Published on 19 September 2026
In ADHD, does the type of physical activity change the cognitive benefit?
The essentials
Twenty-one randomized trials, 915 children with ADHD: structured physical activity programs improve performance on executive function tests compared with control groups that were, in most trials, left without any structured activity at all. The clearest effect concerns inhibitory control after activities the authors label as open, with a standardized effect of 0.75. So-called closed activities outperform open ones on cognitive flexibility and working memory, with more modest effects of 0.61 and 0.45. The practical message that has spread from this work, that every difficulty has a matching sport, does not survive scrutiny of the design: no child was ever randomized between judo and jump rope. Modalities are compared across different subgroups of trials, each set against its own control condition, and no formal test of difference between subgroups is reported. Finally, the outcomes are performances on cognitive tests, not ADHD symptoms or school functioning. What holds up here is the general signal in favor of regular activity. What does not hold up is the ranking between disciplines.
Context
Parents raise this question in consultation, almost always in the same terms: “we were told to get him into sports, but which one?” The request is a reasonable one. Physical activity is accessible, inexpensive, free of notable adverse effects when properly adapted, and it is recommended for every child regardless of ADHD status. For some families, it is also a way to begin management without turning straight to medication.
Until now, the literature supported a general statement: moving around improves certain cognitive performances in children with ADHD. The work analyzed here tries to go further by classifying activities according to how predictable they are, and by asking whether this classification predicts which executive domain improves. It is this promise of differentiation that makes the article interesting, and it is exactly what calls for the most caution.
Mechanism
What the open/closed activity distinction assumes
The hypothesis is cognitive before it is biological. A so-called open activity unfolds in a changing, unpredictable environment that forces the child to inhibit a prepared response, redirect attention, and decide quickly. A so-called closed activity unfolds in a stable environment, built around a repeated, self-paced movement. The assumption is that the former places greater demands on online executive control, and that this repeated demand transfers to laboratory tests. This is a plausible hypothesis. It is not demonstrated by this work, which measured neither the actual cognitive load of the sessions nor any mediator at all.
| Element of the hypothesis | Status in this work |
|---|---|
| Executive demand assumed by the modality | Postulated, not tested |
| FindingThe modality is identified by the name of the discipline, not by a measurement of what it actually demands of the child. Two sessions carrying the same label can place very different cognitive demands depending on how they are run. | |
| Boundary between the two categories | Set by the authors |
| FindingNine trials are classified as closed activities, running-type aerobic exercise, swimming, jump rope and Baduanjin, and twelve as open activities, table tennis, football, basketball, taekwondo, judo, inline skating, structured play, exergaming and combined exercise programs. The classification is not even stable within the publication itself: the characteristics table places the football trial among open activities and one aerobic-exercise trial among closed activities, while the results text does the opposite. Moving a discipline from one category to the other shifts the effect attributed to that category. | |
| Dose: duration and intensity | Approximated through subgroups |
| FindingThe parameters associated with the clearest effects, at least six weeks and moderate-to-vigorous intensity, come from comparisons between subgroups of trials, not from a randomization of dose. On cognitive flexibility, the two intensity levels even yield exactly the same standardized effect, 0.54. | |
| Biological mediators | Not measured in this work |
| FindingThe only data extracted from the trials are performances on executive tests: no measure of physical fitness, imaging, or biological marker. Dopaminergic pathways and BDNF are mentioned only in the discussion, drawing on other studies. This absence is an absence of measurement: it says nothing, in either direction, about whether a mediator exists. | |
The study at a glance
| What the study did | |
|---|---|
| Population | |
| DetailChildren and adolescents aged 6 to 18 years with a clinical diagnosis of ADHD, with trial mean ages ranging from about 7.7 to 14.5 years. Neither the proportion of children on medication, nor ADHD subtypes, nor sex distribution were analyzed, a gap the authors acknowledge themselves. Thirteen of the twenty-one trials were conducted in China, the others in Switzerland, Tunisia, the United States and Iran. | |
| Intervention | |
| DetailPhysical activity programs, all aerobic in nature, classified by the authors into open and closed activities, with sub-analyses contrasting single aerobic exercise against multi-component programs, then five families of disciplines. | |
| Comparator | |
| DetailHeterogeneous control conditions: waiting list, no intervention, sedentary activities such as watching videos, but also standard physical education classes, art workshops and sport games. No trial pitted two physical activity modalities against each other, and none of the described control conditions was a medication. | |
| Outcomes | |
