Authors: Wei Chen, Zhang Chang, Wang Fuhai, He Baiqiao, Xie Peiyu, Xu Meini, Liu Cong, Wang Peng, Hong Qian, Wang Xing
Categories: Systematic Review, Exercise, Executive Function, Cognitive Function, Older Adults, Meta-analysis
Source: BMC Public Health
Authors: Wei Chen, Zhang Chang, Wang Fuhai, He Baiqiao, Xie Peiyu, Xu Meini, Liu Cong, Wang Peng, Hong Qian, Wang Xing
To systematically evaluate the effect of exercise on the executive function of cognitively healthy older adults through a three-level Meta-analysis.
Randomized controlled trials (RCTs) related to the effects of exercise on the executive function of cognitively healthy older adults were retrieved from eight databases, including PubMed, Web of Science, The Cochrane Library, EBSCOhost, Embase, CNKI, Wanfang, and VIP, with a search period from the establishment of the databases to February 27, 2025. The quality of the literature was assessed using the PEDro scale and the evidence quality was evaluated using GRADE. Meta-analysis was performed using a three-level random effects model in R, and publication bias was tested. Hedges’ g value and its 95% confidence interval (CI) were used for assessment, and a 95% prediction interval (PI) was calculated to determine the expected range of effect sizes in future similar studies. Funnel plots and Egger’s regression were used to assess publication bias.
A total of 50 RCTs were included, involving 4,826 cognitively healthy older adults. The PEDro scale scores ranged from 5 to 8, with an average of 6.4, indicating that the overall quality of the included literature was good.
After excluding outliers, the meta-analysis revealed that exercise significantly improved overall executive function in older adults (g= 0.155, 95% CI 0.084, 0.226, p < 0.001). Significant improvements were also observed in the subdomains of working memory (g = 0.208, 95% CI 0.093, 0.322,P<0.001) and cognitive flexibility (g = 0.150, 95% CI 0.001, 0.299,P= 0.049). Regarding exercise parameters, significant benefits were found for a frequency of ≥3 times/week (g = 0.188, 95% CI 0.097, 0.280, p < 0.001), a duration of ≤26 weeks (g = 0.193, 95% CI 0.114, 0.272, p< 0.001), a session time of ≤60 minutes (g = 0.190, 95% CI 0.115, 0.264, p < 0.001), and moderate-intensity exercise (g = 0.254, 95% CI 0.117, 0.390, p < 0.001), with both aerobic (g = 0.130, 95% CI 0.043, 0.218, p = 0.003) and resistance training (g = 0.232, 95% CI 0.077, 0.388, p = 0.003) being effective. Furthermore, these improvements were consistent across all population subgroups, including those aged 60-69 years (g = 0.134, 95% CI 0.049, 0.218, p = 0.002), those ≥70 years (g = 0.203, 95% CI 0.075, 0.330, p = 0.002), those with ≤12 years of education (g = 0.322, 95% CI 0.135, 0.510, p < 0.001), those with >12 years of education (g = 0.149, 95% CI 0.016, 0.283, p = 0.028), studies with ≥50% female participants (g = 0.137, 95% CI 0.067, 0.206, p < 0.001), studies with <50% female participants (g = 0.359, 95% CI 0.083, 0.635, p = 0.010), and studies employing a usual care control design (g = 0.205, 95% CI 0.113, 0.296, p < 0.001).
The Egger's test (t = 2.630，p= 0.009) indicates a potential publication bias. The GRADE evidence quality for the impact of exercise on executive function in cognitively healthy older adults is rated as moderate.
Exercise is an effective way to enhance overall executive function, as well as working memory and cognitive flexibility in cognitively healthy older adults. Compared to the control group, moderate-intensity aerobic exercise and resistance training performed ≥3 times/week, each session lasting ≤60 minutes and with a duration of ≤26 weeks, can improve overall executive function, working memory, and cognitive flexibility, especially for older adults with higher age and lower education levels. However, future exercise programs should be tailored to the individual needs and physical conditions of older adults.
This study has been registered on PROSPERO with the registration number CRD420251000402. Registration www.crd.york.ac.uk.
The online version contains supplementary material available at 10.1186/s12889-026-27418-w.
According to the World Health Organization, it is projected that by 2050, the global population aged 60 years and older will reach 2 billion [1]. With the rapid acceleration of global aging and increasing life expectancy, the incidence of age-related degenerative diseases is also increasing [2], with cognitive decline being particularly pronounced. The number of individuals experiencing subjective cognitive decline, mild cognitive impairment (MCI), and dementia (DE) is increasing [3, 4]. Patients with cognitive impairments typically exhibit varying degrees of decline in attention, memory, language expression, and executive function [5–8], which can significantly affect the lives of older adults. Consequently, finding ways to prevent and delay cognitive decline and maintain the physical and mental health of elderly individuals has become a core issue that urgently needs to be addressed in the context of an aging society.
Exercise, as a promising nonpharmacological intervention, has been utilized to maintain and improve cognitive decline [9]. Research indicates that exercise has a positive effect on healthy aging in older adults [10, 11] and is positively correlated with cognitive function in this population [12, 13]. However, existing studies have focused primarily on the effects of exercise on overall cognitive function, with limited attention given to executive function [14]. Executive function refers to an individual’s ability to regulate and monitor relevant information in a goal-directed manner during the information processing process [15, 16], which is largely controlled by the frontal lobe of the brain. It includes core domains such as working memory, inhibitory control, and cognitive flexibility [17] and reflects higher-order cognitive processes such as planning and reasoning [18, 19]. Executive functions play a crucial role in daily life, not only predicting the decline in activities of daily living among older adults [20], but also affecting their ability to handle complex tasks when executive functions deteriorate, thereby accelerating overall cognitive decline and increasing mortality rates [21, 22]. Numerous studies have shown [23–25] that exercise can improve inhibitory control by altering activation patterns in related brain regions, enhancing functional connectivity between brain networks, and influencing hormone release. Integrating behavioral and imaging evidence [12, 26], exercise has been found to enhance synaptic plasticity in the hippocampus-prefrontal cortex circuit, thereby improving working memory. Regarding cognitive flexibility, research findings have yielded inconsistent results, with some studies indicating that exercise promotes task switching and mental flexibility [27, 28], while others have found no significant effects [29]. As for higher-level planning and reasoning abilities, research on their relationship with exercise remains relatively limited, and the underlying mechanisms require further exploration.
