Autism spectrum disorder (ASD) is characterized by social communication deficits and restricted, repetitive behaviors (1). Beyond these core features, psychiatric comorbidities are highly prevalent. Epidemiological data indicate that 40%–50% of individuals with ASD meet the clinical criteria for anxiety disorders, while comorbidity with attention-deficit/hyperactivity disorder (ADHD) ranges from 30% to 80% (2, 3). In addition, frequent sleep disturbances are associated with the exacerbation of repetitive behaviors and emotional problems. These co-occurring conditions exacerbate the core ASD symptoms and are associated with poorer prognosis and a reduced quality of life (4).

Clinically, anxiety disorders in school-aged children with ASD are often associated with social interaction difficulties, sensory sensitivity, and reactivity to unpredictable environments (5). Similarly, co-occurring ADHD symptoms frequently overlap with the core ASD features, manifesting as inattention, hyperactivity, and impulsivity (6). Sleep problems—including difficulties in initiating sleep, frequent nighttime awakenings, and irregular sleep–wake cycles—are also prevalent and may be linked to dysregulated melatonin secretion and atypical sensory processing (7).

Recent research has demonstrated growing interest in exercise-based interventions as non-pharmacological approaches for addressing the core symptoms of ASD and its common comorbidities (8). Accumulating evidence indicates that physical activity yields benefits across multiple functional domains (9), including enhancements in motor skills (10) and reductions in the core ASD symptoms (11). Specifically, interventions have been shown to reduce repetitive behaviors (12), improve social engagement and communication (13, 14), and enhance executive function (15). Beyond the core symptoms, exercise is recognized as an effective strategy for regulating anxiety and ADHD symptoms (16, 17), potentially through physiological mechanisms such as the promotion of neurotransmitter release (18). Consequently, preliminary evidence supports physical activity as a viable strategy to mitigating anxiety, sleep disturbances, and the ADHD-related deficits in children with ASD (19, 20).

Although evidence supports the benefits of exercise for ASD (21), a key limitation in the literature is that many of the studies approach ASD as a homogeneous condition, overlooking the heterogeneity arising from common comorbidities (22). Children with co-occurring anxiety, sleep disorders, or ADHD often differ substantially in terms of their neurobiological profiles and intervention needs (23). While interventions often target specific symptoms—such as the energy expenditure for ADHD (24) or relaxation for anxiety (25)—direct quantitative comparisons of the effects of exercise across these subgroups remain scarce. Majority of the existing meta-analyses have evaluated ASD or ADHD in isolation, failing to address concurrent presentations. This gap limits the development of precision exercise prescriptions (26). To our knowledge, no systematic review has yet evaluated the combined and comparative effects of exercise on co-occurring anxiety, sleep disorders, and ADHD in this population (27, 28).

Consequently, this systematic review and meta-analysis aimed to: 1) quantify the overall effect size of exercise interventions on the core ASD symptoms; 2) evaluate their effects on comorbid psychiatric conditions—specifically anxiety, sleep disturbances, and ADHD symptoms—via subgroup analyses; and 3) identify potential moderators of intervention efficacy (e.g., exercise modality, intensity, and duration) to generate evidence-based recommendations for clinical practice.

This study employed a systematic search strategy across eight databases, initially identifying 8,270 relevant publications. After removing 1,374 duplicates, 6,896 articles were retained for further evaluation. Following a two-stage screening process based on the titles and abstracts and a subsequent full-text assessment, 12 eligible RCTs, encompassing a total of 482 participants, were ultimately included. This rigorous retrieval and selection process ensured comprehensive and systematic literature coverage, thereby establishing a robust foundation for subsequent analysis (46). Compared with existing reviews, this study addresses the broader clinical context in which high-functioning ASD (ASD-H) co-occur with anxiety, sleep disorders, and ADHD. For example, while Riis et al. (47) included eight studies focusing on anxiety interventions, the present review incorporates 12 studies on anxiety-related outcomes, reflecting broader inclusion criteria. The value of this comprehensive approach lies in its ability to capture a more representative picture of the comorbidity patterns of ASD; however, these insights remain constrained by the limited number of available studies, as also noted by Hossain et al. (48). The limited number of studies included in this review may be attributed to several factors, including the relatively recent emergence of research in this interdisciplinary area, the stringent inclusion criteria applied, and potential biases such as publication bias and language restrictions. The literature search results revealed a growing trend in the existing exercise interventions targeting comorbidities of ASD, such as anxiety, sleep disorders, and ADHD. Nevertheless, there remains a need for more high-quality RCTs to strengthen the evidence base. Consequently, the findings of this study can only be interpreted within the context of the available literature.