| DetailExecutive functions measured by tests, inhibitory control, cognitive flexibility and working memory, using eighteen different instruments, pooled as standardized mean differences. No ADHD symptom scale, no school-performance indicator, no measure of daily functioning. | |
| Design and sample size | |
| DetailSystematic review and meta-analysis of 21 randomized trials, 915 children in total, fewer than fifty children per trial on average. Analyzed sample sizes differ by domain: 18 trials and 763 children for inhibitory control, 11 trials and 503 children for cognitive flexibility, 13 trials and 578 children for working memory. Protocol registered on PROSPERO (CRD420251080013), risk of bias assessed with RoB 2, pooling performed in RevMan 5.4. | |
| Bibliographic sources | |
| DetailSix databases searched through April 20, 2026: PubMed, Cochrane Library, Embase, Web of Science, CNKI and Wanfang. 1,500 records identified, 472 duplicates removed, 21 trials retained. Search restricted to English and Chinese, which excludes other literatures. |
Quality control
| Point checked | Judgment |
|---|---|
| Protocol registration | Registered, then amended |
| FindingThe initial submission is dated June 24, 2025, before the literature search, which is the right order. But the registration was revised on May 14, 2026, after the search closed on April 20, 2026, and this revision added, after the fact, the final title, the restriction to English- and Chinese-language publications, and the search date boundaries. The deposited protocol also stated that no quantitative synthesis was planned: the published meta-analysis therefore departs from what was registered. | |
| Risk-of-bias assessment | One trial at high risk |
| FindingRoB 2 is the reference tool for randomized trials. Twenty-four percent of trials are rated low risk, 71% some concerns, and 5%, one trial, high risk. The main reservations concern the randomization process, rarely described. No sensitivity analysis excludes the weakest trials: robustness was tested only by removing trials one at a time. | |
| Heterogeneity between trials | Low to moderate, reported |
| FindingContrary to what the diversity of programs might suggest, heterogeneity is low: I-squared of 38% for inhibitory control, 0% for cognitive flexibility, 23% for working memory, with a fixed-effects model in all three cases. The authors nonetheless call 38% heterogeneity “significant” and 0% heterogeneity “moderate”, which does not help the reader. Some subgroups remain highly dispersed, up to 85% for ball sports. | |
| Certainty of the evidence | No formal assessment |
| FindingNo GRADE-type confidence rating accompanies the estimates. This absence was announced in advance: the registered protocol explicitly stated that the certainty of the results would not be assessed. The reader therefore receives figures without the degree of confidence the authors themselves assign to them. | |
| Comparator in the included trials | Heterogeneous, rarely active |
| FindingComparing a supervised program with a waiting list or video watching measures the whole package: the exercise, but also adult time, structure, the group, and family expectation. A few trials did, however, offer the control group physical education classes, art workshops or sport games, which makes the control condition uneven from one trial to the next and makes effects less directly comparable across subgroups. | |
| Independence and funding | Public funding declared |
| FindingThe work is funded by the National Social Science Fund of China, project 23BTY078, and by a research fund at Fudan University. The authors declare no commercial or financial relationship constituting a conflict of interest, which is consistent with a non-pharmacological topic. The PROSPERO registration, by contrast, stated that the review had no external funding. | |
Results
Across all trials, without distinction by modality, physical activity improves all three executive domains, with low heterogeneity.
| Executive domain | Trials | Standardized effect, 95% CI | I-squared |
|---|---|---|---|
| Inhibitory control | 18 | -0.69 [-0.84, -0.54] | 38% |
| Cognitive flexibility | 11 | -0.53 [-0.71, -0.35] | 0% |
| Working memory | 13 | -0.43 [-0.59, -0.26] | 23% |
The ranking of modalities that made this work’s reputation appears once these same trials are split into subgroups. Placed side by side, the intervals overlap substantially.
| Modality and executive domain | Trials | Standardized effect | 95% CI |
|---|---|---|---|
| Open activities, inhibitory control | 10 | -0.75 | [-0.97, -0.52] |
| Closed activities, inhibitory control | 8 | -0.64 | [-0.84, -0.45] |
| Closed activities, cognitive flexibility | 6 | -0.61 | [-0.84, -0.38] |
| Open activities, cognitive flexibility | 5 | -0.40 | [-0.69, -0.12] |
| Closed activities, working memory | 6 | -0.45 | [-0.68, -0.23] |
| Open activities, working memory | 7 | -0.40 | [-0.65, -0.15] |
Negative values correspond, under the convention adopted by the authors, to a result favoring physical activity: measures where a higher score means worse performance, reaction time or error count, were reverse-coded before pooling. The intervals exclude zero: the existence of an average effect on these performances is consistent from one domain to another. Their width, however, is a reminder that the size of this effect remains imprecise.
Subgroup analyses place the clearest results with programs lasting at least six weeks and delivered at moderate-to-vigorous intensity. This order of magnitude is consistent with general physical activity recommendations for children, which is reassuring, but it does not come from a randomized comparison of durations or intensities, and the superiority of the higher intensity disappears for cognitive flexibility, where the two levels give the same figure.
