The effects of motor skill and physical activity interventions on preschoolers' cognitive and academic skills : A systematic review

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ The effects of motor skill and physical activity interventions on preschoolers' cognitive and academic skills : A systematic review © 2021 The Authors. Published by Elsevier Inc. Published version Jylänki, Pinja; Mbay, Theo; Hakkarainen, Airi; Sääkslahti, Arja; Aunio, Pirjo Jylänki, P., Mbay, T., Hakkarainen, A., Sääkslahti, A., & Aunio, P. (2022). The effects of motor skill and physical activity interventions on preschoolers' cognitive and academic skills : A systematic review. Preventive Medicine, 155, Article 106948. https://doi.org/10.1016/j.ypmed.2021.106948 2022 Preventive Medicine 155 (2022) 106948 Available online 30 December 2021 0091-7435/© 2021 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Review Article The effects of motor skill and physical activity interventions on preschoolers’ cognitive and academic skills: A systematic review Pinja Jyl¨ anki a , * , Theo Mbay a , Airi Hakkarainen b , Arja S¨ a¨ akslahti c , Pirjo Aunio a a Faculty of Educational Sciences, University of Helsinki, Siltavuorenpenger 5, 00170 Helsinki, Finland b Open University, University of Helsinki, Yliopistonkatu 3, 00100 Helsinki, Finland c Faculty of Sport and Health Sciences, University of Jyv¨ askyl¨ a, Keskussairaalantie 4, 40600 Jyv¨ askyl¨ a, Finland ARTICLE INFO Keywords: Motor skills Physical activity Cognition Academic skills Early intervention Systematic review ABSTRACT The present systematic review aimed to investigate the methodological quality and the effects of fundamental motor skills and physical activity interventions on cognitive and academic skills in typically developing 3 to 7year-old children. The review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. A literature search was carried out in April 2020 using seven electronic databases. The methodological quality of the studies was assessed with the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool. Cohen’s d effect size calculations and post hoc power analyses were conducted for the included studies. A total of 35 studies, representing 2472 children met the inclusion criteria. Two of the studies demonstrated a strong methodological quality, while 24 were considered as methodologically weak. The majority (71%) of the included studies demonstrated the beneficial effects of the intervention on cognitive and academic skills. The most evidence was found for executive functions, language, and numeracy, and the effects were largest in enhancing memory. The effects were larger on cognitive and academic skills in the combined interventions compared to only fundamental motor skill and physical activity interventions, while fundamental motor skill interventions had larger effects than physical activity interventions. These findings indicate that it may be possible to support typically developing preschoolers’ cognitive and academic learning with fundamental motor skill and physical activity interventions. However, most of the studies in this field have a weak methodological quality and thus, the presented evidence was considered weak in nature. 1. Introduction Early childhood is a critical phase in the development of fundamental motor (FMS; Gallahue and Ozmun, 2002), cognitive (Burger, 2010), and academic skills (e.g., early numeracy and language skills; Duncan et al., 2007); which has shown to predict later school adjustment (Bart et al., 2007) as well as academic achievements (Duncan et al., 2007). FMS can be seen as the “building blocks” of more complex movement skills that are required in games or other physical activities (Logan et al., 2018), and consists of balance (e.g. rolling, turning), manipulative (e.g., catching, kicking), and locomotor movement skills (e.g., running, jumping; Gallahue and Ozmun, 2002). Physical activity (PA) includes all voluntary bodily movements produced by skeletal muscles that result in energy expenditure (Caspersen et al., 1985). When young children perform FMS, which are produced by large muscles of the body, the energy expenditure and PA is high (Malina et al., 2004; Gallahue and Ozmun, 2002). Thus, among young children there is very close relationship between FMS and PA (Stodden et al., 2008; Jones et al., 2020). In recent years, the research has increasingly focused on the effects of FMS and PA on cognitive and academic skills and has shown a positive relationship between these skills in preschoolers (Planinˇ sec, 2002). A potential explanation for the close association between FMS and cognitive skills is the co-activation of the same brain areas (i.e., cerebellum and prefrontal cortex; Diamond, 2000). It has also been suggested that the effects of PA on cognitive and academic skills may be mediated by other factors such as physical fitness, health, and psycho-social factors (Tomporowski et al., 2011). Indeed it has been shown that physical fitness and cognitive skills are related in young children (Ruiz-Hermosa et al., 2020), however the relationship between PA and physical fitness will strengthen in the later childhood (Stodden et al., 2008). Although the positive associations between FMS and PA with various * Corresponding author. E-mail address: [email protected] (P. Jyl¨ anki). Contents lists available at ScienceDirect Preventive Medicine journal homepage: www.elsevier.com/locate/ypmed https://doi.org/10.1016/j.ypmed.2021.106948 Received 24 June 2021; Received in revised form 18 November 2021; Accepted 27 December 2021 Preventive Medicine 155 (2022) 106948 2 health indicators (i.e., cardiometabolic health, bone and skeletal health, psychosocial health, self-esteem, and cognitive development) have been demonstrated (Carson et al., 2017; Rasmussen and Laumann, 2013), children have become less physically active during recent years and almost half of preschool-aged children do not meet the recommended daily