Although existing studies have provided important evidence-based support for interventions in older adults with cognitive decline, cognitively healthy older adults also experience gradual decline in cognitive function [30], with executive function deterioration being a hallmark manifestation of this decline [31].More importantly, executive functions are not a unitary structure but a complex system comprising key subdomains such as inhibitory control, working memory, and cognitive flexibility. The research focuses on its four core inhibitory control, working memory, cognitive flexibility, and reasoning/planning. Inhibitory control, working memory, and cognitive flexibility are widely regarded as the most representative and well-measured foundational components of executive functions [32]. In contrast, reasoning and planning, as higher-order integrative abilities, have been less extensively studied in relation to exercise. Including this dimension helps to comprehensively assess the potential impact of exercise on complex cognition. Therefore, separately examining the effects of exercise on the subdomains of executive functions is particularly important for exploring how exercise may prevent or delay declines in various aspects of executive functioning in older adults.However, heterogeneity in exercise protocols (e.g., type, intensity, frequency) may lead to inconsistent conclusions. For example, Cherup et al. [33] demonstrated that higher-intensity resistance exercise significantly improved cognitive function in older adults compared to controls, whereas lower-intensity aerobic exercise showed no such improvement. Liu-Ambrose et al. [34] found that resistance exercise performed 2 times/week significantly enhanced performance on the Flanker task in older women compared to once-weekly sessions. Furthermore, the diversity of outcome measures for executive functions (e.g., Stroop test [35–44], N-back task [45–49], Trail Making Test [27–29, 50–52]) makes direct comparisons of results challenging and obscures a clear understanding of the differential effects of exercise on overall executive functions and its subdomains. Additionally, the differential impacts of potential moderating variables (e.g., age [53], gender [54], years of education [55]) on intervention effects have not been systematically elucidated, further contributing to heterogeneity in research findings.To address these issues, this study comprehensively incorporates randomized controlled trials (RCTs) published in both Chinese and English. It employs a three-level meta-analytic approach, utilizing a three-level model to handle multiple effect sizes from each study. This method aims to maximize the use of original data and optimize statistical power [56]. Within this framework, the study seeks to re-evaluate the effects of exercise interventions on both overall executive functions and specific subdomains in older adults. It also conducts detailed dose-response analyses (type, intensity, frequency, session time, exercise duration) tailored to specific subdomains of executive functions.This study aims to investigate whether the components of exercise protocols, research design characteristics, and sample attributes have potential moderating effects on executive function. The findings are intended to provide a theoretical basis for developing targeted strategies to prevent cognitive decline in healthy older adults within an aging society, as well as for formulating exercise interventions tailored to older adults with cognitive impairment.
This meta-analytical review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement [57] and was registered in an international database of systematic reviews in health and social care (registration CRD420251000402).
This study includes research that meets the PICOS (1) Participants: Cognitively healthy older adults aged 60 years and above. (2) Intervention: In the experimental group, exercise was used as the intervention method. (3) Control: The control group does not receive specific exercise interventions, apart from maintaining daily activities or using a placebo control (such as balance and stretching training, health education, etc.). (4) Outcomes: Executive function or its subdomains are measured via neuropsychological tests. (5) Study Design: RCTs. (6) The studies were published in both Chinese and English.
Exclusion (1) Participants were diagnosed with MCI, DE, neurological disorders, or mental health disorders. (2) Exercise programmes that include confounding factors of nonexercise interventions, such as exercise games that incorporate cognitive training, vitamin supplements, or pharmacological interventions. (3) Data that could not be extracted and original data were not obtained even after the authors were contacted. (4) Nonrandomized controlled trials, reviews, or conference papers. (5) Literature that presents duplicate data.
The literature search included eight PubMed, the Cochrane Library, EMBASE, Web of Science, EBSCOhost, China National Knowledge Infrastructure (CNKI), Wanfang, and VIP. The search timeline spans from the inception of each database until February 27, 2025. Additionally, a manual search will be conducted for the references of the included literature. Two researchers independently reviewed the titles and abstracts of the identified studies. If a study preliminarily met the inclusion criteria, the full text was obtained for further evaluation on the basis of the inclusion and exclusion criteria. In cases where there is a disagreement in judgment, a discussion will be held with a third researcher to determine eligibility for inclusion. The literature search strategy is detailed in Table 1.
Table 1Literature search strategiesDatabasePubMed、The Cochrane Library#1 (“physical exercise“[Title/Abstract] OR “physical activity“[Title/Abstract] OR “aerobic exercise“[Title/Abstract] OR “resistance exercise“[Title/Abstract] OR “strength exercise“[Title/Abstract] OR “mind-body exercise“[Title/Abstract] OR “flexibility exercise“[Title/Abstract] OR “coordinative training“[Title/Abstract] OR “multicomponent exercise“[Title/Abstract])#2 (“cognition“[Title/Abstract] OR “executive function“[Title/Abstract] OR “executive control“[Title/Abstract] OR “inhibitory control“[Title/Abstract] OR “working memory“[Title/Abstract] OR “cognitive flexibility“[Title/Abstract] OR “planning“[Title/Abstract])#3 (“old people“[Title/Abstract] OR “elderly“[Title/Abstract] OR “old age“[Title/Abstract] OR “the aged“[Title/Abstract] OR “senior citizen“[Title/Abstract] OR “older adults“[Title/Abstract])#4 Randomized controlled trial [Publication Type]#5 #1 AND #2 AND #3 AND #4Web of Science#1 TS=(“physical exercise” OR “physical activity” OR “aerobic exercise” OR “resistance exercise” OR “strength exercise” OR “mind-body exercise” OR “flexibility exercise” OR “coordinative training” OR “multicomponent exercise”)#2 TS=(“cognition” OR “executive function” OR “executive control” OR “inhibitory control” OR “working memory” OR “cognitive flexibility” OR “planning”)#3 TS=(“old people” OR “elderly” OR “old age” OR “the aged” OR “senior citizen” OR “older adults”)#4 TS=(“Randomized controlled trial”)#5 #1 AND #2 AND #3 AND #4EMbase#1 ' physical exercise ‘:ab, ti OR ' physical activity ‘:ab, ti OR ' aerobic exercise ‘:ab, ti OR ' resistance exercise ‘:ab, ti OR ' mind-body exercise ‘:ab, ti OR ' flexibility exercise ‘:ab, ti OR ' coordinative training ‘:ab, ti OR ' multicomponent exercise ‘:ab, ti#2 ' cognition ‘:ab, ti OR ' executive function ‘:ab, ti OR ' executive control ‘:ab, ti OR ' inhibitory control ‘:ab, ti OR ' working memory ‘:ab, ti OR ' cognitive flexibility ‘:ab, ti OR ' planning ‘:ab, ti#3 ' old people ‘:ab, ti OR ' elderly ‘:ab, ti OR ' old age ‘:ab, ti OR ' senior citizen ‘:ab, ti OR ' the aged ‘:ab, ti OR ' older adults ‘:ab, ti#4 ' Randomized controlled trial ‘:ab, ti#5 #1 AND #2 AND #3 AND #4EBSCOhost#1 XB (physical exercise OR physical activity OR aerobic exercise OR resistance exercise OR strength exercise OR mind-body exercise OR flexibility exercise OR coordinative training OR multicomponent exercise)#2 XB (cognition OR executive function OR executive control OR Inhibitory control OR Working memory OR Cognitive flexibility OR planning)#3 XB (old people OR elderly OR old age OR the aged OR senior citizen OR older adults)#4 XB (Randomized controlled trial)#5 #1 AND #2 AND #3 AND #4CNKITS =(运动 + 体育锻炼 + 有氧运动 + 体力活动+ 抗阻运动 + 身心运动) AND TS =༈执行功能 + 抑制控制 + 工作记忆+ 认知灵活性 + 抑制 + 刷新 + 转换༉AND TS =༈老人 + 老年人༉Wangfang、VIPTS=(运动 OR 体育锻炼 OR 有氧运动 OR 体力活动OR 抗阻运动 OR 身心运动) AND TS =༈执行功能 OR 抑制控制 OR 工作记忆OR 认知灵活性 OR 抑制 OR 刷新 OR 转换༉AND TS =༈老人 OR 老年人༉
Two researchers independently extracted information from the literature included in the analysis, documenting and coding relevant data. In cases where data were missing or could not be extracted from the literature, the authors of the studies were contacted to provide the missing information. The extracted content included the first author, publication year, study type, basic information about the study population (sample size, age, sex), elements of the exercise prescription (type of exercise, intensity, frequency, duration, and duration), and outcome measures. If a study included both exercise interventions and other forms of interventions, only the exercise group and the control group were included in the analysis. Studies that reported multiple groups of data were counted as multiple studies. If a study included a mixed population of both cognitively healthy older adults and those with other health conditions, only data from the cognitively healthy subgroup were extracted and included in the analysis. When there were directional differences in outcome measures, the average values were multiplied by −1 to ensure consistent interpretation of effect sizes, following the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions [58]. In this study, exercise intensity from the included literature was converted to average values via a triple coding strategy to eliminate fixed intensity values were retained; interval intensities (e.g., 40%–50%) were converted to their means (the mean value was 45%) [59]; and progressive program execution durations were weighted for conversion [60].
