Introduction
The benefits of physical exercise have been widely recognised both in the literature (e.g., McMorris et al.
2009) and the media (Leavy et al.
2011; Marcus et al.
1998). Its reported positive effects can be broadly classified into physical health (e.g., WHO
2010), behavioural (e.g., Sowa and Meulenbroek
2012), cognitive (e.g., Kramer and Erickson
2007), and psychosocial health or functioning (e.g., Netz et al.
2005). One specific focus of research has been on the relationship between exercise and cognitive functions, ranging from studies of children (e.g., Tomporowski et al.
2008), to young adults (e.g., Lambourne and Tomporowski
2010), geriatric populations (e.g., Kramer and Erickson
2007), and non-clinical (e.g., McMorris and Hale
2012) and clinical populations (e.g., Eggermont et al.
2009).
Most of the research conducted concerns the general population and has reported mostly positive effects of exercise on cognitive performance, mainly on executive functions (e.g., Kramer and Erickson
2007; Tomporowski et al.
2008). Two key points have emerged from this research. Firstly, the effect of exercise on cognition is likely to be selective (e.g., Kramer and Erickson
2007; Tomporowski et al.
2008), even within executive functions (EF). Indeed, a recent meta-analysis by Verburgh et al. (
2013), examined 24 studies (N = 944) on the effects of acute and chronic exercise on domains of EF in healthy individuals aged 6–35 years. The authors reported a larger effect size on inhibition (i.e., d = 0.46), followed by planning (i.e., d = 0.16), and working memory (i.e., d = 0.05). However, the authors highlighted that the working memory and planning results and the effects of chronic exercise on executive functions should be interpreted with caution due to the small number of studies investigating these areas. Similarly, another meta-analysis of 79 studies involving healthy populations (N = 2072) also found variability of exercise effects across various cognitive tasks reflecting different EF (Chang et al.
2012a). Secondly, the effect of exercise varies among individuals (e.g., Kramer and Erickson
2007; Tomporowski et al.
2008) and not every study reported cognitive improvements (e.g., Audiffren
2009; Kramer and Erickson
2007). For instance, it has been reported that individual differences such as fitness level moderate the magnitude of the effect of exercise on cognition; with greater improvements among those who are fitter (e.g., Chang et al.
2012a). Together, the two key points from the typical developing population literature has revealed the potential and limitations of using exercise in enhancing cognitive functions.
One of the implications provided by the literature from the typical developing population is that exercise may be particularly beneficial to individuals with learning difficulties (e.g., Fedewa and Ahn
2011; Sibley and Etnier
2003), such as those with attention deficit hyperactivity disorder (ADHD) or autism spectrum disorders (ASD). Unlike in the literature from the general population, however, the relationship between exercise and cognition in the neurodevelopmental population appears to be less clear, with studies on ASD samples focusing more on behavioural symptoms (e.g., Petrus et al.
2008; Sowa and Meulenbroek
2012) compared to ADHD studies that specifically examine EF (e.g., Chang et al.
2012b; Smith et al.
2013). This trend is not surprising as research in the general population regularly reveals specific improvements on EF following exercise (e.g., Tomporowski et al.
2008), and thus it has been theorised that exercise interventions may compensate for the impaired EF observed in individuals with ADHD (Wigal et al.
2013). Indeed, this hypothesis has been investigated in ADHD individuals but with inconclusive findings. For example, an earlier study conducted by Craft (
1983), found that 1–10 min of stationary cycling did not produce cognitive benefits on working memory performance in 31 children with ADHD or in comparison with 31 healthy children. In partial support of this, Chang et al. (
2012b) investigated the effects of running on a treadmill for 30 min on aspects of EF (i.e., inhibition and divided attention) in children with ADHD. Compared with participants in a sedentary control group (n = 20), the exercise group (n = 20) did not demonstrate greater EF performance, even though the authors reported post-exercise improvements from baseline scores in some aspects of EF tasks. In contrast to these studies though, Kang et al. (
2011) reported improved cognitive performance (i.e., divided attention and working memory) in children with ADHD following a series of aerobic exercises (n = 15) for 55 min compared to an educational control group (n = 13). Despite conflicting findings, ADHD studies generally report positive findings of the effects of exercise interventions on aspects of EF (e.g., Choi et al.