PA guidelines (i.e., at least 60 min of physical activity and several hours of unstructured play per day; National Association for Sport and Physical Education, 2002; Tucker, 2008). As a consequence, children’s FMS, especially coordinative skills, have been declining (Roth et al., 2010), and children spent excessive time being sedentary (e.g., engage in screen-based entertainments; Hinkley et al., 2012). This increased sedentary time during early childhood has been negatively associated with children’s cognitive and academic skills (Carson et al., 2015). Due to the widespread attendance, preschools exhibit a crucial role as a supporter of beneficial development in FMS, PA habits (Tonge et al., 2016) and cognitive skills (Burger, 2010). Previous systematic reviews that have examined the effects of FMS or PA interventions on cognitive and academic skills, have concentrated mainly on school-aged children (i.e., 5–12-year-olds) and included only PA interventions (Watson et al., 2017), have used inadequate quality assessment and included a small number of studies (Zeng et al., 2017) or have not conducted effect size calculations (Carson et al., 2016). Thus, the present systematic review aimed to investigate 1) the methodical quality and 2) the effects of FMS and PA interventions on cognitive and academic skills in 3 to 7-year-old children. In addition, the effectiveness of only FMS and/or PA interventions were compared with combined interventions (FMS and/or PA combined with cognitive or academic skill practices, e.g., FMS practices combined with story reading; Bedard et al., 2017). Because this systematic review concentrated in 3 to 7-yearold children, both FMS and PA interventions were included. In order to separate FMS and PA interventions from each other, in this review PA interventions were conceptualized as interventions that do not include any specific skill practices but rather focuses on increasing PA. 2. Methods The systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement, which can be used especially in evaluations of intervention studies (Moher et al., 2009). The following steps were carried out in this systematic review: Step 1: A literature search including abstract rating and full-text screening based on the predetermined eligibility criteria. Step 2: Methodological quality of the studies was assessed with the Effective Public Health Practice Project Quality Assessment Tool for Quantitative Studies (EPHPP). Step 3: Cohen’s d effect sizes were calculated and post hoc power analyses were conducted to determine the statistical power of the selected studies. 2.1. Eligibility criteria Peer-reviewed intervention studies, investigating the effects of FMS or PA interventions on preschoolers’ cognitive or academic skills, that were published in English were eligible to be included in the present systematic review. Eligibility criteria included the following: The main content of studies: intervention studies with at least one program including only FMS and/or PA practices or a program that combined FMS and/or PA with cognitive skill practices. Target population: typically developing children between the ages 3 and 7. Based on the participant description, children with special needs, low socioeconomic status, or migrant background were excluded from the present review since the lower baseline scores might have affected the generalizability of the results. Study design: all intervention designs, except case studies, were included; i.e., randomized controlled trials (RCT), controlled clinical trials (CCT), pre-post designs with a control group (PPD), and acute interventions. Outcome measures: intervention effects were assessed with cognitive or academic skills (i.e., executive functions, cognitive performance, language skills, numeracy, or memory) measures. Type of publication: published peer-reviewed journal articles written in English. All publications before April 16th, 2020 were searched. The systematic literature search was carried out in April 2020 by two authors using seven electronic databases including ERIC, Scopus, Web of Science, PsycINFO, CINAHL, PubMed and SPORTDiscus. Search terms were designed in accordance with the PICO framework (Melnyk and Fineout-Overholt, 2005). The final search string consisted of the following: “early education” OR child* AND motor* OR “physical activity” AND intervention OR program* OR treatment OR training OR instruction AND cognit* OR academic*. When possible, additional search filters were used to exclude studies that investigated children older than 7 years. 2.2. Study selection Two authors performed the article selection according to the predetermined eligibility criteria. Initially, the articles were screened based on the title and the abstract. In terms of inclusion, the abstracts were coded as “yes”, “maybe” or “no”. The inter-rater agreement was determined during the abstract rating processes by calculating Cohen’s weighted kappa. Both authors rated the first 40% (n =2266) of the abstracts independently, after which the inter-rater agreement was 0.718, which can be considered as a good agreement between the raters (Byrt, 1996). After the abstract screening, all of the eligible articles underwent full-text screening. Both authors decided independently whether to “exclude”, “include” or “maybe” include each article. Disagreements between the two authors were solved in consensus meetings with all of the five authors. 