The 12 included studies involved a total of 482 children with ASD and employed a variety of intervention approaches, such as aerobic exercise, yoga combined with cognitive therapy, dance, martial arts, soccer, rhythmic balance training, aquatic activities, basketball, and equestrian therapy. Quality assessments rated three studies as high quality, seven studies as moderate quality, and two studies as low quality, suggesting that the overall evidence is generally acceptable despite some limitations. Although the diversity of interventions enriched insights into the efficacy of different exercise modalities, it also contributed to the increased heterogeneity in the results (49). Similar to the synthesis by Healy et al. (50) of 29 studies, the exercise interventions examined in this analysis showed moderate effects on symptom presentation in children with ASD. However, notable methodological variations were observed across the included studies, such as insufficient blinding procedures and heterogeneity in the control conditions (lack of blinding and inconsistent controls). The overall quality of the evidence may have been compromised by several sources of bias, including the small sample sizes, the challenges associated with the implementation of effective blinding, and the inconsistencies in the intervention duration and frequency (51). This suggests that the conclusions may be susceptible to certain biases related to the experimental design and reporting. For instance, the predominant reliance on subjective assessments by parents or teachers introduces the possibility of evaluator-dependent bias, which could affect the reliability of the outcomes. Consistent with this concern, individual studies have reported discrepancies among different observers, potentially undermining the consistency of the findings. While the overall evidence points to potential benefits of exercise interventions, methodological and implementation limitations indicate that the results should be interpreted with caution. Future research should focus on improving the study design to strengthen the validity and generalizability of the evidence.

A three-level random effects meta-analysis incorporating 12 studies (27 effect sizes) revealed that exercise interventions produced a moderate-to-large, statistically significant improvements in the core symptoms among children with ASD and comorbid anxiety, sleep disorders, or ADHD: Hedges’ g = 0.879 (95%CI = 0.449–1.310, p < 0.001). The sensitivity analysis further confirmed the robustness of the findings: sequentially removing each individual study yielded an effect size ranging from 0.695 to 0.914, with all estimates remaining statistically significant (p < 0.01) and without any change in the direction of the effects. These results are consistent with prior evidence, including a reported moderate overall effect size (g = 0.62) of exercise interventions in adolescents with ASD. Furthermore, other network meta-analyses have similarly demonstrated the beneficial effects of physical exercise on the social and behavioral outcomes in individuals with ASD. Statistically, the analysis revealed considerable heterogeneity (I² = 88.5%), indicating substantial differences in the intervention protocols, measurement tools, and participant characteristics across the included studies. Therefore, a random effects model was adopted for the meta-analysis. Although Egger’s test suggested potential publication bias (p < 0.001), a trim-and-fill analysis demonstrated that the core findings remained consistent after imputing theoretically missing studies, thereby strengthening the reliability of the results (52). From a theoretical perspective, exercise may enhance neuroplasticity through the modulation of neurotransmitters—such as serotonin and norepinephrine—and the cortisol levels, which could in turn alleviate the core symptoms of ASD and related comorbidities (Liang et al., 2022). Nonetheless, the interpretation of these findings should be tempered by the presence of high heterogeneity and limited sample sizes. These limitations may introduce biases resulting from inconsistent intervention quality and non-uniform outcome measures. The overall findings should be interpreted in conjunction with the context of the individual studies included (53). This study provides robust evidence supporting the value of exercise interventions as an adjunct therapeutic approach, establishing a solid foundation for future detailed subgroup and dose–response analyses. In the sub-analyses focusing on the core symptoms of children with ASD, the exercise interventions demonstrated a moderate effect on ASD symptoms (Hedges’ g = −0.56, 95%CI = 0.32–0.80, p < 0.001). The heterogeneity analysis revealed a moderate level of variation across the included studies (I² = 37.5%, Q = 17.39), with a high degree of statistical consistency. The trim-and-fill meta-analysis indicated no need for adjustment, suggesting limited publication bias. These results support the conclusion that exercise interventions can produce moderate improvements in the core symptoms of ASD in children, including social communication and repetitive behaviors. This finding is consistent with prior research, such as the study by Woodman et al. (54), which reported that high-intensity exercise significantly reduces repetitive behaviors in children with ASD. Yano et al. (55) further proposed that exercise interventions act as a key mechanism linking the immune, muscular, and brain systems, leading to improvements in social communication and a reduction in repetitive behaviors among children with ASD. From a mechanistic standpoint, exercise may enhance neural function by elevating the levels of brain-derived neurotrophic factor (BDNF) and modulating connectivity within neural networks. Concurrently, exercise interventions have been shown to promote beneficial changes in white matter connectivity and to enhance executive function networks in children with ASD (56, 57). Overall, this study supports the use of exercise as an adjunctive intervention for improving the core symptoms of ASD and provides preliminary evidence to further elucidate the mechanisms underlying different exercise modalities (58).

