Published on 18 September 2026
Neurodevelopmental disorders: does the type of physical activity matter more than the amount?
The essentials
This network meta-analysis pools 31 randomized trials, 66 arms and 1,403 children and adolescents aged 5 to 17 with a neurodevelopmental disorder. It compares four categories of physical activity (aerobic exercise, exergaming, mind-body exercise and multi-component physical activity) within a seven-node network that also includes three comparators: relaxation techniques, neurofeedback and usual care. Mind-body exercise, meaning yoga, qigong, tai chi and body-based interventions with a meditative component, ranks first for attention, with a standardized mean difference of 1.91 (95% CI 0.57 to 3.25) versus usual care, ahead of exergaming at 1.58 (95% CI 0.14 to 3.02). Multi-component physical activity, which combines at least two types of activity such as racket sports, reaches 0.79 (95% CI 0.48 to 1.11) for executive functions. Aerobic exercise is not statistically significant in any of the three cognitive domains, and it is less well accepted than usual care. In the attention-deficit/hyperactivity disorder subgroup, multi-component physical activity is the only modality significant across all three domains at once. The authors state the caveats themselves: very low to low certainty for every comparison, small trials, very wide intervals and a sample that is 83.2% boys.
Context
Physical activity appears in almost every care plan in child and adolescent psychiatry, and it often appears in a vague form: move more. The question asked here is sharper and more useful in the consulting room. For the same amount of exercise time, does content that demands coordination, anticipation or sustained attention do something that purely cardiovascular content does not?
The underlying hypothesis is that cognitive engagement during exertion recruits circuits that exertion alone does not. It is plausible and has already been discussed in other populations. This study tests it in a network that combines the available direct comparisons between modalities with indirect comparisons reconstructed through shared comparators, which makes it possible to rank interventions that are rarely pitted against each other within the same trial.
The study at a glance
| Question, population, outcome | |
|---|---|
| Population | |
| Children and adolescents aged 5 to 17 with a diagnosed neurodevelopmental disorder. Attention-deficit/hyperactivity disorder in 18 of the 31 trials, as well as autism spectrum disorder, developmental coordination disorder and specific learning disorder. 1,403 participants, mean age 10.0 ± 1.9 years, 83.2% boys. | |
| Interventions | |
| Four categories of physical activity: aerobic exercise, exergaming, mind-body exercise (yoga, qigong, tai chi, body-based interventions with a meditative component) and multi-component physical activity (a combination of at least two types of activity, for example racket sports). Multi-component physical activity is the most represented modality, in 19 trials. Doses and durations vary from one trial to another. | |
| Comparators | |
| Three comparator nodes: usual care (no treatment, waiting list or activities that do not fit the other nodes, for example educational activities), relaxation techniques and neurofeedback. The estimates reported here are those against usual care. | |
| Outcomes | |
| Primary outcomes: three cognitive domains measured immediately after the intervention, namely attention, memory and executive functions. Secondary outcome: acceptability, defined as the proportion of participants who completed the assigned treatment. No behavioral outcome and no measure of processing speed are analyzed. | |
| Design | |
| Systematic review and frequentist network meta-analysis, 31 randomized trials, 66 arms, search of seven databases up to September 4, 2023, standardized mean differences, random-effects models, ranking by surface under the cumulative ranking curve, certainty graded with the CINeMA framework, protocol registered on PROSPERO (CRD42023409606). |
Quality control
| Item checked | Judgment |
|---|---|
| Protocol registered before analysis | Yes, PROSPERO CRD42023409606 |
| FindingReported as conducted according to the PRISMA statement and its extension for network meta-analyses. Transitivity examined, global and local inconsistency tested, sensitivity analyses, meta-regressions and subgroup analyses performed. | |
| Certainty of the results | Very low to low |
| FindingUnfavorable CINeMA grading for attention, memory and executive functions. The authors indicate that downgrading is driven mainly by within-study risk of bias, imprecision and heterogeneity. Point estimates, even large ones, remain uncertain. | |
| Risk of bias of included trials | Moderate for 26 trials |
| FindingTwenty-six trials are rated as having some concerns and five as high risk. The authors cite the open-label nature of the trials, unblinded assessors and imperfect handling of missing data. Blinding participants is impossible in any case for this kind of intervention. The domain-by-domain detail is in a supplementary file that was not attached to the text reviewed. | |
| Network density | Sparse network |
| FindingThe authors themselves report poorly connected networks and estimates underpowered to detect some differences. The attention network rests on 10 trials and 441 participants, the memory network on 11 trials and 484 participants, the executive function network on 26 trials and 1,003 participants. | |
| Sample composition | 83.2% boys |