2.3. Methodological quality The methodological quality of the eligible studies was assessed with the EPHPP tool (National Collaborating Centre for Methods and Tools, 2008; Thomas et al., 2004). The tool is suitable for evaluating the quality of a variety of study designs (e.g., RCTs and PPDs; Armijo-Olivo et al., 2012) and it has been used previously in systematic reviews in this particular field (e.g., Norris et al., 2015). The inter-rater agreement has shown to be more consistent in the EPHPP tool comparison to the Cochrane Collaboration Risk of Bias Tool (Armijo-Olivo et al., 2012). Two authors rated each study procedure as “strong”, “moderate” or “weak”. Final ratings were formed based on six sections (selection bias, study design, confounders, blinding, data collection methods, and withdrawals and drop-outs) with the following criteria: studies with no weak ratings and at least four strong ratings were considered as “strong”; studies with less than four strong ratings and one weak rating were considered as “moderate”; studies with two or more weak ratings were considered as “weak”. For more detailed criteria see Quality Assessment Tool for Quantitative Studies Dictionary (National Collaborating Centre for Methods and Tools, 2008). Disagreements were solved in consensus meetings with all of the authors. 2.4. Data extraction Data were extracted from the eligible studies independently by two authors. Extracted data included the geographical location, study design, sample size, children’s age and gender, cognitive and academic skill measures, intervention exposure, intervention details (only FMS and/or PA interventions and combined interventions), control conditions, and data for effect size calculations. If missing data was encountered, the corresponding author was contacted in order to receive the required information. P. Jyl¨ anki et al. Preventive Medicine 155 (2022) 106948 3 2.5. Effect size calculations Cohen’s d effect sizes (Cohen, 1988) were calculated to allow for the quantification and comparison of the effects across the studies. Effect sizes were calculated for the studies that demonstrated significant effects and provided sufficient information (i.e., pre and post-scores as well as associated standard deviations or standard errors). If a study demonstrated significant effects for multiple outcomes, all of them were included. Cohen’s d effect sizes of <0.2, 0.2, 0.5, and 0.8, correspond to trivial, small, medium, and large effects, respectively (Cohen, 1988). Between-group effects were calculated followingly. ES(d)=(Mpost,E–Mpre,E)–(Mpost,C–Mpre,C) SDpooledpre where SDpooledpre = ((nE–1)SDpre2,E+ (nC–1)SDpre2,C) (nE+nC–2) √ , and within-group effects were calculated as ES(d)=(Mpost–Mpre) SDpre For the studies assessing the learning of previously unknown items, i.e., where the baseline score was approximating zero (Padial-Ruz et al., 2019; Mavilidi et al., 2015; Mavilidi et al., 2016a; Mavilidi et al., 2017a; Mavilidi et al., 2018a; Toumpaniari et al., 2015), the effect sizes were calculated as: ES(d)=(Mpost,E–Mpost,C) SDpooledpost where SDpooledpost = ((nE–1)SDpost2,E+ (nC–1)SDpost2,C) (nE+nC–2) √ ES (d) =Cohen’s d effect size Mpost =mean post score Mpre =mean pre score E =experimental group C =control group SDpooled =pooled standard deviation n =sample size. 2.6. Power analyses Power calculations were carried out with G*power 3.1.9.6 (Faul et al., 2007). If a study had multiple groups, the power calculations were conducted on a sub-group basis in order to determine the power of specific group comparisons. Type 1 error probability ( α ) was computed as 0.05, corresponding to a significance level of 5%. A medium effect size (0.5) was used as the reference point to establish observed power for each outcome and type 2 error probability (β) of 0.2, corresponding to a power of 0.8 (1 – β), or 80%, was selected as the cut-off point for adequate power (Schulz and Grimes, 2005). 3. Results 3.1. Search results In the systematic literature search a total number of 10,995 articles, 5665 after duplicates were removed, were found. Additional 4 articles were found through other sources. A total number of 5669 articles were screened based on the abstract and title. After that, 5471 articles were excluded due to not meeting the eligibility criteria and 198 full-text articles were further assessed for eligibility. Finally, 35 articles were included in the main analysis (see Fig. 1). 3.2. Study characteristics and population All study characteristics, including authors, study design, participants, outcome measures, intervention duration, and intervention/ control activities, are presented in detail in Supplementary Table S1. Included studies were published between March 2010 (Ellemberg and St-Louis-Deschˆ enes, 2010) and February 2020 (Botha and Africa, 2020). The studies were conducted in 17 countries, representing North America, Europe, Asia, Australia, and Africa (see Supplementary Table S1). Participants’ ages ranged from 3.3 years (Bedard et al., 2020) to 7.8 years (Ellemberg and St-Louis-Deschˆ enes, 2010). One study (Ellemberg and St-Louis-Deschˆ enes, 2010) included only boys and other studies included both girls and boys (i.e., 37.8% (Jarraya et al., 2019) - 78.6% (Gao et al., 2019) of the participants were boys). The total sample in the 35 studies included 2472 children. 3.3. Intervention characteristics In total, 35 articles with 46 intervention programs were included in the present review. Eleven studies (Barnard et al., 2014; Callcott et al., 2015a; Chang et al., 2013; Duncan et al., 2019; Jarraya et al., 2019; Mavilidi et al., 2015; Mavilidi et al., 2016a; Mavilidi et al., 2017a; Mavilidi et al., 2018a; Shoval et al., 2018; Xiong et al., 2019a) included two programs that met the eligibility criteria. Seventeen of the intervention programs focused on FMS practises, 15 programs on PA practises and 14 programs on both FMS and PA practises. Twenty-four of the interventions were only FMS and/or PA programs and 22 interventions combined FMS and/or PA practices with cognitive and/or academic skill practices in the program. Intervention duration varied from one week (Mavilidi et al., 2016a; Mavilidi et al., 2016b) to one academic year (Bala et al., 2013; Have et al., 2018). Two of the interventions (Ellemberg and St-Louis-Deschˆ enes, 2010; Stein et al., 2017) were acute interventions, and in three studies (Callcott et al., 2015a; Habibi et al., 2014; Raney et al., 2017; Shoval et al., 2018) the intervention duration