (mild, moderate, severe) will be accepted. Study will include at least one diagnosis criteria regarding Diagnostic and statistical Manual of Mental Disorder [1], International Statistical Classi cation of Diseases and Related Health Problems (ICD), or Autism Diagnostic Observation Schedule (ADOS). In addition, demographic information including subject's age, gender, and sample size should be reported.

3.2.2.
Intervention. Any sort intended to increase cognitive functions in subjects with ASD using physical activity or exercise interventions. Physical activity exhibited in school-based physical education, exergaming, mindfulness practice (e.g., yoga, martial arts, Tai Chi, etc.), aerobic exercise and sports, and any types of therapeutic physical activity interventions will be included. 3.2.6. Exclusion Criteria. Studies will be excluded with participants who were older than 12 years or unable to conduct psychomotor movements; RCT was not used; intervention lasting less than 8 weeks; manuscript was not written in English.

Study Selection and Data Extraction
Recording and managing related literature will be accomplished with Mendeley software (Mendeley Ltd, London, UK). Two authors will independently screen the titles and abstracts of the searched peer-review studies which will be based on the pre-established criteria for inclusion. A third reviewer will be involved to initiate a full-text evaluation if any con icted selections occur. The PRISMA owchart illustrates the process of search and screening ( Figure 2). Study characteristics including author information, publication year, mean age in sample, diagnostic criteria for ASD, study period, intervention arms, and study design will be summarized using Microsoft Excel (Microsoft Excel, Redmond, Washington: Microsoft Corporation, 2010). Speci cally, we will conduct a thematic organization approach to categorize intervention arms within all relevant RCTs into six pre-de ned categories: 1) school physical education programs; 2) exergaming, 3) aerobic exercise and sports, 4) mindfulness practice; 5) computerized and non-computer games combination; 6) therapeutic physical activity interventions [32,37]. The project administrator will serve as the coordinator, dealing with potential disagreements or unclearly de ned PAIS categories. Internal discussions and consultation with an experienced researcher in physical activity and health promotion are considered two solutions for disagreements and uncertainties. The inter-rater agreement rate Cohen's k between reviewers for the eligibility criteria of the study will be reported [62]. To note, grouped/ungrouped or supervised/unsupervised intervention arms will be separated as one effect size per treatment [51]. Quanti ed outcomes [mean of intervention arms and comparators, SDs, 95% con dence interval (CI)s] will be extracted from the aforementioned categories including measurement results regarding cognitive functions (e.g., global EFs, working memory, thinking exibility, inhibitory control, attention, or memory outcomes). The NMA will compare intervention effects of the pre-de ned intervention categories on participants' cognitive performances. Raw data (e.g., sample sizes, t, F, or p values) will be extracted to help estimate the effect size if the means and SDs were not reported [51]. When necessary, the corresponding authors of selected studies will be contacted to obtain unpublished and missing data to allow for effect size estimates.

Assessment of Risk of Bias
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) will be used to assess the quality of selected studies in the NMA and pairwise meta-analysis [63-65]. Two authors will independently assess bias due to randomization, deviations from intervention, missing data, outcome measurement, and selection of reported results in selected RCTs. Each item is classi ed as high, some concerns, low or unclear/no information upon risk of bias and results of the grading of evidence quality will then be generated as a  [40,58]. Consistency of indicating direct and indirect evidence should essentially agree with each other and should not result in discrepancies [40]. The R package developed by van Valkenhoef and colleagues [67] will be employed to evaluate the inconsistency between indirect and direct comparisons via the node-splitting method. We evaluated consistency with a nodesplitting technique that compares the direct and indirect estimates for each comparison. Transitivity will be qualitatively examined to ensure direct trials do not differ with respect to the distribution of effect modi ers which all treatments are jointly randomizable [68].