| FindingThe data do not support extrapolation to girls, and the authors acknowledge this. The limitations section of the publication gives a proportion of boys of 79.9%, which differs from the 83.2% reported in the results, and the discrepancy is not explained. Attributing this imbalance to underdiagnosis in girls is a matter of opinion and does not appear in the publication. | |
| Geographic origin of the trials | Twelve countries and regions |
| FindingThe trials come from Iran, Taiwan, Hong Kong, South Korea, China, Switzerland, Italy, the Netherlands, the United States, Brazil, Tunisia and Egypt. Recruitment therefore does not come from a single cultural setting, but no trial was conducted in France, and school and sports supervision differ from one country to another. | |
| Heterogeneity | Moderate, τ² from 0.08 to 0.68 |
| FindingThe authors themselves describe this heterogeneity as moderate. Cognitive test batteries vary widely between trials, which weighs on pooling by domain. Funnel plot asymmetry and Egger’s test suggest publication bias for executive functions, but not for attention or memory. | |
Results
| Outcome | Value |
|---|---|
| Mind-body exercise, attention | 1.91 (95% CI 0.57 to 3.25) |
| ReadingLargest effect in the network and top ranking, with a surface under the cumulative ranking curve of 83.4%. The interval is very wide. An effect size of this magnitude in a cognitive domain calls for caution in itself. | |
| Exergaming, attention | 1.58 (95% CI 0.14 to 3.02) |
| ReadingA significant result, but the lower bound is close to zero. In the attention-deficit disorder subgroup, this superiority over usual care is no longer found, for either attention or memory. | |
| Exergaming, memory | 0.97 (95% CI 0.10 to 1.84) |
| ReadingThe only significant comparison versus usual care for memory when all disorders are pooled, with the best ranking in this domain, 88.0%. | |
| Multi-component physical activity, executive functions | 0.79 (95% CI 0.48 to 1.11) |
| ReadingA more modest effect and the most precise estimate in the network. In the same domain, exergaming, 0.94 (95% CI 0.42 to 1.45), and mind-body exercise, 0.92 (95% CI 0.21 to 1.62), are also superior to usual care. | |
| Aerobic exercise | No significant difference in any of the three domains |
| ReadingAttention 0.76 (95% CI -0.46 to 1.99), memory 0.01 (95% CI -0.68 to 0.69), executive functions 0.15 (95% CI -0.39 to 0.70). This is an absence of detected difference, not a demonstration that there is no effect, in a network the authors describe as underpowered. Two modalities do, however, perform significantly better than aerobic exercise on executive functions: exergaming, 0.78 (95% CI 0.12 to 1.45), and multi-component physical activity, 0.64 (95% CI 0.11 to 1.18). | |
| Attention-deficit disorder subgroup | Multi-component physical activity significant in all three domains |
| ReadingAttention 1.87 (95% CI 0.61 to 3.14), memory 0.81 (95% CI 0.24 to 1.38), executive functions within a set of differences ranging from 0.57 to 1.82. Mind-body exercise is significant for attention, 2.00 (95% CI 0.89 to 3.10), and for executive functions, but not for memory. These estimates come from supplementary files that were not attached to the text reviewed and are reported as the authors cite them in the body of the article. | |
| Acceptability | Aerobic exercise less well accepted than usual care |
| ReadingOdds ratio 0.38 (95% CI 0.15 to 0.97). No difference in dropout is observed between the other modalities and usual care. The only unfavorable signal in the study therefore concerns adherence to aerobic exercise alone, not its tolerability. | |
Critical appraisal
| Domain | Judgment |
|---|---|
| Consistency between methods and conclusion | Respected |
| FindingThe authors openly state the low certainty of their results, write that their findings suggest effectiveness without presenting it as established, and call for head-to-head trials. The temptation to quote the effect sizes without that caveat comes from the reader, not from the source. | |
| Size of the effects on attention | Unlikely |
| FindingEditorial opinion: a standardized difference close to two for attention exceeds what most documented behavioral interventions achieve. The trials involved are small and outcome assessment is not consistently blinded, which is enough to explain part of the observed magnitude. | |
| Network consistency | No inconsistency detected |
| FindingGlobal and local inconsistency were tested without revealing any disagreement between direct and indirect evidence. This test has low power in a sparse network, however: it does not demonstrate that no disagreement exists. | |
| Clinical relevance of the outcome | Intermediate |
| FindingCognitive tests are intermediate measures, recorded immediately after the intervention. No school, independence or quality-of-life outcome is assessed, and no longer-term follow-up is reported. | |
| Benefit-risk balance | Favorable |
| FindingEditorial opinion: the interventions studied carry little risk and have value in their own right. Uncertainty about the size of the cognitive benefit is not a reason to forgo them, provided nothing is promised on the cognitive side. | |
Level of evidence
A network meta-analysis of randomized trials, level 1a on the Oxford scale for the rank of the design. That rank should not hide the very low to low certainty that the authors themselves assign to all of their comparisons. What is demonstrated rests on two direct and significant comparisons for executive functions: exergaming and multi-component physical activity outperform aerobic exercise alone. What is suggested, with very low to low certainty, is the ranking between modalities and every numerical value. Expert opinion covers the idea that cognitive engagement during exertion explains this gap, as well as extrapolation to a girl followed in France, which no data from this study support.