was not reported. The duration of one intervention session varied from five minutes (St. Laurent et al., 2018) to three hours (Shoval et al., 2018), and sessions were held once a week (Bedard et al., 2017; Bedard et al., 2020; Fisher et al., 2011; Mavilidi et al., 2017a; Mavilidi et al., 2017b; Mavilidi et al., 2018a; Mavilidi et al., 2018b; Pienaar et al., 2011) to six times a week (Have et al., 2018). Outcome measures were assigned to six categories based on the provided description of the measure. The categories were as follows: executive functions (e.g., Eriksen Flanker task (Beck et al., 2016; Chang et al., 2013; Have et al., 2018), visuo-spatial short-term memory; Have et al., 2018), language (e. g., the ESSI reading and spelling tests; Bala et al., 2013; Shoval et al., 2018), cognitive skill (e.g., Raven Progressive Matrices assessment test; Bala et al., 2013; Shoval et al., 2018), numeracy (e.g., Mathematics Achievement Test; Shoval et al., 2018), memory (e.g., free-recall and cued recall tests; Mavilidi et al., 2015; Mavilidi et al., 2017a) and other tests (curriculum test; Raney et al., 2017). 3.4. Methodological quality The methodological quality was determined based on the following factors: selection bias, study design, confounders, blinding, data collection methods, and withdrawals and drop-outs. Only two of the included 35 studies (6%) demonstrated strong methodological quality, while nine of the studies (26%) had moderate quality and 24 of the studies (69%) were considered as methodologically weak. The rating for each section as well as the overall quality of the studies are presented in Table 1. P. Jyl¨ anki et al. Preventive Medicine 155 (2022) 106948 4 Of the 35 studies, 30 were CCTs (i.e., quasi-experimental designs with a control group and RCTs that did not report the randomization process) and five studies were RCTs. In 19 studies (Bala et al., 2013; Barnard et al., 2014; Battaglia et al., 2019; Beck et al., 2016; Botha and Africa, 2020; Callcott et al., 2015a; Duncan et al., 2019; Fisher et al., 2011; Have et al., 2018; Mavilidi et al., 2015; Mavilidi et al., 2016a; Mavilidi et al., 2017a; Mavilidi et al., 2018a; S´ anchez-L´ opez et al., 2019; Shoval et al., 2018; Stein et al., 2017; Toumpaniari et al., 2015; Xiong Records excluded (n = 5471) Full-text articles excluded, with reasons (n = 163) Reasons: Children older than 7 years (n = 50) No FMS or PA intervention (n = 31 ) Not typically developing children (n = 20) •No cognitive or academic outcomes (n = 17) Children younger than 3 years (n = 14) Case studies (n = 12) Duplicates (n = 4) No control group (n = 4) Follow-up study (n = 3) Not published in English (n = 1) Book (n = 1) Conference presentation (n = 1) Animal study (n = 1) Studyplan(n= 1) Review (n = 1) Qualitative study (n = 1) Not found (n = 1) Only FMS/PA interventions (n = 24) Combined FMS/PA and cognitive/academic interventions (n = 22) Additional records identified through other sources (n = 4) Records identified through database searching (n = 10995) ERIC (n = 318) Pubmed (n = 1742) SportDISCUS (n = 383) CINAHL (n = 685) Web of Science (n = 4165) PsycINFO (n = 902) Scopus (n = 2800) Records after duplicates removed (n = 5669) Records screened (n = 5669) Full-text articles assessed for eligibility (n = 198) Studies included in qualitative synthesis (n = 35) Included Eligibility Screening Identification Fig. 1. PRISMA flow diagram of the stages associated with the systematic selection of studies. P. Jyl¨ anki et al. Preventive Medicine 155 (2022) 106948 5 et al., 2017; Xiong et al., 2019a), the participants were referred from a source (e.g., preschool) in a systematic manner, and thus the participants were considered only somewhat likely to be representative of the target population. None of the studies referred the participants through randomization. In 18 studies (Bedard et al., 2017; Mavilidi et al., 2017a; Chang et al., 2013; Have et al., 2018; Botha and Africa, 2020; Bala et al., 2013; Bedard et al., 2020; Ellemberg and St-Louis-Deschˆ enes, 2010; Ramah, 2014; Jarraya et al., 2019; Habibi et al., 2014; Battaglia et al., 2019; Gao et al., 2019; Lee et al., 2020; Wen et al., 2018; Duncan et al., 2019; Xiong et al., 2019a; Stein et al., 2017) there were not any important confounders between groups (i.e., participants’ age, gender, health status or pre-intervention score) and in four studies (Beck et al., 2016; Fisher et al., 2011; Shoval et al., 2018; St. Laurent et al., 2018) the confounders were 80–100% controlled. Only six studies (Botha and Africa, 2020; Fisher et al., 2011; Gao et al., 2019; Have et al., 2018; Jarraya et al., 2019; Wen et al., 2018) reported that the outcome assessors were not aware, and two studies (Battaglia et al., 2019; Raney et al., 2017) reported that the outcome assessors were aware of the intervention exposure. The validation of the data collection methods was reported in 12 studies (Bedard et al., 2017; Bedard et al., 2020; Callcott et al., 2015a; Duncan et al., 2019; Gao et al., 2019; Have et al., 2018; Lee et al., 2020; Pienaar et al., 2011; S´ anchez-L´ opez et al., 2019; Shoval et al., 2018; Wen et al., 2018; Xiong et al., 2019a) and of these, only six studies (Callcott et al., 2015a; Gao et al., 2019; Lee et al., 2020; Pienaar et al., 2011; Wen et al., 2018; Xiong et al., 2019a) reported that the data collection methods demonstrated good reliability in that particular data. Most of the studies (Beck et al., 2016; Bedard et al., 2017;: Bedard et al., 2020; Botha and Africa, 2020; Ellemberg and St-Louis-Deschˆ enes, 2010; Fischer et al., 2011; Fisher et al., 2011; Gao et al., 2019; Have et al., 2018; Habibi et al., 2014; Lee et al., 2020; Mavilidi et al., 2015; Mavilidi et al., 2016a; Mavilidi et al., 2017a; Mavilidi et al., 2018a; Padial-Ruz et al., 2019; S´ anchez-L´ opez et al., 2019; St. Laurent et al., 2018; Xiong et al., 2017) reported both withdrawals and drop-outs in terms of numbers and reasons. Thirty-one (89%) of