Statistical Analysis
The primary outcome for this study will be changes in cognitive functions. However, unlike physiological and biochemical indexes, cognitive performances were measured through pre-established validated psychometric instruments (illustrated in Table 1). Guiding through the approach applied by Désaméricq and colleagues [69], the effect size will be calculated for each extracted result and these effect sizes will be then combined into three cognitive scores, as listed in the Supplementary. An omnibus composite unit free score of cognitive outcomes will be created by averaging all subordinate domain effect sizes within a study, per intervention arm. Study included multiple treatments or outcomes will provide the corresponding amount of effect sizes per study. Extracted data such as sample size, treatment and control means and SDs for between group designs; pre-and post-intervention mean and SDs for within group designs, will be converted to correlated summary results, standardized mean differences (SMDs) for NMA to explore the e cacy of PAIS upon cognitive performances [70]. The effect sizes will be computed using SMD with 95% CIs between groups [32,71]. Hedges' g in a random model will be used in consideration of potential variances across the selected studies if the study number is fewer than 20 [32,51,72]. Effect sizes will be classi ed as small, moderate, or high respectively, regarding to the cutoff values of 0.2, 0.5, and 0.8 [32,73]. The calculations will be computed through between-case SMD estimator, the web-application scdhlm [74]. Heterogeneity was evaluated through the means of Q test and I 2 values regarding the statistical heterogeneity in 25% as low, 50% as medium, and 75% as high [32,75,76]. Sensitivity analyses will be conducted to determine if any study separately in uenced the overall results [77]. To minimize the impact of heterogeneous outcomes or extreme values, we will apply the threshold of Z score 3.29 for screening out outliers in mean effect scores [46, 53,78]. The level of signi cance will be set at .05 in this study. The NMA will be performed upon RStudio (version 3.6; [79]) with the BUGSnet package (Version 1.1.0; [80]).

Moderation Analysis
Moderation analysis will be conducted to examine the variability among potential sources owing to heterogeneous PAIS settings. Moderators such as participant age, exercise intensity, and intervention setting (e.g., group versus individual) will be examined via robust variance estimation [51]. Our expectation is providing another perspective of synthesis results to explicit the higher order outcomes from NMA.

Ethics and Dissemination
The ethical approval process is exempted because this study will be based on ndings of previous studies. This review protocol has been registered in the PROSPERO international prospective register of systematic reviews with aims regarding promoting transparency and minimizing the risk of bias [81]. Future results will be submitted to a peer-reviewed journal of the relative eld for consideration of publication.

Discussion
We expect to compare the effects of different PAIS on cognitive outcomes (overall and by domain), over a period of treatment (at least 8 weeks). We will perform an NMA; we will select all randomized controlled trials of considering PAIS as a treatment(s) on cognitive outcomes in children with ASD. This study is going to bridge the gap and provide a novel perspective and robust evidence for clinicians and practitioners for future decision making in children with ASD. Forecasted results will include study characteristics, qualitative appraisal on cognitive tasks and intervention technologies, and quality of selected studies; demonstration visual synthesis results such as network diagram(s) and associated geometry will be generated and evaluated, respectively. Expected results will be able to rank the e cacy of multiple PAIS (e.g., aerobic exercise, mindlessness practices, school-based physical education programs, exergaming, etc.) with effect size differences demonstrated between multiple treatments. In addition to network plots, results will be also illustrated in tables of network characteristics, data plots, league tables and league heat plots, SUCRA plots, and rankograms. In-depth discussion will be generated. In general, results may reveal the rank e cacy of PAIS, in combination or single treatment, on promoting cognitive functions and further social and communication skills in children with ASD.
In addition, confronting with the COVID-19 pandemic, social restriction and isolation were carried out with increasing sedentary behavior and physical inactivity which increased risks of cardiovascular disease and mental wellness crisis in children [82][83][84]. Especially, there are more concerns regarding children with ASD [85,86]. We will include an in-depth discussion about feasible interventions along with safety suggestions for programing during the global pandemic. Promoting effective PAIS to reduce the impact on children with disabilities, such as children with ASD, should be considered as urgent needs nowadays as well as in the post-pandemic era. To the best of our knowledge, this will be the rst systematic review and NMA protocol to shed light on the e cacy of PAIS on cognitive performances in children with ASD. Our study is giving advantages in combining large existing evidence under similar scheme to make indirect comparisons among different forms of PAIS feasible. Therefore, completing this protocol will provide comprehensive and robust outcomes to further explore the multifaceted nature of PAIS in empirical studies. Concomitantly, we have con dence that this study may provide valuable insights and key recommendations for parents, clinicians, and policymakers to make well-informed decisions to improve the cognition development in children with ASD population in the future.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.