All test results pertaining to executive function were pooled to generate an overall effect size, coded as global executive function. For more detailed analysis, individual tests were classified into subdomains—inhibitory control, working memory, cognitive flexibility, and reasoning/planning—based on established literature [61], with each study outcome coded into one or more subcategories to allow independent effect size calculations for each subdomain. Moderator variables were categorized as mean age (60–69 years and ≥ 70 years), average years of education (≤ 12 years and > 12 years), type of control groups (active control, including stretching and balance, and conventional control group [62]), sex distribution (female proportion ≥ 50% and < 50% [63]), exercise type (aerobic, resistance, and multimodal exercise [combining two or more modalities such as aerobic, strength/resistance, balance/coordination, and flexibility training] [64, 65]), exercise intensity (low to relatively low, moderate, and relatively high to high), session duration (≤ 60 min and > 60 min), exercise duration (≤ 26 weeks and > 26 weeks), and exercise frequency (< 3 times/week and ≥ 3 times/week [60]).
The quality of the included studies was evaluated via the Physiotherapy Evidence Database (PEDro) scale, with a score of ≥ 6 indicating high-quality research. Additionally, the GRADE (Grading of Recommendations Assessment, Development and Evaluation) system was utilized to assess the quality of evidence for outcome measures [66]. Each outcome measure was rated from high to low quality and categorized into four high, moderate, low, and very low. Two researchers independently evaluated the included literature on the basis of the assessment criteria, and in cases of disagreement, a third researcher participated in the discussion until a consensus was reached.
Data analysis was conducted via R software. To manage the dependency of effect sizes within studies, a three-level meta-analysis using a random-effects model was employed. This model accounts for the variability of effect sizes across three sampling variance (level 1), within-study variance (level 2), and between-study variance (level 3). Standardized mean differences and their variances were calculated on the basis of the posttest means, standard deviations, and sample sizes of the intervention and control groups. Using the restricted maximum likelihood (REML) method, Hedges’ g values and their 95% confidence intervals (CIs) were assessed, where g ≤ 0.20 indicated a small effect, 0.20 < g < 0.49 indicated a small to moderate effect, 0.5 ≤ g < 0.79 indicated a moderate effect, and g ≥ 0.80 indicated a large effect [67]. The statistical significance of within-study (level 2) and between-study (level 3) variances was tested via likelihood ratio tests (LRTs). A 95% prediction interval (PI) was calculated [68] to determine the expected range of effect sizes in future similar studies. Statistical significance was defined as p < 0.05. Funnel plots and Egger’s regression test were employed to assess publication bias.
Through systematic searching, a total of 35,880 initial records were identified. After removing 13,228 duplicates via EndNote software, 22,652 records remained. By screening titles and abstracts, 22,384 irrelevant records were excluded. The full texts of 268 studies were assessed for eligibility. This led to the exclusion of 21 studies for nonconformity with outcome measures, 49 for nonconformity with intervention methods, 13 due to data that could not be extracted, 19 for being non-randomized controlled trials, 59 for population mismatches, 4 as duplicate publications, and 53 for being reviews or conference proceedings. Ultimately, 50 studies [27–29, 33–52, 69–95] met the inclusion criteria (Fig. 1).
Fig. 1Flowchart of Literature Retrieval
The quality assessment of the included studies is presented in Table 2. All studies met the eligibility criteria, including random allocation, baseline similarity, and intergroup statistical analysis, as well as point measurements and difference values. 15 studies implemented allocation concealment, and 17 studies employed assessor blinding. In addition, 44 studies had a participant retention rate exceeding 85%, and 45 studies utilized intention-to-treat analysis. Among the included studies, 44 were classified as high quality, 6 as moderate quality, and no low-quality studies were identified. The PEDro scores ranged from 5 to 8, with an average score of 6.4, indicating a generally good quality of the overall literature.
Table 2Quality evaluation resultsStudy1234567891011TSQuality gradeCai et al.,2024 [45]YYYYNNYYNYY7High Liu 2020 [69]YYNYNNNYYYY6HighYin et al.,2022 [70]YYNYNNNNYYY5ModerateZhao et al.,2020 [35]YYYYNNNYYYY7HighAlbinet et al.,2016 [71]YYNYNNNNYYY5ModerateAlbinet et al.,2010 [72]YYNYNNNYYYY6HighAlexandra et al.,2020 [28]YYYYNNYYYYY8HighBerryman et al.,2014 [73]YYNYNNNYYYY6HighBest et al.,2015 [36]YYNYNNYYYYY7HighByun et al.,2024 [37]YYNYNNNYYYY6HighCassilhas et al.,2007 [74]YYNYNNNYYYY6HighChapman et al.,2013 [38]YYNYNNNYYYY6HighCherup et al.,2018 [33]YYNYNNNYYYY6HighCoetsee et al.,2017 [39]YYNYNNNYYYY6HighErickson et al.,2011 [75]YYNYNNNNYYY5ModerateFabre et al.,2002 [76]YYYYNNNYYYY7HighFerreira et al.,2015 [77]YYNYNNYNNYY5ModerateFranco et al.,2020 [29]YYYYNNYYNYY7HighGerritsen et al.,2021 [78]YYNYNNNYYYY6HighGothe et al.,2016 [46]YYNYNNYYYYY7HighHariprasad et al.,2013 [40]YYNYNNNYYYY6HighHong et al.,2013 [79]YYNYNNNYYYY6HighIuliano et al.,2015 [50]YYNYNNNYYYY6HighJamrasi