the included studies were found to be underpowered to detect a medium effect size, and only nine of the included studies reported having conducted a priori power analyses (Raney et al., 2017; Jarraya et al., 2019; Habibi et al., 2014; Gao et al., 2019; Lai et al., 2020; Stein et al., 2017; Wen et al., 2018; Xiong et al., 2019a; Have et al., 2018); of which only one was adequately powered (Have et al., 2018) Table 1 Methodological quality of the included studies. Reference Selection bias Study design Confounders Blinding Data collection methods Withdrawal and dop outs Overall quality scores Bala et al., 2013 Moderate Strong Strong Moderate Weak Weak Weak Barnard et al., 2014 Moderate Strong Weak Moderate Weak Weak Weak Battaglia et al., 2019 Moderate Strong Strong Weak Weak Weak Weak Beck et al., 2016 Moderate Strong Strong Moderate Weak Strong Moderate Bedard et al., 2017 Weak Strong Strong Moderate Moderate Strong Moderate Bedard et al., 2020 Weak Strong Strong Moderate Moderate Strong Moderate Botha and Africa, 2020 Moderate Strong Strong Strong Weak Strong Moderate Callcott et al., 2015a Moderate Strong Weak Moderate Strong Weak Weak Chang et al., 2013 Weak Strong Strong Moderate Weak Weak Weak Duncan et al., 2019 Moderate Strong Weak Moderate Moderate Weak Weak Ellemberg and St-Louis-Deschˆ enes, 2010 Weak Strong Strong Moderate Weak Strong Weak Fischer et al., 2011 Weak Strong Weak Moderate Weak Strong Weak Fisher et al., 2011 Weak Strong Strong Strong Weak Strong Weak Gao et al., 2019 Weak Strong Strong Strong Strong Strong Moderate Habibi et al., 2014 Weak Strong Strong Moderate Weak Strong Weak Have et al., 2018 Moderate Strong Strong Strong Moderate Strong Strong Jarraya et al., 2019 Weak Strong Strong Strong Weak Weak Weak Lai et al., 2020 Weak Strong Weak Moderate Weak Weak Weak Lee et al., 2020 Weak Strong Strong Moderate Strong Strong Moderate Mavilidi et al., 2015 Moderate Strong Weak Moderate Weak Strong Weak Mavilidi et al., 2016a Moderate Strong Weak Moderate Weak Strong Weak Mavilidi et al., 2017a Moderate Strong Strong Moderate Weak Strong Moderate Mavilidi et al., 2018a Moderate Strong Weak Moderate Weak Strong Weak Padial-Ruz et al., 2019 Weak Strong Weak Moderate Weak Strong Weak Pienaar et al., 2011 Weak Strong Weak Moderate Strong Weak Weak Ramah, 2014 Weak Strong Strong Moderate Weak Weak Weak Raney et al., 2017 Weak Strong Weak Moderate Weak Weak Weak S´ anchez-L´ opez et al., 2019 Moderate Strong Strong Moderate Moderate Moderate Moderate Shoval et al., 2018 Moderate Strong Strong Moderate Moderate Weak Moderate Stein et al., 2017 Moderate Strong Strong Moderate Weak Weak Moderate St. Laurent et al., 2018 Weak Strong Strong Moderate Weak Moderate Weak Toumpaniari et al., 2015 Moderate Strong Weak Moderate Weak Weak Weak Wen et al., 2018 Weak Strong Strong Strong Strong Weak Weak Xiong et al., 2017 Moderate Strong Weak Moderate Weak Strong Weak Xiong et al., 2019a Moderate Strong Strong Moderate Strong Strong Strong Note. Some modifications were made to the EPHPP tool to solve misunderstandings between the raters. Study design: Studies that used quasi-experimental design were coded as CCT. Confounders: The confounders of interest included age, gender, health status and pre-intervention score. Blinding: In question 2 “Were the study participants aware of the research question?” we chose to code “no”, if there was no mention that participants were aware of the research question. This decision was made based on the young age of the participants. Data collection methods: The outcome of interest (cognitive or academic measurement) was evaluated. Methods were coded to be “valid” if the validity was mentioned in the article or if there was a citation to test manual or other article where the validity was reported. Some well known methods were seen as valid methods without mentioning (e.g., Wechsler Intelligence Scale for Children or Bayley Scales of Infant and Toddler Development). Methods were coded as “reliable” only, if the reliability was measured and reported in that specific data. Withdrawals and drop-outs: In question 1, “Were withdrawals and drop-outs reported in terms of numbers and/or reasons per group?”, if both numbers and reasons are reported then it was code as “yes”, otherwise “no”. Withdrawals and drop-outs were considered as children that did not finish the intervention, i.e. not missing data. P. Jyl¨ anki et al. Preventive Medicine 155 (2022) 106948 6 and five stated low statistical power as a limitation of the study (Gao et al., 2019; Jarraya et al., 2019; Lai et al., 2020; Wen et al., 2018; Xiong et al., 2019a). 3.5. Effect sizes In terms of the intervention effects (see Supplementary Table S2), twelve of the included studies assessed executive functions as an outcome variable; three with combined interventions (Gao et al., 2019; Have et al., 2018; Beck et al., 2016) and nine with FMS and/or PA only interventions (Chang et al., 2013; Ellemberg and St-Louis-Deschˆ enes, 2010; Fisher et al., 2011; Jarraya et al., 2019; Lai et al., 2020; Stein et al., 2017; Wen et al., 2018; Xiong et al., 2017; Xiong et al., 2019a). 