The decisive point comes down to one sentence: the modalities were never compared against each other. Each activity category was tested against its own control condition, in separate trials, and the results of those trials were then pooled by category. The apparent difference between open and closed activities is therefore an indirect comparison between subgroups of trials, exposed to everything else that distinguishes those trials: age, severity, ongoing treatment, tests used, country, quality of supervision. No formal test of difference between subgroups is reported either, and the intervals overlap almost entirely, 0.45 versus 0.40 for working memory. A subgroup is not a trial. Until a trial randomizes children between two disciplines, the ranking between modalities remains a hypothesis to test, not a result to apply.
Critical appraisal
| Domain | Judgment |
|---|---|
| Comparison between modalities | Indirect, between subgroups |
| FindingThis is the main limitation, and it bears precisely on what makes the article interesting. Randomization protects the comparison between exercise and no exercise, not the comparison between two types of exercise. No formal test of difference between subgroups is reported. | |
| Internal consistency of the publication | Abstract and tables disagree |
| FindingThe abstract reports a standardized effect of 0.81 on inhibitory control for open activities, while the results table gives 0.75 with the same confidence interval, the only value compatible with that interval. The conclusion further attributes the best effect on cognitive flexibility to ball sports, whereas the abstract and the table point to running and cycling. The figures used here are those from the results tables. | |
| Nature of the outcome | Surrogate outcome |
| FindingA performance on an executive test is not a symptom, a grade, or homework handed in. Transfer from the laboratory to the classroom is not established here, and the recent history of cognitive remediation in ADHD counsels against presuming it. | |
| Blinding and the effect of the setting | Blinding not possible |
| FindingChildren and families know which group they are in. The review does not report whether test assessors were blinded in the included trials: the corresponding RoB 2 domain, measurement of the outcome, is judged without major concern in most trials, which is not the same as a documented blinding procedure. Against controls left to their usual activities, part of the measured effect belongs to the attention received and to familiarization with the tests, not to the exercise itself. | |
| Size of the included trials | Small trials |
| FindingFewer than fifty children per trial on average, with the smallest trial including only sixteen. Small trials produce scattered estimates and, when published selectively, pull the average upward. | |
| Classification of activities | Debatable boundary |
| FindingHow disciplines are assigned to categories is a matter of judgment, and that judgment drives the result. Two trials even switch sides between the characteristics table and the results text. Any practical conclusion built on a discipline’s label inherits this fragility. | |
| Risk of publication bias | Dominant regional literature |
| FindingEgger’s test does not detect asymmetry, with p-values of 0.103 for inhibitory control, 0.769 for cognitive flexibility and 0.299 for working memory. This result is only moderately reassuring: the test has low power over about a dozen trials, and thirteen of the twenty-one trials come from the same country, whose exercise literature is thought to publish null results less often. | |
| Precision of the estimates | Wide intervals |
| FindingThe direction of effect is consistent and overall heterogeneity is low, but subgroup intervals remain wide, up to [-1.80, -0.15] for the four ball-sport trials on inhibitory control, with an I-squared of 85%. These are precisely the subgroups on which discipline-specific recommendations rest. | |
Level of evidence
The design corresponds to CEBM level 1a, a systematic review of randomized trials with a registered protocol. This level qualifies the architecture of the work, not the certainty of each of its conclusions, and the gap between the two is what matters most here.
What is reasonably established: in children with ADHD, a supervised physical activity program is associated with improved performance on executive function tests compared with control conditions, with standardized effects of 0.69 for inhibitory control, 0.53 for cognitive flexibility and 0.43 for working memory. The direction of effect is consistent across all three domains and withstands the removal of any single trial.
What is only suggested: that certain modalities act preferentially on certain executive domains, and that a duration of at least six weeks at moderate-to-vigorous intensity is preferable. These are hypotheses generated by subgroup comparisons, without a formal test of difference, consistent with a plausible cognitive rationale, but untested by any trial.
What is not addressed: the effect on ADHD symptoms, on schooling, on family life, the durability of gains after the program ends, the influence of ongoing medication and of ADHD subtypes, and any comparison with existing treatments. The study’s silence on these points is not a negative result: it is an absence of analysis.
The colleague test
What an experienced colleague would say if shown this study in two minutes, between two consultations.
That exercise does these children good is something already believed and already recommended, so nothing changes there. What would not happen is telling a mother that judo trains inhibition and jump rope trains working memory. No one was randomized between the two; these are different trials being compared. And these are tests, not homework. The advice stays a regular activity, one the child will stick with for two months, with nothing else in the care plan touched.
Translation for practice: advising regular physical activity is legitimate, useful for general health, and safe when properly adapted. The choice of discipline, however, should be decided by the child’s enjoyment and consistency, not by a chart matching a sport to a cognitive function.