2015; Pontifex et al.
2013) but the efficacy remains unclear.
A recent meta-analysis of eight studies reviewed the effects of exercise on various symptoms of ADHD children and provided the first indication of the magnitude of the effect on cognition (Cerrillo-Urbina et al.
2015). The authors reported moderate to large effect sizes (standardised mean differences of 0.58 and 0.84) on measures of EF and attention, respectively. However, although the meta-analysis detected various levels of risk for publication bias among the included articles, the analysis conducted could not account for the existing bias (e.g., reporting bias). Additionally, the reported positive effect of exercise on global EF was too broad. Therefore, it is unknown which aspects of EF are specifically impacted by exercise. This is important as it has been acknowledged in the general population literature that not all EF improves following exercise intervention (Tomporowski et al.
2008; Verburgh et al.
2013). Furthermore, as inhibition has been found to be commonly impaired in ADHD (e.g., Pennington and Ozonoff
1996), it is imperative to investigate whether this specific domain and other aspects of EF are influenced by exercise. Nonetheless, these limitations in the meta-analysis are likely due to the limited number of papers available and the focus being more on the general ADHD symptomology than cognition per se.
The ADHD literature appears to focus on positive findings of exercise intervention on EF. In particular, papers either reported a beneficial effect of exercise on global EF (Cerrillo-Urbina et al.
2015) or only statistically significant findings (e.g., Choi et al.
2015; Pontifex et al.
2013); and in those studies where non-improvements were acknowledged, this was attributed to the lack of task-sensitivity (e.g., Gapin et al.
2015; Pan et al.
2015). Although it is possible that the cognitive tasks used in the studies may be insensitive in detecting subtle cognitive changes, what is also possible or more likely based on the larger number of studies conducted on the general population, is that the effect of exercise is specific to some aspects of EF rather than to cognition in general and is evident in some individuals more than others (e.g., Tomporowski et al.
2008).
In terms of the ASD literature, what appears to be missing is an investigation of the effects of exercise on EF, as there have been suggestions that aspects of EF such as planning and set-shifting are impaired in ASD individuals (e.g., Pennington and Ozonoff
1996). To date, only one study has examined this issue. Anderson-Hanley et al. (
2011) administered a single bout of exercise through the use of a gaming system (i.e., exergaming) for 20 min in 22 adolescents with ASD, to investigate the effect of exercise on aspects of EF and stereotyped behaviours. Compared with a sedentary control condition, post-exergaming participants demonstrated improvements in an EF task reflecting working memory but the results were less clear on set-shifting and inhibition. However, the authors did not report the findings of all the EF tasks involved in the study, further limiting the conclusions regarding the effect of exercise on EF in this population.
The gaps identified here in the literature may have important implications, particularly for the selection of appropriate interventions, as the symptoms of ASD and ADHD often do not appear in isolation (Gargaro et al.
2011; Joshi et al.
2014), even though non-comorbid cases also exist (e.g., Chantiluke et al.,
2014). Furthermore, given that comorbidity between the two disorders is reported to be as high as over 80 % of the cases (e.g., Joshi et al.
2010; Mukaddes et al.
2010), it is not known if exercise can be used as an effective intervention in improving aspects of cognitive functions in these clinical populations (i.e., comorbid and non-comorbid individuals). Although it is still contentious as to whether the two disorders are a common or unique subtype that falls under the domain of either ASD or ADHD exclusively (e.g., Chantiluke et al.
2014), there is a general agreement that symptom overlap exists even if individuals may not fulfil the diagnostic criteria of both disorders. Therefore, it has been recommended that the uniqueness and co-morbid characteristics between both disorders be acknowledged (Gargaro et al.