In the subgroup analysis of children with ASD with comorbid anxiety symptoms (six studies, 219 participants), exercise interventions were associated with a significant reduction in anxiety levels, demonstrating a moderately large effect size (Hedges’ g = −0.68, 95%CI = −1.04 to −0.33, p < 0.001). Similar to the analysis of the core ASD symptoms, the negative effect size indicates a significant reduction in anxiety symptoms. The heterogeneity among the studies was low (I² = 37%), supporting the robustness of the findings. Subgroup analyses revealed variations in anxiety reduction according to exercise type: aerobic exercise (e.g., dance and aerobic training) demonstrated the largest effect size (mean effect size = 1.18). Our subgroup analysis specifically highlights that aerobic exercise is superior to other modalities for anxiety reduction. Clinically, this suggests that interventions aiming to reduce physiological arousal in anxious children with ASD should prioritize continuous, rhythmic aerobic activities over complex motor skills training, followed by mind–body exercises (e.g., yoga and karate, g ≈ −0.54) and team sports (g = 0.53), whereas animal-assisted therapy (e.g., equine therapy) was associated with a relatively smaller effect (g = 0.40). Meta-regression further indicated that a weekly intervention frequency was a significant moderator of anxiety improvement, with higher frequencies associated with greater reductions in anxiety (β = −0.49, p = 0.009). In contrast, neither the total intervention duration nor the participant age reached statistical significance. This finding indicates that exercise programs characterized by a short duration and high frequency may be particularly beneficial in reducing anxiety among children with ASD. The anxiolytic effects of exercise are likely mediated through multiple physiological and psychological mechanisms, including regulation of the hypothalamic–pituitary–adrenal (HPA) axis, the release of endorphins, and the overall mood enhancement. Additional contributing factors may involve the stabilization of daily routines and improvements in self-efficacy (59, 60). These results are consistent with the systematic review by Trzmiel et al. (61), which supports the efficacy of various forms of physical activity—such as yoga, soccer, group-based sports, and equestrian therapy—in alleviating anxiety symptoms in individuals with ASD (3, 40, 62).