et al.,2024 [41]YYYYNNNYNYY6HighJonasson et al.,2016 [47]YYNYNNNYYYY6HighKimura et al.,2010 [80]YYNYNNNYYYY6HighKlusmann et al.,2010 [42]YYYYNNYYYYY8HighLeckie et al.,2014 [48]YYNYNNNYYYY6HighLegault et al.,2011 [81]YYNYNNNYYYY6HighLiu-Ambrose et al.,2008 [51]YYYYNNYYYYY8HighLiu-Ambrose et al.,2012 [34]YYYYNNYYYYY8HighLu et al.,2016 [43]YYNYNNYYYYY7HighMaki et al.,2012 [27]YYNYNNNYYYY6HighMortimer et al.,2012 [82]YYNYNNNYYYY6HighNguyen et al.,2012 [52]YYNYNNNNYYY5ModerateNouchi et al.,2014 [44]YYYYNNNYYYY7HighOken et al.,2006 [83]YYNYNNNYYYY6HighPereira et al.,2020 [84]YYYYNNNYYYY7HighQi et al.,2021 [85]YYYYNNYYYYY8HighSexton et al.,2020 [86]YYYYNNYYNYY7High Shatil 2013 [87]YYNYNNNYYYY6HighSink et al.,2015 [88]YYNYNNNNYYY5ModerateVaughan et al.,2014 [89]YYYYNNYYYYY8HighWalsh et al.,2015 [90]YYNYNNNYYYY6HighWang et al.,2024 [91]YYNYNNNYYYY6HighWelford et al.,2023 [92]YYYYNNYYYYY8HighWilliamson et al.,2009 [93]YYNYNNYYYYY7HighWu et al.,2021 [94]YYNYNNYYYYY7HighYang et al.,2019 [95]YYNYNNYYYYY7HighZhao et al.,2022 [49]YYNYNNNYYYY6HighMean score6.4
All included studies were RCTs, with a total of 2,493 participants in the experimental group and 2,183 participants in the control group, all with an average age of over 60 years. Among these, 3 studies did not report the gender distribution of participants, 7 studies focused on female participants, and 1 study focused on male participants. 22 studies provided information on the average years of education of the participants, whereas 28 studies did not report this information. An overview of the participants is presented in Table 3.Table 3Overview of participants’ characteristics in the included studiesStudySampleFemale-to-male ratioAgeYears of educationCai et al.,2024 [45]E:25 C:2731/2183.64 ± 6.539.68 ± 5.1 Liu 2020 [69]E:22 C:19N/A65.57 ± 5.12N/AYin et al.,2022 [70]E:14 C:1117/869.08 ± 4.6813 ± 3.46Zhao et al.,2020 [35]E:60 C:6040/8064.42 ± 1.429.42 ± 4.12Albinet et al.,2016 [71]E:19 C:1726/1067 ± 511.89 ± 3.87Albinet et al.,2010 [72]E:12 C:1213/1170.9 ± 4.911.3 ± 4.6Alexandra et al.,2020 [28]E:55 C:52107/069.98 ± 7.83N/ABerryman et al.,2014 [73]E1:16C:1628/18E1:69.3 ± 3.9E1:15.3 ± 3.4E2:15E2:69.5 ± 6.1E2:14.9 ± 2.4Best et al.,2015 [36]E1:46C:42135/0E1:69.4 ± 3E1: N/AE2:47E2:69.5 ± 2.7E2: N/AByun et al.,2024 [37]E:40 C:4161/2068.713.4Cassilhas et al.,2007 [74]E1:20 C:230/62E1:68.4 ± 0.67E1: N/AE2:19E2:69.01 ± 1.1E2: N/AChapman et al.,2013 [38]E:18 C:1927/864 ± 4.3N/ACherup et al.,2018 [33]E1:9 C:720/6E1:72.2 ± 2.6E1: N/AE2:8E2:68.5 ± 2.8E2: N/ACoetsee et al.,2017 [39]E1:13 C:1931/14E1:61.6 ± 5.8E1: N/AE2:22E2:62.4 ± 5.1E2: N/AErickson et al.,2011 [75]E:60 C:6080/4067.6 ± 5.81N/AFabre et al.,2002 [76]E:8 C:8N/A65.4 ± 2.211.2 ± 1.3Ferreira et al.,2015 [77]E:22 C:2234/1066.5 ± 5.612.9 ± 2.7Franco et al.,2020 [29]E:35 C:3665/668.6 ± 7.2N/AGerritsen et al.,2021 [78]E:20 C:2320/2363.95 ± 7.25N/AGothe et al.,2016 [46]E:61 C:5792/2662.1 ± 5.82N/AHariprasad et al.,2013 [40]E:62 C:5872/4875.74 ± 6.4613.05 ± 4.09Hong et al.,2013 [79]E:12 C:1316/975.53 ± 4.48N/AIuliano et al.,2015 [50]E1:20C:2035/25E1:68.44 ± 6.4E1:11.08 ± 4.18E2:20E2:68.5 ± 6.32E2:12.4 ± 3.45Jamrasi et al.,2024 [41]E1:25 C:2754/25E1:73 ± 4.6E1: N/AE2:27E2:73.9 ± 4.2E2: N/AJonasson et al.,2016 [47]E:29 C:2933/2568.4 ± 2.5413.69 ± 3.49Kimura et al.,2010 [80]E:65 C:5470/4973.6 ± 4.7N/AKlusmann et al.,2010 [42]E:91 C:76167/073.6 ± 411.8 ± 2.5Leckie et al.,2014 [48]E:47 C:4561/3167.23 ± 5.39N/ALegault et al.,2011 [81]E:18 C:1821/1577.5 ± 4.8N/ALiu-Ambrose et al.,2008 [51]E:31 C:2841/1881.4 ± 6.2N/ALiu-Ambrose et al.,2012 [34]E1:20 C:1752/0E1:69.7 ± 2.8E1: N/AE2:15E2:68.9 ± 3.2E2: N/ALu et al.,2016 [43]E:15 C:1631/072.8 ± 6.7N/AMaki et al.,2012 [27]E:75 C:75106/4471.9 ± 4.111.8 ± 2.5Mortimer et al.,2012 [82]E1:30 C:3059/31E1:67.3 ± 5.3E1: N/AE2:30E2:67.8 ± 5E2: N/ANguyen et al.,2012 [52]E:48 C:4848/4869.23 ± 5.3N/ANouchi et al.,2014 [44]E:32 C:32N/A66.75 ± 4.6113.44 ± 1.85Oken et al.,2006 [83]E1:44 C:44101/33E1:71.5 ± 4.9E1:15.4 ± 2.2E2:47E2:73.6 ± 5.1E2:14.8 ± 2.8Pereira et al.,2020 [84]E:24 C:2530/1866 ± 513 ± 7Qi et al.,2021 [85]E:22 C:2633/1563.91 ± 4.0613.14 ± 3.11Sexton et al.,2020 [86]E:23 C:2329/1765.5 ± 4N/A Shatil 2013 [87]E:31 C:2941/1979 ± 5.76N/ASink et al.,2015 [88]E:735 C:741999/47770–89N/AVaughan et al.,2014 [89]E:25 C:2449/069 ± 3.112.3 ± 2.9Walsh et al.,2015 [90]E:29 C:3140/2063.94 ± 8.0217.13 ± 23.41Wang et al.,2024 [91]E:18 C:1626/865.22 ± 4.32N/AWelford et al.,2023 [92]E1:27C:2661/19E1:72.1 ± 5.5E1: N/AE2:29E2:73.3 ± 5.5E2: N/AWilliamson et al.,2009 [93]E:50 C:5272/3076.8 ± 4.37N/AWu et al.,2021 [94]E:19 C:1933/563.6 ± 413.6 ± 2.2Yang et al.,2019 [95]E:13 C:1326/066.31 ± 4.2513.46 ± 2.11Zhao et al.,2022 [49]E:18 C:1626/865.2 ± 4.3N/AE = exercise group; C = control group; N/A = unspecified
A total of 50 RCTs were included, comprising 61 experimental groups. Among these, 36 experimental groups engaged in aerobic exercise, 13 groups participated in resistance training, and 12 groups were involved in multimodal exercise. The duration of the exercise interventions ranged from 8 to 52 weeks, with exercise sessions lasting between 30 and 90 min and a frequency of 1–5 times/week. 31 experimental groups reported on exercise intensity, which varied from low to high intensity. 25 studies measured cognitive flexibility, 27 studies assessed inhibition control, 34 studies evaluated working memory, and 2 studies measured reasoning/planning. The characteristics of the exercise interventions and outcome measures are presented in Table 4.