2011). This review attempts to study the efficacy of exercise intervention on cognition in both disorders, concurrently and separately.
In terms of executive dysfunction, some specific EF domains are generally reported to be impaired in individuals with ASD and ADHD, when compared separately to healthy individuals. In ASD individuals, EF domains including aspects of planning (e.g., Chen et al.
2016; Hill
2004), set-shifting and working memory (e.g., Chen et al.
2016; Andersen et al.
2015) are usually impaired. Conversely, deficits are commonly found in aspects of inhibition (e.g., Willcutt et al.
2005) and working memory (e.g., Schreiber et al.
2014) in individuals with ADHD. There are also findings that EF in general tend to be more impaired in individuals with ASD than those with ADHD (e.g., Corbett et al.
2009; Goldberg et al.
2005; Pennington and Ozonoff
1996). It is noteworthy that there are inconsistencies within ASD and ADHD literature on which aspects of EF in these individuals are impaired or intact, which is beyond the scope of this review. Nevertheless, studies that investigated EF between individuals with ASD and ADHD have demonstrated the difficulty in establishing a distinct executive dysfunction profile (e.g., Corbett et al.
2009; Geurts et al.
2004; Goldberg et al.
2005); in that EF deficits can sometimes overlap in both disorders such as aspects of working memory, sustained attention and even inhibition. Considering that the literature in the healthy population generally reported exercise benefits on EF (e.g., Kramer and Erickson
2007; Tomporowski et al.
2008), especially on inhibition and probably other aspects of EF (Verburgh et al.
2013), exercise may have the potential to be used as an intervention that targets EF deficits in individuals with ASD and/or ADHD.
In summary, the existing literature has provided valuable information on the effects of exercise on cognition in individuals with ASD/ADHD. Nevertheless, what seems to be lacking is also a practical interpretation of effect size (Ellis
2010) that would be useful for clinicians and parents. The purpose of this review is to investigate the efficacy of exercise intervention on individuals with ASD/ADHD, and explore the practical significance of applying exercise to cognition based on the meta-analytic findings. To the authors’ knowledge, this is the first meta-analysis that reviews the relationship between exercise and cognition, in both ASD and ADHD populations.
Limitations/Future Directions
The present meta-analysis bridges the gap in the literature regarding the exercise and cognition relationship in the ASD/ADHD and typical developing populations. Although the overall findings are encouraging, it should be regarded as a tentative conclusion in guiding future research and future large-scale randomised-controlled trials would be required to validate the current findings. Moreover, the findings should also be interpreted in the context of the existing limitations. Firstly, the number of articles available and included in this review (i.e., 22 studies) is still considered relatively small compared to other meta-analyses in the typical developing population literature. In particular, the number of ASD studies examining exercise-cognition relationship is very limited (i.e., six studies), especially on EF. Secondly, five studies (one ASD and four ADHD articles) did not report non-significant results. This reporting bias is consistent with the results of the sensitivity analysis. As mentioned earlier regarding the limitations of exercise interventions, future studies should try to include at least some basic information on non-significant findings (e.g., effect size, means, and standard deviations). Furthermore, similar to drug trials, it would be informative to also report the number of participants that show improvements in their cognition. This is because most studies in the literature only rely on the mean statistics and it may be possible that the number of individuals that truly improve in cognition after exercise interventions is low but the mean score of the experimental group/condition is high or significantly different to the control group/condition (see Speelman and McGann
2013 regarding limitations of the mean). Thirdly, variability in individual differences such as developmental and fitness levels, and the intensity of exercise interventions are identified in the literature as important moderators in understanding the exercise-cognition relationship (e.g., Chang et al.
2012a; Kramer and Erickson
2007; Tomporowski et al.