This study, which included five randomized trials involving a total of 274 children with ASD, evaluated the effects of intervention on comorbid sleep disorders. The results revealed extreme heterogeneity across studies (Cochran’s Q = 169.91, p < 0.001, I² = 98.8%), which compromised the reliability of the pooled effect estimate (g = 1.49, 95%CI = −1.33 to 4.30, including zero). Subgroup analysis based on exercise type revealed no statistically significant differences (QM = 1.047, p = 0.79), suggesting inconsistent—but not statistically significant—effects across intervention types. The meta-regression analysis revealed that the total intervention duration was the primary source of heterogeneity (QM = 16.265, p < 0.001, I² = 90%), suggesting that longer exercise interventions are generally associated with greater sleep benefits. In contrast, neither the mean age nor the exercise frequency demonstrated a significant association with the outcomes. The observed heterogeneity may be attributed to the inconsistencies across studies regarding the assessment criteria for sleep disorders, the assessment methods (e.g., subjective sleep questionnaires vs. objective monitoring), and the variations in the intervention content (63–65). The limited sample sizes and the considerable heterogeneity in the data constrained the statistical power of the analysis; therefore, caution is advised against overinterpreting the subgroup findings (66). Given the extreme heterogeneity (I² > 90%) and the wide confidence intervals, current evidence remains inconclusive with regard to the effect of physical exercise on sleep disorders in children with ASD (67). Consequently, future research should aim to enroll larger cohorts, employ standardized assessment protocols, and further investigate underlying mechanisms—such as the influence of exercise on the circadian rhythm—to establish more conclusive outcomes (68).

The subgroup analysis of children with ASD with comorbid ADHD symptoms (four studies, 148 participants) showed a pooled effect size of Hedges’ g = −0.48 (95%CI = −0.81 to −0.16, p < 0.01), indicating a moderate reduction in the core ADHD symptoms following exercise interventions. Heterogeneity was low (I² = 0%), suggesting consistent results across the included studies without significant statistical heterogeneity. Among the included studies, the effect sizes varied according to the exercise modality: rhythmic coordination activities (e.g., soccer) yielded the greatest effect (g = −0.66). This finding has significant clinical relevance, implying that exercises requiring cognitive engagement and complex motor planning (e.g., team sports and martial arts) are more effective at addressing executive function deficits than simple repetitive movements, followed by karate training (g = −0.56). In contrast, aerobic exercise (g = −0.45) and progressive combined training (g = −0.26) were associated with relatively smaller effects. This pattern may be attributable to the cognitive demands inherent in rhythm-based and team sports, which are known to engage attention and executive function (69). These results are consistent with the findings of Seiffer et al. (70), who observed small to moderate effects of regular moderate-to-vigorous exercise on ADHD symptoms (g = −0.33). Together, these findings support the role of exercise as an adjunctive therapeutic intervention in ADHD.

A mixed effects meta-regression analysis was conducted to assess the moderating effects of training volume, single-session duration, and weekly frequency on the overall effect size. None of these moderators reached statistical significance: the training volume (QM = 0.47, p = 0.495), single-session duration (QM = 0.62, p = 0.432), and weekly frequency (QM = 0.07, p = 0.788) did not significantly influence treatment efficacy. Each regression coefficient was non-significant, with a model R² = 0, indicating that these training dose metrics could not explain the heterogeneity across studies. Concurrently, the residual heterogeneity remained exceptionally high (I² > 94%, p < 0.0001), suggesting that additional variables likely contributed to the observed differences in the effects. These findings imply that quantifying the efficacy of the exercise interventions based solely on the training frequency or the duration is challenging, possibly due to more complex factors such as exercise quality, type specificity, or individual participant characteristics (71). Previous studies have frequently highlighted the significance of exercise frequency and intensity—as reflected, for instance, in the anxiety group outcomes of this study, underscoring the role of frequency. However, the multivariate regression analysis conducted at the overall sample level did not reveal significant associations (72). The limited sample size (12 studies) may have been insufficient to support complex regression modeling. Furthermore, the observed relationships suggest that the effects of exercise prescriptions are not simply linearly incremental but may involve threshold effects or interaction mechanisms (73). Due to the limitations of the available data, this study cannot confirm that specific training scale parameters exert universal moderating effects. Future research should focus on more nuanced moderating variables, such as the quality of the intervention content, social factors, and individual baseline characteristics (74).