Table 4Overview of intervention and executive function assessment characteristics in the included studiesStudyGroupExercise interventionAdherence rateIntensityOutcome measureCai et al.,2024 [45]E: ResistanceC: No exercise/ADL3 times/w, 40 min/session, 16 w83.3%66.5% HRmax (Moderate)Working memory (N-back) Liu 2020 [51]E: MultimodalC: No exercise/ADL2 times/w, 90 min/session, 24 w100%RPE 13 (Moderate)Working memory (Backward Digit Span, Forward Digit Span)Yin et al.,2022 [70]E: MultimodalC: No exercise/ADL3 times/w, 50–70 min/session, 20 w93.3%60% HRmax (Relatively high)Working memory (Verbal fluency, Backward Digit Span, Forward Digit Span)Zhao et al.,2020 [35]E: Tai Chi (Aerobic)C: No exercise/ADL5 times/w, 60 min/session, 12 w100%N/AInhibition (Stroop)Albinet et al.,2016 [71]E: Swimming (Aerobic)C: Stretching2 times/w, 60 min/session, 21 w100%52.5% HRR (Moderate)Inhibitory control (Stroop test, Random number generation, Hayling task)Working memory (Spatial running span, Verbal running span, 2-back)Cognitive flexibility (Dimension switching, Plus-minus, Digit-letter)Albinet et al.,2010 [72]E: AerobicC: Stretching3 times/w, 60 min/session, 12 w80.3%50% HRR (Moderate)Cognitive flexibility (Wisconsin Card Sorting Test)Alexandra et al.,2020 [28]E: Pilates (Aerobic)C: No exercise/ADL2 times/w, 60 min/session, 12 w94.5%N/AWorking memory(Isaacs test)Cognitive flexibility (Trail Making Test-B)Berryman et al.,2014 [73]E1: MultimodalE2: MultimodalC: Stretching3 times/w, 60 min/session, 8 w94.1%N/AInhibition & Working memory (Random Number Generation)Best et al.,2015 [36]E1: ResistanceE2: ResistanceC: BalanceE1: 2 times/w, 60 min, 52 wE2: 1 time/w, 60 min, 52 w83.3%90% 1RM (High)Inhibition (Stroop)Working memory (Backward Digit Span)Cognitive flexibility (Trail Making Test B-A)Byun et al.,2024 [37]E: Cycling (Aerobic)C: No exercise/ADL3 times/w, 50 min/session, 12 w88.4%N/AInhibition (Stroop)Cassilhas et al.,2007 [74]E1: ResistanceE2: ResistanceC: Stretching3 times/w, 60 min/session, 24 w87.0%E1: 50% 1RM (Moderate)E2: 80% 1RM (Relatively high)Working memory (Backward Digit Span, Forward Digit Span, Corsi’s block-tapping)Chapman et al.,2013 [38]E: AerobicC: Waitlist3 times/w, 60 min/session, 12 w72.7%62.5% HRmax (Light)Inhibition (Stroop)Working memory (Backward Digit Span)Cognitive flexibility (Trail Making Test B-A)Cherup et al.,2018 [33]E1: ResistanceE2: Treadmill (Aerobic)C: No exercise/ADL3 times/w, 35 min/session, 12 w100%E1: (Relatively high)E2: 55% HRR (Moderate)Inhibition (NIH Toolbox)Working memory (NIH Toolbox)Cognitive flexibility (NIH Toolbox)Coetsee et al.,2017 [39]E1: AerobicE2: ResistanceC: No exercise/ADL3 times/w, E1: 45 min, E2: 30 min, 16 w100%E1: 72.5% HRmax (Relatively high)E2: 80% 1RM (Relatively high)Inhibition (Stroop)Erickson et al.,2011 [75]E: AerobicC: Stretching3 times/w, 40 min/session, 52 w100%62.5% HRR (Relatively high)Working memory (Spatial Memory Task)Fabre et al.,2002 [76]E: AerobicC: No exercise/ADL2 times/w, 60 min/session, 8 w90%N/AWorking memory (Forward Digit Span)Ferreira et al.,2015 [77]E: Walking (Aerobic)C: No exercise/ADL3 times/w, 40–50 min/session, 24 w80%70% HRR (Relatively high)Working memory (Backward Digit Span, Forward Digit Span, Corsi’s block-tapping)Cognitive flexibility (Wisconsin Card Sorting Test)Franco et al.,2020 [29]E: Dance (Aerobic)C: No exercise/ADL2 times/w, 60 min/session, 12 w100%N/ACognitive flexibility (Trail Making Test B-A)Gerritsen et al.,2021 [78]E: Tai Chi (Aerobic)C: No exercise/ADL2 times/w, 45 min/session, 10 w91.6%N/AInhibitory control (Stop-signal task)Gothe et al.,2016 [46]E: Yoga (Aerobic)C: Stretching3 times/w, 60 min/session, 8 w82.7%N/AWorking memory (Running memory span, N-back)Cognitive flexibility (Trail Making Test B-A)Hariprasad et al.,2013 [40]E: Yoga (Aerobic)C: Waitlist1 time/w, 60 min/session, 24 w100%N/AInhibition (Stroop)Working memory (Spatial span, Backward Digit Span, Forward Digit Span)Cognitive flexibility (Trail Making Test-B)Hong et al.,2013 [79]E: ResistanceC: No exercise/ADL2 times/w, 60 min/session, 12 w64.7%65% 1RM (Moderate)Working memory (Backward Digit Span, Forward Digit Span, Controlled Oral Word Association Test)Iuliano et al.,2015 [50]E1: AerobicE2: ResistanceC: No exercise/ADL3 times/w, 40 min/session, 12 w85.3%E1: 75% HRR (Relatively high)E2: 82.5% 1RM (Relatively high)Inhibition (Stroop)Cognitive flexibility (Trail Making Test-B)Planning (Raven’s matrices)Jamrasi et al.,2024 [41]E1: MultimodalE2: Walking (Aerobic)C: Stretching2 times/w, 50 min/session, 12 w100%RPE 13 (Moderate)Inhibition (Stroop)Jonasson et al.,2016 [47]E: AerobicC: Stretching3 times/w, 30–60 min/session, 24 w95.1%60% HRmax (Relatively high)Inhibition (Flanker)Working memory (Backward Digit Span, N-back, Letter Memory)Cognitive flexibility (Trail Making Test-4)Planning (Raven’s matrices, Letter sets)Kimura et al.,2010 [80]E: MultimodalC: Health Education2 times/w, 90 min/session, 12 w70.9%60% 1RM (Moderate)Cognitive flexibility (Task switching)Klusmann et al.,2010 [42]E: MultimodalC: No exercise/ADL3 times/w, 90 min/session, 25 w80%N/AInhibition (Stroop)Leckie et al.,2014 [48]E: Walking (Aerobic)C: Stretching3 times/w, 40 min/session, 52 w100%67.5% HRR (Relatively high)Cognitive flexibility (Task switching)Legault et al.,2011 [81]E: Walking (Aerobic)C: Health Education2 times/w, 75 min/session, 16 w100%N/AInhibition (Flanker)Working memory (N-back)Cognitive flexibility (Trail Making Test B-A, Task