2008); however, 18 studies did not report IQ levels and 12 studies did not provide quantification of their exercise interventions (e.g., heart rate, oxygen consumption). Therefore, these factors could not be examined in this meta-analysis. Fourthly, the current results could only support the use of exercise in enhancing aspects of cognition for individuals up to 25 years of age. Therefore, it is unknown if exercise interventions are equally effective in older age groups with ASD/ADHD. Lastly, another limitation is the inherent issues within EF tasks. Due to the complex nature of EF, an EF task is usually unable to provide an isolated measure of a specific cognitive process that it intends to measure and invariably also captures other aspects of EF and non-EF processes (e.g., Pennington and Ozonoff
1996; Suchy
2009). Moreover, most studies included in the meta-analysis examined each EF domain with a single neuropsychological task, making it difficult to ascertain whether an improvement (or not) on a single task after exercise intervention is indeed a reflection of an actual change in that particular EF. In addition, recent literature has supported EF as a multifaceted latent construct where various domains are separable but interdependent (e.g., Cassidy
2016). Thus, future experimental studies could consider using a number of neuropsychological tasks to assess each EF domain (Ziereis and Jansen
2015) and if possible, to include a range of EF measures (see e.g., Smith et al.
2013). Although the present meta-analysis is unable to evaluate the exercise effect on other EF such as planning, sustained attention and working memory due to limited number of papers examining these areas; nevertheless, various neuropsychological tasks used in the studies (refer to “
Appendix”) are combined to evaluate the effect of exercise intervention on inhibition, set-shifting and memory functions (including working memory). Specifically, there is support that exercise benefits inhibitory function in individuals with ADHD.
Clinical Implications: Binomial Effect Size Display
A major issue that has been overlooked in the literature is the practical interpretation of what effect size actually means in the context of applying exercise interventions in improving cognitive functions. The binomial effect size display (BESD) by Rosenthal and Rubin (
1982) enables the interpretation of effect size in meaningful terms: the estimated percentage of individuals that improved aspects of their cognition by exercise interventions. Notwithstanding the limitations, the conversion of overall
r to BESD (Table
7) demonstrates that relative to control groups or baseline measures, overall, following exercise interventions, 61.75 % of the individuals with ASD and ADHD improved on aspects of their cognitive performance. Specifically, exercise benefited 76.30 % of ASD individuals by enhancing their on-task behaviour and performance on simple learning task. For ADHD individuals, 59.05% reported cognitive benefits on aspects of their EF after exercising; especially on inhibitory control (i.e., 58.70 %). Based on Table
7, it is evident that exercise interventions mostly only account for a small variance on cognitive improvements, supporting the consensus in the literature that the relationship between exercise and cognition is complex and is moderated by many other factors such as individual differences (e.g., Chang et al.
2012a; Tomporowski et al.
2008). Nonetheless, it should be noted that the implications of utilising exercise interventions with individuals with ASD and/or ADHD should not be underestimated, as it is likely to be particularly useful in addressing areas of their cognitive function in which they typically do not perform well (e.g., inhibition). Furthermore, apart from exercise-induced improvement in aspects of EF, which is likely to be related to neurobiological pathways such as catecholamine (see e.g., Wigal et al.
2013), neurotrophic and growth factors (see e.g., Ratey and Loehr
2011); indirect pathways including elevated self-efficacy, mood, and other psychosocial functioning (e.g., Davis and Lambourne
2009; Tan et al.
2013b; Tomporowski et al.
2011) may also be beneficial to individuals with ASD and/or ADHD.
Table 7
Binomial effect size display (BESD): efficacy of post-exercise interventions on cognition
ASD/ADHD | .235 | 61.75 | 38.25 | .06 |
ASD (on-task/learning task) | .526 | 76.30 | 23.70 | .28 |
ADHD (executive function) | .181 | 59.05 | 40.95 | .03 |
ADHD (inhibition) | .174 | 58.70 | 41.30 | .03 |
ADHD (memory) | .286 | 64.30 | 35.70 | .08 |