The positive effects of exercise on anxiety and psychiatric comorbidities in children with ASD may be mediated by several neurobiological mechanisms. Firstly, regarding anxiety reduction, physical activity acts as a physiological stressor that helps regulate the HPA axis, leading to reduced basal cortisol levels and improved stress resilience. Crucially, aerobic exercise has been shown to dampen the hyperactivity of the amygdala—the brain’s fear center often dysregulated in ASD—thereby modulating emotional reactivity (49). Secondly, for ADHD-related symptoms, exercise stimulates the secretion of monoamine neurotransmitters, such as dopamine and serotonin, which play pivotal roles in mood regulation and attentional control (69). Specifically, complex motor activities (e.g., rhythmic or team sports) may activate the prefrontal cortex–cerebellum loop, which underpins executive function and inhibitory control. Furthermore, regular physical activity has been shown to upregulate BDNF, facilitating neuroplasticity and optimizing brain network connectivity, which are often dysregulated in children with ASD (49, 75).

This study offers systematic evidence regarding the efficacy of exercise interventions for children with ASD and co-occurring psychiatric conditions. However, several limitations must be noted. Firstly, the majority of the included studies are single-center trials with small sample sizes. While small sample sizes in primary studies increase the risk of type II errors, this meta-analysis mitigates that limitation by synthesizing the data into a larger pooled sample (N = 482), thereby enhancing the statistical power to detect significant treatment effects. Nevertheless, the generalizability of the findings may still be constrained by the variability in the intervention protocols (76). Secondly, the reliance on subjective assessment tools may introduce detection bias due to inter-rater variability (77). In particular, findings related to sleep outcomes are less reliable owing to measurement heterogeneity (78). Thirdly, signs of publication bias were detected (Egger’s test, p < 0.001). Although we actively engaged with grey literature (e.g., ProQuest and CNKI) to evaluate potential biases—as recommended in recent guidelines (79, 80)—the risk of bias persists. However, the trim-and-fill analysis demonstrated that the core findings remained consistent after imputing theoretically missing studies, suggesting that our primary conclusions are robust despite this potential limitation (5, 29). To advance the field, subsequent research should prioritize large-scale, multicenter RCTs and combine subjective questionnaires with objective physiological measures—such as actigraphy or neuroimaging—to reduce assessment bias (81). Furthermore, future research should prioritize elucidating the neurobiological mechanisms through which exercise exerts its effects, such as the modulation of brain network dynamics (82).

In light of current evidence, specific clinical recommendations can be formulated. For individuals with ASD and co-occurring anxiety, prioritizing moderate-to-high intensity aerobic exercise is recommended (83). For those with attention deficits (ADHD symptoms), interventions emphasizing rhythmic coordination activities and structured team sports may be particularly beneficial (84). Comprehensive reporting of intervention parameters—including heart rate zones and social context—will further facilitate dose–response analysis in clinical practice (85).

This study systematically evaluated the evidence-based effects of exercise interventions on the core symptoms of children with ASD with comorbid anxiety and ADHD while comparing outcomes across different exercise modalities. The results partially supported our hypotheses: overall, exercise interventions produced beneficial therapeutic effects, with aerobic exercise yielding the greatest reduction in anxiety symptoms and rhythmic coordination exercises demonstrating stronger regulatory effects on attention control. The core contribution of this study lies in being the first to analyze children with comorbid ASD and anxiety alongside ADHD as a distinct clinical subgroup, elucidating the specific benefits of various forms of exercise across different diagnostic dimensions. Based on the accumulated evidence, we recommend implementing a combined intervention for this population. Specifically, a “FITT” (frequency, intensity, time, and type of exercise) framework is suggested for consideration based on current evidence: frequency of three to four sessions per week; intensity at moderate-to-vigorous levels; time ranging from 30 to 60 min per session; and type focusing on aerobic exercise (for anxiety reduction) combined with rhythm-based team sports (for attention regulation). The subgroup analyses further supported this model, indicating that high-frequency interventions yield superior outcomes compared with low-frequency regimens.