switching)Liu-Ambrose et al.,2008 [51]E: MultimodalC: Health Education2–3 times/w, 30 min/session, 24 w83.3%N/AInhibition (Stroop)Working memory (Verbal Digits Backward)Cognitive flexibility (Trail Making Test-B)Liu-Ambrose et al.,2012 [34]E1: ResistanceE2: ResistanceC: BalanceE1: 2 times/w, 60–90 min, 52 wE2: 1 time/w, 60–90 min, 52 w90%90% 1RM (High)Inhibition (Flanker)Lu et al.,2016 [43]E: Tai Chi (Aerobic)C: Music/Fall Prevention3 times/w, 90 min/session, 16 w96.6%N/AInhibition (Stroop)Maki et al.,2012 [27]E: Walking (Aerobic)C: Health Education1 time/w, 90 min/session, 12 w83.7%N/ACognitive flexibility (Trail Making Test-B)Mortimer et al.,2012 [82]E1: Tai Chi (Aerobic)E2: Walking (Aerobic)C: No exercise/ADL3 times/w, 50 min/session, 40 w87.9%N/AInhibition (Stroop)Working memory (Spatial span, Backward Digit Span, Forward Digit Span)Cognitive flexibility (Trail Making Test-B)Nguyen et al.,2012 [52]E: Tai Chi (Aerobic)C: No exercise/ADL2 times/w, 60 min/session, 24 w97.8%N/ACognitive flexibility (Trail Making Test-B)Nouchi et al.,2014 [44]E: MultimodalC: No exercise/ADL3 times/w, 30 min/session, 40 w76%70% HRmax (Relatively high)Inhibition (Stroop)Working memory (Backward Digit Span, Forward Digit Span)Oken et al.,2006 [83]E1: Yoga (Aerobic)E2: AerobicC: No exercise/ADL1 time/w, 90 min/session, 24 w86.1%N/AInhibition (Stroop)Working memory (Letter number)Cognitive flexibility (Set shifting)Pereira et al.,2020 [84]E: ResistanceC: No exercise/ADL3 times/w, 50–60 min/session, 12 w75.1%N/AInhibition (Stroop)Working memory (Backward Digit Span, Forward Digit Span)Qi et al.,2021 [85]E: Qigong (Aerobic)C: Stretching1–2 times/w, 120 min/session, 12 w79.2%N/AWorking memory (Backward Digit Span)Sexton et al.,2020 [86]E: Cycling (Aerobic)C: No exercise/ADL3 times/w, 30 min/session, 12 w86.6%N/AWorking memory (Backward Digit Span, Forward Digit Span, Verbal fluency, N-back)Cognitive flexibility (Trail Making Test-B) Shatil 2013 [87]E: MultimodalC: Reading3 times/w, 45 min/session, 16 w94.6%N/AInhibition (CogniFit)Working memory (CogniFit)Sink et al.,2015 [88]E: MultimodalC: Health Education3–4 times/w, 50 min/session, 26 w96.6%(Moderate)Inhibition (Flanker)Working memory (N-back)Cognitive flexibility (Task switching)Vaughan et al.,2014 [89]E: MultimodalC: No exercise/ADL2 times/w, 60 min/session, 16 w90%N/AInhibition (Stroop)Working memory (Letter-Number Sequencing, Controlled Oral Word Association Test)Cognitive flexibility (Trail Making Test-B)Walsh et al.,2015 [90]E: Tai Chi (Aerobic)C: No exercise/ADL2 times/w, 30 min/session, 24 w81.2%N/AWorking memory (Letter fluency, Backward Digit Span, Forward Digit Span)Cognitive flexibility (Trail Making Test-B)Wang et al.,2024 [91]E: Dance (Aerobic)C: No exercise/ADL3 times/w, 60 min/session, 12 w93.7%N/AWorking memory (Operation span)Welford et al.,2023 [92]E1: Yoga (Aerobic)E2: AerobicC: Waitlist3 times/w, 60 min/session, 12 w77.6%E2: (Moderate)Working memory (Verbal fluency)Williamson et al.,2009 [93]E: MultimodalC: Health Education2 times/w, 75 min/session, 52 w68.7%N/AInhibition (Stroop)Working memory (Digit Symbol Substitution Test)Wu et al.,2021 [94]E: Tai Chi (Aerobic)C: Telephone Consult3 times/w, 60 min/session, 12 w96%N/ACognitive flexibility (Set shifting)Yang et al.,2019 [95]E: Tai Chi (Aerobic)C: No exercise/ADL3 times/w, 45 min/session, 8 w78.5%N/AInhibition (Flanker)Zhao et al.,2022 [49]E: Dance (Aerobic)C: No exercise/ADL3 times/w, 75 min/session, 12 w92%70% HRmax (Moderate)Inhibition (Stroop)Working memory (N-back)ADL = Activities of Daily Living; C = Control group; E = Exercise group; HRmax = Maximum Heart Rate; HRR = Heart Rate Reserve; N/A = Not Available/Not Applicable; RPE = Rating of Perceived Exertion; RM = Repetition Maximum; min = minute; w = week
A total of 50 studies were included, yielding 263 effect sizes that explored the effect of exercise on executive function. The results of the three-level meta-analysis indicated a small but significant effect (g = 0.139, 95% CI 0.074, 0.205, P<0. 001, 95% PI − 0.001, 0.281). The results of the two-level meta-analysis also demonstrated a significant effect (g = 0.135, 95% CI 0.074, 0.196, P<0. 001, 95% PI 0.074, 0.196), confirming the overall significance of the effect.
An influence analysis was conducted to check for outliers that might affect the results of the meta-analysis. 7 outliers were identified, as shown in Fig. 2. After these outliers were removed, the results of the three-level meta-analysis results(g = 0.155, 95% CI 0.084, 0.226, P <0. 001, 95% PI −0.032, 0.333). The two-level meta-analysis results (g = 0.153, 95% CI 0.088, 0.218, P <0.001, 95% PI 0.088, 0.218) revealed that the overall effects remained significant. To avoid the influence of outliers on the results, these studies were not included in subsequent analyses.The detailed forest plots are provided in Supplementary Material.
Fig. 2Influence analysis
After removing outliers, the heterogeneity analysis yielded Q(df = 254) = 133.309, P >0.05, indicating nonsignificant heterogeneity. In terms of total variance, the sampling variance (level 1) was 97.28%, whereas the within-study variance (level 2) was < 1%. The between-study variance (level 3) was 2.66%. LRT indicated that the between-study variance (level 3) was χ² = 0.280, P > 0.05; the within-study variance (level 2) was nonsignificant, χ² < 0.01, P > 0.05, suggesting that the three-level model did not provide a better fit than the two-level model.