What you can do with this
- Recommend a regular, sustained activity. The parameter best supported by this work is persistence: at least six weeks, at least two sessions a week, at moderate-to-vigorous intensity, in a supervised setting. This order of magnitude matches general physical activity recommendations for children, which makes the advice easy to sustain.
- Leave the choice of discipline to the child and family. What the child will actually keep doing matters more than the category their sport belongs to. Steering toward one specific discipline on the strength of this work would treat a subgroup comparison as though it were a head-to-head trial.
- Name clearly what was measured. Performances on laboratory tests, not symptoms or school results. This is a useful sentence in consultation; it heads off disappointment two months later.
- Place physical activity within the overall care plan. It complements parent guidance, school accommodations and, when indicated, medication. Nothing in this work supports presenting it as an alternative, since none of the described control conditions was an ADHD treatment.
- Know how to answer the parent who brings in the article. Yes, the signal in favor of physical activity is real and convergent. No, it does not yet allow prescribing one sport over another based on the child’s specific difficulty.
Frequently asked questions
Should an impulsive child be steered toward judo and a child who forgets instructions toward jump rope?
That would go beyond what the design supports. Each activity category was compared only with its own control condition, in separate trials, never against each other in the same children, and no test of difference between categories is reported. A correspondence between a type of sport and an executive domain is a research lead, not a referral rule.
What duration and intensity should be recommended in practice?
The clearest effects appear beyond six weeks, with at least two sessions a week, at moderate-to-vigorous intensity. Treat this as an order of magnitude, since these parameters come from subgroup analyses. It happens to coincide with what is already recommended for every child, which simplifies the message.
Will a child who improves on the tests do better in class?
Not demonstrated here, and caution is warranted. Cognitive training interventions in ADHD have shown gains on the trained tasks without a corresponding improvement in symptoms rated blind. Moving from test performance to daily life requires its own demonstration.
Are these effects comparable to those of an ADHD medication?
This work does not settle the question, and the figures are not interchangeable: medication trials are judged on symptom scales, these trials on cognitive tests. None of the described control conditions was a medication, and the review does not stratify by the treatment the included children were taking. Physical activity adds to care; it does not replace it.
Are there children for whom sport should not be recommended?
Few, and for somatic rather than psychiatric reasons: known cardiac or respiratory disease, orthopedic issues, or a situation where repeated failure in a competitive activity risks harming self-esteem. In those cases, adapting the format, rather than giving up on activity altogether, remains the rule.
Annotated bibliography
Source study. Chen C, Fan M, Yu C, Wan S. Effects of different physical activity modalities on executive function in children with attention deficit hyperactivity disorder: a systematic review and meta-analysis. Frontiers in Psychiatry, 2026 ; 17 : 1824121. DOI 10.3389/fpsyt.2026.1824121, PMID 42422527. PROSPERO protocol CRD420251080013. Contribution: quantifies the effect of physical activity programs on three executive domains and proposes a reading by activity modality. Limitations: comparisons between modalities are non-randomized and without a test of difference, discipline classification is unstable from one table to another, no formal certainty assessment, small trials, and a source literature with limited geographic diversity.
Sonuga-Barke EJS et al. Nonpharmacological interventions for ADHD: systematic review and meta-analyses of randomized controlled trials of dietary and psychological treatments. American Journal of Psychiatry, 2013 ; 170(3) : 275-289. Contribution: shows that the effect of non-pharmacological interventions collapses once outcomes are rated by a blinded assessor. This is the reference argument for reading cautiously an effect obtained against a no-intervention group. Limitations: does not concern physical activity and predates most of the trials included here.
Cortese S et al. Cognitive training for attention-deficit/hyperactivity disorder: meta-analysis of clinical and neuropsychological outcomes from randomized controlled trials. Journal of the American Academy of Child and Adolescent Psychiatry, 2015 ; 54(3) : 164-174. Contribution: illustrates precisely the risk of substitution, with gains on working memory tasks and no corresponding effect on symptoms rated blind. Limitations: concerns computerized cognitive training, not directly transposable to physical exercise.
Cortese S et al. Comparative efficacy and tolerability of medications for attention-deficit hyperactivity disorder in children, adolescents, and adults: a systematic review and network meta-analysis. The Lancet Psychiatry, 2018 ; 5(9) : 727-738. Contribution: reference framework for pharmacological management, useful for placing physical activity as a complement rather than a substitute. Limitations: symptom-based outcomes, with no direct comparison to exercise interventions.
Editorial collections
Tags
Verified on August 13, 2026 against the full text of the publication and its supplementary material where available. This analysis underwent an independent double reading. The English version was checked for conformity on September 19, 2026, against the figures of the French version and against the source. How we verify what we publish
This analysis is intended for healthcare professionals. It does not constitute a prescribing recommendation and does not replace individual clinical judgment.
Analysis from Psychiatry Evidence Base, evidence-based psychiatry, explained with rigor.