1 Two of the three (67%) combined interventions (Gao et al., 2019; Beck et al., 2016), and eight out of nine (89%) FMS and/or PA only interventions (Chang et al., 2013; Ellemberg and St-Louis-Deschˆ enes, 2010; Fisher et al., 2011; Jarraya et al., 2019; Lai et al., 2020; Stein et al., 2017; Xiong et al., 2017; Xiong et al., 2019a) demonstrated significant effects. The effect sizes were medium (d =0.50–0.68, x‾ 0.59) for FMS and/or PA only interventions and ranged from trivial to large for combined interventions (d =0.12–1.47, x‾ 0.68). Out of the two null-findings (Wen et al., 2018; Have et al., 2018) one study (50%; Wen et al., 2018) was underpowered to detect a medium effect. A total of ten studies assessed language skills as an outcome variable; four were FMS and/or PA only interventions (Barnard et al., 2014; Battaglia et al., 2019; St. Laurent et al., 2018; Habibi et al., 2014), three were combined interventions (Bedard et al., 2017; Bedard et al., 2020; Botha and Africa, 2020), and three included both (Callcott et al., 2015a; Duncan et al., 2019; Shoval et al., 2018). Three (Duncan et al., 2019; Barnard et al., 2014; Callcott et al., 2015a) of the seven (43%) FMS and/or PA only interventions and four (Botha and Africa, 2020; Callcott et al., 2015a; Duncan et al., 2019; Shoval et al., 2018) out of six (67%) combined interventions demonstrated significant effects. The effect sizes ranged from trivial to small (d =0.16–0.25, x‾ 0.20) for FMS and/or PA only interventions and from small to large for combined interventions (d =0.31–2.2, x‾ 0.90). All of the null findings (Battaglia et al., 2019; St. Laurent et al., 2018; Shoval et al., 2018; Habibi et al., 2014; Bedard et al., 2017) were underpowered to detect a medium effect. Furthermore, three of the studies (Callcott et al., 2015a; Duncan et al., 2019; Shoval et al., 2018) compared FMS and/or PA only interventions to combined interventions and demonstrated small effects (d =0.23–0.36, x‾ 0.31) in favor of the combined interventions. Seven studies included a numeracy measure as an outcome variable; one was a FMS and/or PA only intervention (Barnard et al., 2014), five were combined interventions (Mavilidi et al., 2018a; St. Laurent et al., 2018; Have et al., 2018; Beck et al., 2016; Fischer et al., 2011) and one included both (Shoval et al., 2018). None of the FMS and/or PA only interventions, and five (83%) of the six combined interventions (Mavilidi et al., 2018a; Shoval et al., 2018; Have et al., 2018; Beck et al., 2016; Fischer et al., 2011) demonstrated significant effects, which ranged from small to large (d =0.38–0.94, x‾ 0.57). Furthermore, one of the studies compared FMS and/or PA only to a combined intervention (Shoval et al., 2018) and found a large effect (d =0.89) in favor of the combined intervention. Of the three null findings (Barnard et al., 2014; St. Laurent et al., 2018; Shoval et al., 2018), one was underpowered to detect a medium effect (St. Laurent et al., 2018). Five studies (Padial-Ruz et al., 2019; Mavilidi et al., 2017a; Mavilidi et al., 2016a; Mavilidi et al., 2015; Toumpaniari et al., 2015) investigated the effects of adding FMS and/or PA on a learning task to enhance memory. All of the studies demonstrated large significant effects (d = 1.23–2.2, x‾ 1.52) when the FMS and/or PA was directly related to the learning task. Three of these studies also included a group performing FMS and/or PA unrelated to the learning activity (Mavilidi et al., 2015; Mavilidi et al., 2016a; Mavilidi et al., 2017a), and two (67%) found significant large effects (d =0.82–1.20, x‾ 1.0; Mavilidi et al., 2016a; Mavilidi et al., 2017a) while one did not find any significant effects (Mavilidi et al., 2015). The one null finding (Mavilidi et al., 2015) was underpowered to detect a medium effect. A total of seven studies assessed cognitive skills as an outcome variable, of which six were FMS and/or PA only interventions (Habibi et al., 2014; Bala et al., 2013; Pienaar et al., 2011; Lee et al., 2020; Ramah, 2014; S´ anchez-L´ opez et al., 2019) and one included a FMS and/or PA only intervention as well as a combined intervention (Shoval et al., 2018). Of the seven FMS and/or PA only interventions, three (43%) demonstrated significant effects (S´ anchez-L´ opez et al., 2019; Shoval et al., 2018; Ramah, 2014), with effect sizes ranging from small to large (d =0.40–1.3, x‾ 0.77). The one combined intervention found a significant medium effect (d =0.60) and when compared to FMS and/or PA only, demonstrated a small effect in favor of the combined intervention (d =0.35). All of the four null findings (Bala et al., 2013; Habibi et al., 2014; Pienaar et al., 2011; Lee et al., 2020) were underpowered to detect a medium effect. Finally, one study used a biology curriculum test as a measure and did not observe a significant impact of a combined intervention (Raney et al., 2017). The one null finding (Raney et al., 2017) was underpowered to detect a medium effect. 3.6. Methodological quality and effect sizes Methodological quality and effect sizes are presented in Table 2. Methodologically strong studies found medium (Xiong et al., 2019a) and trivial effects (Xiong et al., 2019b) on executive functions and small effects (Have et al., 2018) on numeracy. Large effects were found in 13 studies and seven of these (54%) had a weak, while six (46%) had a moderate methodological quality. Of these, only two studies (S´ anchez-L´ opez et al., 2019; Shoval et al., 2018) used outcome measures that were shown to be valid, while three studies (Mavilidi et al., 2016a; Mavilidi et al., 2017a; Mavilidi et al., 2018a) used outcome measures that were shown to be reliable. Six of these studies (Jarraya et al., 2019; Stein et al., 2017; Lai et al., 2020; Botha and Africa, 2020; Mavilidi et al., 2015; Toumpaniari et al., 2015) used outcome measures that were neither shown to be valid or reliable. The studies that received a strong rating in the data collection methods section demonstrated medium effects in three studies (Callcott et al., 2015b; Gao et al., 2019; Xiong et al., 2019a), small effects in one study (Callcott et al., 2015b), and trivial effects in one study (Xiong et al., 2019b). 