This study conducted subgroup analyses on the basis of the subdomains of executive function, research design (type of control group), participant characteristics (gender, average age, average years of education), and characteristics of exercise interventions (type, intensity, duration, frequency, and duration). Table 5 summarizes the results of all subgroup analyses.Table 5Subgroup analysis of the effects of exercise on executive functionSubgroupkg (95% CI)PF (df1, df2)SubdomainsF(3, 252) = 1.061, P = 0.365Inhibition control950.084−0.026,0.193)0.127working memory1090.208(0.093,0.322)< 0.001cognitive flexibility460.150(0.001,0.299)0.049reasoning/planning60.070(−0.666,0.807)0.815Mean ageF(1, 254) = 0.796, P = 0.37360–691820.134(0.049,0.218)0.002≥ 70740.203 (0.075,0.330)0.002The proportion of femaleF(2, 253) = 1.821, P = 0.163≥ 50%2380.137(0.067,0.206)< 0.001<50%150.359(0.083,0.635)0.010N/A30.543(−0.139,1.226)0.118Years of educationF(2, 253) = 1.897, P = 0.152≤ 12430.322(0.135,0.510)< 0.001>12700.149(0.016,0.283)0.028N/A1430.114(0.018,0.210)0.019FrequencyF(1, 254) = 1.306, P = 0.254≥ 3 times/w1580.188(0.097,0.280)< 0.001<3 times/w980.104 (−0.008,0.217)0.069DurationF(1, 254) = 4.148, P = 0.042≤ 26w2080.193(0.114,0.272)< 0.001>26w480.020(−0.127,0.167)0.785Exercise typeF(2, 253) = 0.636, P = 0.530Aerobic exercise1560.130(0.042,0.218)0.003Multimodal exercise440.153(−0.010,0.317)0.066Resistance exercise560.232(0.077,0.388)0.004Session timeF(1, 254) = 5.081 P = 0.025>60 min44-0.011(-0.171,0.148)0.887≤ 60 min2120.190(0.115,0.264)< 0.001IntensityF(3, 252) = 0.941, P = 0.421Low to relatively low240.126(−0.099,0.352)0.270Moderate690.254(0.117,0.390)< 0.001Relatively high to high530.106(−0.045,0.258)0.169N/A1100.126(0.022,0.229)0.007Types of control groupF(1, 254) = 2.830, P = 0.093Active control group1050.083(−0.026,0.192)0.132Conventional control group1510.205(0.113,0.296)<0.001
The subdomains of executive function were not found to be moderating factors affecting the improvement in executive function through exercise (F = 1.061, P = 0.365). Exercise significantly improved working memory (g = 0.208, 95% CI 0.093, 0.322, P <0.001) and cognitive flexibility (g = 0.150, 95% CI 0.001, 0.299, P = 0.049) but did not significantly improve inhibitory control (g = 0.084, 95% CI −0.026, 0.193, P = 0.127) or reasoning/planning (g = 0.070, 95% CI −0.666, 0.807, P = 0.815).
Both exercise session time (F = 5.081, P = 0.025) and exercise duration (F = 4.148, P = 0.042) were identified as moderating factors affecting the effect of exercise on executive function, whereas exercise type (F = 0.636, P = 0.530), exercise frequency (F = 1.306, P = 0.254), and exercise intensity (F = 0.941, P = 0.421) did not have moderating effects.
The results regarding exercise frequency indicated that a frequency of ≥ 3 times/w (g = 0.188, 95% CI 0.097, 0.280, P <0.001) significantly improved executive function in cognitively healthy older adults. Whereas a frequency of < 3 times/w (g = 0.104, 95% CI −0.008, 0.217, P = 0.069) did not significantly improve executive function.
The results revealed that an exercise duration of ≤ 26 w (g = 0.193, 95% CI 0.114, 0.272, P <0.001) significantly improved executive function in cognitively healthy older adults, whereas an exercise duration of > 26 w (g = 0.020, 95% CI −0.127, 0.167, P = 0.785) did not significantly improve executive function.
In terms of exercise type, aerobic exercise (g = 0.130, 95% CI 0.043, 0.218, P = 0.003) and resistance training (g = 0.232, 95% CI 0.077, 0.388, P = 0.004) significantly improved executive function in cognitively healthy older adults, whereas multimodal exercise (g = 0.153, 95% CI −0.010, 0.317, P = 0.066) did not significantly improve executive function.
The results indicated that a session time of ≤ 60 min (g = 0.190, 95% CI 0.115, 0.264, P <0.001) significantly improved executive function in cognitively healthy older adults, whereas a session time of > 60 min (g = −0.011, 95% CI −0.171, 0.148, P = 0.887) had a nonsignificant negative effect on executive function.
In terms of exercise intensity, moderate-intensity exercise (g = 0.254, 95% CI 0.117, 0.390, P < 0.001) significantly improved executive function in cognitively healthy older adults, whereas relatively high- to high-intensity exercise (g = 0.106, 95% CI −0.045, 0.258, P = 0.169) and low- to relatively low-intensity exercise (g = 0.126, 95% CI −0.099 0.352, P = 0.270) did not significantly improve executive function.
None of mean age (F = 0.796, P = 0.373), proportion of female participants (F = 1.821, P = 0.163), average years of education (F = 1.897, P = 0.152), or type of control group (F = 2.830, P = 0.093) were significant moderators of the effect of exercise on executive function.
The results regarding age revealed that exercise significantly improved executive function in cognitively healthy older adults aged 60–69 years (g = 0.134, 95% CI 0.049, 0.218, P = 0.002) and those aged ≥ 70 years (g = 0.203, 95% CI 0.075, 0.330, P = 0.002).
For the proportion of female participants, exercise significantly improved executive function in studies with a female proportion of ≥ 50% (g = 0.137, 95% CI 0.0667, 0.206, P < 0.001) as well as in studies with a proportion of < 50% (g = 0.359, 95% CI 0.083, 0.635, P = 0.010) among cognitively healthy older adults.
With respect to average years of education, exercise significantly improved executive function in cognitively healthy older adults with an average of ≤ 12 years (g = 0.322, 95% CI 0.135, 0.510, P < 0.001) and those with > 12 years (g = 0.149, 95% CI 0.016, 0.283, P = 0.028).
In types of control group characteristics, exercise significantly improved executive function in cognitively healthy older adults in studies with a conventional control group (g = 0.205, 95% CI 0.113, 0.296, P < 0.001), whereas no significant improvement was observed in studies with an active control group (g = 0.083, 95% CI −0.026, 0.192, P = 0.136).
To eliminate multicollinearity among the moderating variables, multiple regression analysis was performed on the significant moderating variables [96]. The results indicated (Table 6) that exercise duration and session time are significant moderating factors that effectively influence the performance of the dependent variable.
Table 6Multiple Regression Analysis of the Moderating VariablesModerating Variableβ P 95%CIIntercept−0.1330.181(−0.329,0.068)Exercise Duration≤ 26w0.1620.043(0.004,0.319)Session Time≤ 60 min0.2110.026(0.023,0.368)
Publication bias was assessed as shown in Fig. 3. Egger’s test (t = 2.630, P = 0.009) indicated potential publication bias. To evaluate the effect of publication bias on the pooled results, we employed the nonparametric trim-and-fill method. The analysis suggested that 28 potentially unpublished studies might exist. After imputing these studies, the pooled effect size under the random-effects model increased from 0.150 (95% CI: 0.105, 0.195) to 0.219 (95% CI: 0.171, 0.267). This finding indicates that the original results may have underestimated the true intervention effect due to publication bias.
Fig. 3Funnel plots of publication bias
GRADE analysis indicated that the effect of exercise on executive function was moderate-quality evidence (Table 7).
Table 7GRADE quality of evidence evaluationOutcomekQuality RatingRisk of BiasIndirectnessInconsistencyImprecisionPublication BiasExecutive Function49(256)Not SeriousNot SeriousNot SeriousNot SeriousSeriousModerate
Although previous meta-analyses [97–100] have demonstrated the positive effects of exercise on executive function in cognitively healthy older adults, most have focused on single exercise modalities (including aerobic, resistance, and multimodal exercise). This study conducted a three-level meta-analysis of RCTs examining the effects of exercise on executive function in cognitively healthy older adults. The results indicate that exercise has a significant improving effect on global executive function in this population. According to the GRADE evidence quality assessment, the evidence was rated as moderate quality due to potential publication bias, which aligns with the moderate-quality evidence for executive function outcomes reported in the Physical Activity Guidelines for Americans (2018) [101]. Although the three-level model did not demonstrate a statistically significant advantage, maintaining such a model remains meaningful. As the included studies generally reported multiple effect sizes (50 studies with 263 effect sizes in total), the three-level model more accurately reflects the inherent characteristic of effect sizes being nested within studies in meta-analysis. The wide PI that included zero suggests uncertainty regarding its validity when applied to future individual studies. This suggests that the current conclusions may have limited applicability to future single studies. This uncertainty may be related to the following Regarding the quality of the included studies, the PEDro scale scores ranged from 5 to 8, indicating some variation in methodological quality. The characteristics of the study participants also showed significant the average age of older adults ranged from 61 to 83 years, spanning a broad spectrum; the average years of education varied between 9 and 17 years; and some studies included only single-gender participants (either exclusively female or male). In terms of exercise intervention protocols, notable differences existed across exercise types included aerobic exercise, resistance exercise, and multimodal exercise; intensity ranged from low to high; frequency was mostly 1 to 3 times/week; duration varied from 8 to 52 weeks; session time ranged from 30 to 120 min, and exercise adherence rates were between 64.7% and 100%. Additionally, methodological differences, such as the type of control group (conventional control group vs. active control), may also have influenced the pooled results. These wide variations in intervention parameters and participant characteristics may lead to inconsistent dose-response effects, thereby affecting the stability of the pooled results. Future research should conduct more in-depth exploration targeting specific subgroups to clarify which populations are most likely to benefit from exercise interventions, thereby enabling the proposal of personalized exercise programs.