The colleague test
What an experienced colleague would say if you walked them through this study in two minutes, between two appointments.
“An effect size of 1.91 on attention, with an interval running from 0.57 to 3.25, in trials where nobody can be blinded? I don’t take that at face value. But the idea that running laps doesn’t give a child the same thing as an activity that makes them think, that one I’m willing to keep.”
In practice: steer activity advice toward content that demands coordination, anticipation or attention rather than toward sheer volume of effort, without promising any quantified cognitive gain.
What you can do with this
- When you recommend an activity, specify the content and not just the frequency: racket sports, martial arts, ball games, motor skills circuits or mind-body practices, rather than running alone. This is the kind of content found in the highest-ranked trials.
- Do not present this activity as a treatment for the disorder or as an alternative to the care already in place. Nothing here supports that kind of substitution, since no trial compared physical activity with medication.
- Be upfront with the family about the uncertainty: the direction of the result is plausible, its size is not. A quantified promise would be contradicted by the width of the confidence intervals themselves.
- Take into account what the family can realistically put in place. In this study, the only modality less well accepted than usual care is aerobic exercise alone, which matches the clinical observation that this type of activity tends to feel monotonous.
- For a girl, consider the question still open, since the sample is overwhelmingly male.
Frequently asked questions
Can physical activity replace medication in ADHD?
This study does not test that and cannot support the claim. Trials combining physical activity with medication or psychological treatment were in fact excluded. The study compares activity modalities with one another, with relaxation techniques, with neurofeedback and with usual care.
Why does aerobic exercise alone not stand out?
No significant difference is detected versus usual care for attention, memory or executive functions. This may reflect a true absence of effect, a lack of power or an ill-suited measure, and the study cannot tell which. Two other modalities, however, are significantly superior to it for executive functions, which is a positive comparative finding and not merely an absence of effect.
Is exergaming really useful?
There is a signal for attention, memory and executive functions when all disorders are pooled, with very low to low certainty. It disappears for attention and memory as soon as the analysis is restricted to attention-deficit disorder. Offer it as an option, not as a recommendation.
Is a standardized difference of 1.91 a large effect?
It would be considerable, given that a value of 0.8 is already classed as large under the convention the authors use. That is precisely why caution is warranted, all the more so because the associated confidence interval runs from 0.57 to 3.25. Values like this often come from small trials without blinded assessment.
Do these results apply to patients in France?
With reservations. Five trials were conducted in Europe, in Switzerland, Italy and the Netherlands, but none in France, and most trials come from Asia and the Middle East. School settings, supervision and activity habits differ.
Annotated bibliography
Source study. Tao R, Yang Y, Wilson M, Chang JR, Liu C, Sit CHP. Comparative effectiveness of physical activity interventions on cognitive functions in children and adolescents with Neurodevelopmental Disorders: a systematic review and network meta-analysis of randomized controlled trials. International Journal of Behavioral Nutrition and Physical Activity, 2025;22(1):6. DOI 10.1186/s12966-024-01702-7. PMID 39806448.
Protocol registration. PROSPERO, CRD42023409606.
Funding. Supported by the Research Grants Council of Hong Kong, General Research Fund (14612122) and Research Impact Fund (R4035-22F).
Conflicts of interest. Professor Cindy H. P. Sit is a member of the editorial board of the International Journal of Behavioral Nutrition and Physical Activity and took no part in the peer review of, or decisions about, this manuscript. The authors declare no other conflicts of interest.
What was reviewed. Verification was carried out on the full text of the publication, including Tables 1 to 5 and Figures 1 and 2, as well as the funding, conflicts of interest and reference sections. The bibliographic record, DOI and PMID, was checked against the PubMed entry. The sixteen supplementary files cited by the authors, including the search strategy, the trial-by-trial risk of bias assessment, the rankings by surface under the curve and the results by type of disorder, are not included in the document reviewed, which contains only their download link: figures drawn from these files are reported here solely as the authors cite them in the body of the article and could not be checked at source. This article underwent an independent double reading.