4. Discussion The present systematic review aimed to investigate the methodological quality and the effects of FMS and PA interventions on cognitive and academic skills in typically developing 3–7-year-old children. These findings demonstrated that only 6% of the included studies had a strong methodological quality, while 69% were considered methodologically weak, which in general was the result of inadequate reporting practices. The majority (71%) of the included studies (i.e., both FMS and/or PA only and combined interventions) demonstrated beneficial effects on cognitive and academic performance of preschoolers. The most evidence was found for the benefits on executive functions, language and numeracy, and the effect sizes were largest in enhancing memory. However, the studies that found large effects had low-to-moderate methodological quality and none of these studies used both validated and reliable outcome measures. In general, combined interventions resulted in beneficial outcomes more consistently than FMS and/or PA only interventions, as well as demonstrated greater effects in magnitude. In the comparison of FMS or PA only interventions, FMS only interventions appeared to have larger effects on cognitive and academic skills than PA only interventions. 1 Chang et al., 2013 and Xiong et al., 2019a included active control groups and thus, the effects were analysed using a within group analysis. P. Jyl¨ anki et al. Preventive Medicine 155 (2022) 106948 7 Table 2 Summary of the relationship between methodological quality and intervention effects. Intervention Outcome Effect size (d) Not significant Trivial Small Medium Large FMS Executive functions Stein et al. 2017 ᶜ Stein et al. 2017 ᵇ Jarraya et al. 2019a ᶜ; Stein et al. 2017 ᵇ Language skills Battaglia et al. 2019ᶜ; Callcott et al. 2015a ᶜ Duncan et al. 2019a ᵇ Cognitive skills Bala et al. 2013 ᶜ; Pienaar et al. 2011 ᶜ; Lee et al. 2020 ᶜ Ramah et al. 2014 ᶜ PA Executive functions Wen et al. 2018 ᶜ Ellemberg et al. 2010 ᶜ Ellemberg et al. 2010 ᶜ Language skills Habibi et al. 2014 ᶜ Cognitive skills Habibi et al. 2014 ᶜ FMS & PA Executive functions Fisher et al. 2011 ᶜ; Jarraya et al. 2019bᶜ; Lai et al. 2020 ᶜ Xiong et al. 2019b ᶜ Chang et al. 2013b ᵇ Chang et al. 2013a ᵇ; Fisher et al. 2011 ᶜ; Xiong et. al. 2017 ᶜ; Xiong et al. 2019a ᶜ Lai et al. 2020 ᵇ Language skills Barnard et al. 2014ᵃ; Shoval et al. 2018a ᶜ Barnard et al. 2014 ᵃ Cognitive skills Shoval et al. 2018a ᶜ ᵉ Shoval et al. 2018a ᶜ Sánchez‐López et al. 2019 ᶜ Numeracy Barnard et al. 2014ᵃ; Shoval et al. 2018a ᶜ FMS combined Executive functions Beck et al. 2016 ᶜ Beck et al. 2016 ᵇ Beck et al. 2016 ᵇ Language skills Bedard et al. 2017 ᶜ; Bedard et al. 2020 ᶜ; Callcott et al. 2015b ᶠ Callcott et al. 2015b ᶜ ᵈ Duncan et al. 2019b ᵈ ᶠ Botha et al. 2020 ᶜ; Callcott et al. 2015bᶜ; Duncan et al. 2019b ᵇ Botha et al. 2020 ᶜ Numeracy Beck et al. 2016 ᶜ; Fischer et al. 2011 ᶜ Beck et al. 2016 ᵇ (Fischer et al. 2011 ᶜ) PA combined Executive functions Have et al. 2018 ᶜ Numeracy Mavilidi et al. 2018bᶜ (Have et al. 2018 ᶜ) Mavilidi et al. 2018a ᶜ Memory performance Mavilidi et al. 2015bᶜ Mavilidi et al. 2015a ᶜ; Mavilidi et al. 2016a ᶜ; Mavilidi et al. 2016bᶜ; Mavilidi et al. 2017a ᶜ; Mavilidi et al. 2017b ᶜ; Toumpaniari et al. 2015ᶜ Biology curriculum test Raney et al. 2017 ᶜ FMS & PA combined Executive functions Gao et al. 2019 ᶜ Language skills St. Laurent et al. 2018 ᶜ Shoval et al. 2018b ᶜ ᵈ Cognitive skills Shoval et al. 2018b ᶜ ᵈ Shoval et al. 2018b ᶜ Numeracy St. Laurent et al. 2018 ᶜ Shoval et al. 2018b ᶜ ᵈ Shoval et al. 2018b ᵈ Note. a and b in the end of the citation refers to different intervention programs within one study. ᵃ Barnard et al. 2014a and Barnard et al. 2014b analyzed together ᵇ Within group analysis ᶜ Intervention compared to control ᵈ Combined intervention compared to only FMS/PA intevention ᵉ Only FMS/PA intevention compared to combined intervention ᶠ Combined intervention compared to academic only Methodological quality based on the EPHPP: weak, moderate, strong If it was not possible to calculate effect sizes but the study reported a self-calculated Cohen's d effect size, it was presented in parentheses. P. Jyl¨ anki et al. Preventive Medicine 155 (2022) 106948 8 In terms of the intervention effects, the largest effects of FMS and/or PA interventions were found for memory, and these benefits were consistently shown only when the PA was directly related to the learning task (e.g., moving in the manner of a particular animal when learning the names of animals in foreign language), and less consistently when the PA was unrelated to the activity (e.g., running in circles after the learning task). Nonetheless, more research is required to elucidate whether these effects translate to general outcomes. The most commonly assessed outcomes in the interventions were executive functions and language skills. For executive functions, 83% of the interventions demonstrated significant benefits with medium effect sizes, with no difference in terms of the magnitude of the effect in FMS and/or PA only and combined interventions. For language skills, 54% of the interventions found significant benefits, with combined interventions resulting in large effects, while FMS and/or PA only interventions yielded small effects. Similar trend was also observed in terms of numeracy, where 83% of the combined interventions found significant medium effects, while none of the FMS and/or PA only interventions did. While these findings suggest that combined interventions are superior to FMS and/or PA interventions, it is important to note that in the vast majority of the combined interventions, the outcome measure was directly related to the contents of the intervention. Therefore, it cannot be concluded that combining cognitive tasks to FMS and/or PA is superior per se; rather it is possible that the differences stem simply from direct practice and assessing of a particular skill. In order to address this question, studies should include an additional control group that receives the same academic content without the FMS and/or PA component. In the present review, only two of the included studies were designed in accordance (Duncan et al., 2019; Callcott et al., 2015a). Although both studies demonstrated the superiority of a combined intervention to an FMS and/or PA only intervention, only one found the combined intervention to be superior to the academic control group (Duncan et al., 2019), while the other showed no difference between the two (Callcott et al., 2015a). Thus, more studies with the requisite design are needed before drawing definitive conclusions. When examining the effects of FMS and PA interventions, FMS interventions demonstrated greater effects on cognitive and academic skills compared to PA interventions (small-to-large vs. trivial-to-small effects). The superiority of FMS interventions might be explained by the fact that especially complex FMS tasks activate the same brain areas as complex cognitive tasks, and the co-activation in these brain areas is one potential explanation for the close associations between learning these skills (Diamond, 2000). In terms of PA, it is suggested that especially PA with higher duration and frequency can be effective on children’s cognitive and academic skills, albeit this evidence is based on a small number of studies with weak quality (Carson et al., 2016). It is also possible that higher PA intensity plays a role in this association (Carson et al., 2016) as it was demonstrated by Chang et al. (2013) reporting superior effects of moderateintensity in comparison to low-intensity training. However, further high-quality studies in this age group are required to confirm these findings. An important consideration is that in four of the included studies, the significant effects were limited to within-group comparisons despite the inclusion of a control group (Bedard et al., 2020; Stein et al., 2017; Lai et al., 2020; Beck et al., 2016). This raises a general concern in regard to the validity of the presented within-group effects in the current review as well as highlights the importance of applying a control-group design for establishing intervention effects. Moreover, only four of the included studies had sufficient statistical power to detect a medium effect size (Callcott et al., 2015a; Shoval et al., 2018; Have et al., 2018; S´ anchezL´ opez et al., 2019). In part, the lack of statistical power appeared to be the consequence of the absence of a priori power calculations; which only nine of the included studies reported having conducted (Have et al., 2018; Habibi et al., 2014; Stein et al., 2017; Raney et al., 2017; Gao et al., 2019; Jarraya et al., 2019; Lai et al., 2020; Wen et al., 2018; Xiong et al., 2019a). Interestingly, however, even within the aforementioned nine studies, only one was adequately powered (Have et al., 2018) and five stated low statistical power as a limitation of the study (Gao et al., 2019; Jarraya et al., 2019; Lai et al., 2020; Wen et al., 2018; Xiong et al., 2019a). This suggests that the concept of statistical power, similarly to other modern research fields (Dumas-Mallet et al., 2017), remains widely under-appreciated – and perhaps misunderstood – in the present field of research. This is an important consideration as underpowered studies may indicate substantially inflated effects as well as lead to nonsignificant results when a true effect is observed (La Caze and Duffull, 2011). Nonetheless, considering the unequivocal inability of the studies to detect small intervention effects, it is probable that the observed average effects in the present review are inflated compared to the true effects. Therefore, in order to improve the validity of scientific research, authors are strongly encouraged to ensure adequate statistical power by conducting a priori power analyses. Finally, most of the included intervention studies had a weak methodological quality and thus the results have to be dealt with caution. The finding is in line with previous systematic reviews (Carson et al., 2016). In general, low ratings resulted from inadequate reporting regarding several fundamental components such as relevant confounders, blinding, validity and reliability of the outcome measures, the agreement percentage of the selected participants as well as withdrawals and drop-outs. Previous systematic reviews with older children have found similar insufficiencies in reporting practices (Watson et al., 2017). In order to improve the quality of research evidence, adherence to published reporting guidelines for intervention studies (e.g., CONSORT 2010 Statement for RCT; Schulz et al., 2010) is highly encouraged for future studies. 4.1. Study limitations and strengths The present systematic review had numerous strengths. FMS and PA interventions as well as combined interventions were examined, included studies were conducted in 17 countries representing almost every continent, and all studies, except one, included both girls and boys. In addition, effect size calculations allowed the quantification and comparison of intervention effects across studies. However, this review also included some limitations. First, only studies published in English were included; making it possible that some relevant studies were missed. Further, the target population was limited to typically developing children and thus, future research should also cover children with special needs. 5. Conclusions Collectively, these results indicate that it may be possible to support typically developing preschoolers’ cognitive and academic learning with FMS and PA interventions. When comparing FMS and PA only interventions to combined interventions, the latter appeared to have larger effects. However, due to the lack of essential control groups, this notion remains unelucidated. In addition, FMS interventions were found to have larger effects compared to PA interventions. Nonetheless, most of the analysed studies had low methodological quality and thus these results should be considered with caution. This highlights the need for high quality intervention studies with representative and sufficient sample sizes, control groups, as well as validated data collection tools and adequate reporting practices. Supplementary data to this article can be found online at https://doi. org/10.1016/j.ypmed.2021.106948. Declaration of Competing Interest None. Funding sources had no involvement in the preparation, conducting nor in the decision to submit the article for publication. P. Jyl¨ anki et al.