Though a three-level meta-analysis, this study examining the long-term effects of exercise, demonstrates that exercise has selective benefits on specific subdomains of executive function in cognitively healthy older adults. Significant improvements were observed in working memory (g = 0.208) and cognitive flexibility (g = 0.150), while inhibitory control showed a non-significant improving trend (g = 0.084). For reasoning/planning (g = 0.070), only 2 studies incorporating relevant tasks (k = 6) were available, resulting in a wide confidence interval and a statistically non-significant effect despite a positive trend. Future research should include more high-quality RCTs to further explore the effect of exercise on reasoning/planning abilities in cognitively healthy older adults.
The frontal lobe is the primary brain region governing executive function. Brain areas associated with working memory include the left anterior cingulate cortex, left premotor cortex, and right inferior frontal gyrus, while those related to cognitive flexibility are distributed across the bilateral frontal cortex [102]. This neuroanatomical distribution may underlie the differential effects of exercise on various executive subdomains. Research indicates that exercise can increase the volume of the frontal lobe and hippocampus. The hippocampus is a complex region deep within the brain that plays a key role in memory formation and retrieval. Exercise may exert its benefits through mechanisms such as neuroplasticity, increased cerebral blood flow, and modulation of neurochemical pathways related to executive function [103]. Furthermore, one study found that 12 weeks of high-intensity resistance training effectively remodeled the prefrontal-hippocampal circuit, thereby optimizing resource allocation in working memory and the flexibility of executive control [104]. The functional significance of this neural remodeling is further evidenced in individuals with high aerobic fitness, who demonstrate stronger prefrontal activation, shorter reaction times, and higher accuracy during high cognitive load tasks, possibly indicating that aerobic fitness optimizes cognitive performance by enhancing the compensatory allocation of neural resources [105]. In contrast, the neural networks supporting inhibitory control and reasoning/planning are likely more complex or widely distributed, potentially requiring longer durations or more targeted training approaches for significant improvement. The conclusions of this study are partially contradictory to those of another traditional meta-analysis on the long-term effects of exercise on executive function in older adults [14], which reported significant improvements in working memory (g = 0.127), cognitive flexibility (g = 0.511), and inhibitory control (g = 0.136) compared to control groups, but no significant effect on reasoning/planning (g = 0.433). A potential reason for this discrepancy is that the aforementioned study extracted multiple effect sizes from single studies without accounting for the correlations between these effect sizes, potentially leading to overly optimistic results. More high-quality research is needed in the future to substantiate the differential effects of exercise on the various subdomains of executive function.
The characteristics of exercise interventions include type, intensity, duration, session time and frequency. This study revealed that session time and exercise duration may serve as moderators influencing improvements in executive function. Although some potential moderating variables—such as exercise intensity, frequency, and type—were not statistically significant, they may still offer valuable insights for future research and should be considered in future studies and practical applications. This study revealed that both aerobic exercise and resistance exercise can effectively enhance executive function in older adults. The benefits of aerobic and resistance exercise for improving executive function in older adults have been well documented [65, 106, 107], and the current study reaffirms these findings, underscoring the importance of these exercise modalities for the aging population.Some studies suggest that aerobic exercise may enhance executive function in older adults by increasing hippocampal activity and promoting the expression of brain-derived neurotrophic factor (BDNF), which is associated with the volume and connectivity of the prefrontal and temporal lobes [14]. Resistance training has been shown to increase the serum levels of insulin-like growth factor 1 (IGF-1) in older adults. IGF-1 crosses the blood‒brain barrier and exerts neuroprotective effects, including promoting myelin formation and facilitating remyelination following pathological damage [36, 108, 109]. However, compared with the above two forms of exercise, multimodal exercise shows a certain but non-significant improvement in the executive function of older adults. The reason for this may be attributed to the varied combinations of multimodal exercise protocols across studies. The multi-modal exercise programs included in this review predominantly featured moderate to relatively high intensity, with session durations concentrated between 50 and 90 min. Only three multimodal studies employed session durations shorter than 50 min. Excessively high intensity or prolonged exercise sessions may easily induce fatigue in older adults [110], thereby reducing exercise adherence and subsequently diminishing the potential benefits of exercise on executive function. Furthermore, the included studies utilized a wide range of outcome measures for assessing executive function. Since different tasks target distinct aspects of executive function, this diversity may have reduced statistical power in the pooled analysis.
This study found that engaging in moderate-intensity exercise for >3 times/week, with each session lasting ≤ 60 min and for a duration of ≤ 26 weeks, can effectively enhance executive function. Regarding exercise frequency, a meta-analysis by Ye et al. [111] indicated that aerobic exercise performed 3–7 times/week can improve cognitive flexibility and working memory in middle-aged and older adults, while aerobic exercise 3–4 times/week can enhance inhibitory control. Combined with the findings of this study, it suggests that exercising ≥ 3 times/week may yield superior improvements in executive function among older adults. In terms of exercise duration, interventions lasting ≤ 26 weeks showed a significant improving effect on executive function, while those > 26 weeks, though still beneficial, did not reach statistical significance. This may be related to the age-related cognitive decline in older adults [31], with executive function decline being a primary manifestation of this process [112]. As the exercise duration extends, the degree of cognitive decline in older adults may deepen, partially offsetting the improving effects of exercise. Additionally, adherence is a critical factor influencing the effectiveness of exercise interventions, with intervention adherence positively correlated with cognitive improvement [113]. A 12-month remotely supervised study [114] revealed that only 54% of older adults maintained exercise ≥ 3 times/week, while 28.6% exercised only 1–2 times/week, highlighting the practical challenges of adhering to long-term regular exercise. The study by Northey et al. [113] suggested that exercise programs > 26 weeks significantly improved cognition in adults aged ≥ 50 years, which is inconsistent with the findings of this study. The discrepancy may be due to the older population in the present study (mean age > 60 years), who are at higher risk of cognitive decline and may have relatively lower brain plasticity.
Exercise duration and intensity are important factors influencing the effectiveness of a single exercise session. The effect of exercise intensity on executive function follows an inverted U-shaped dose-response relationship, while a session time of ≤ 60 min was found to be optimal for improving executive function in older adults. This study also observed that excessively long exercise sessions time(> 60 min/session) may have certain negative effects on executive function. Sessions time ≤ 60 min may adjust the physiological-cognitive level of older adults to an appropriate state. Exercise can increase the concentrations of substances such as BDNF, serotonin, and dopamine in the human body [115–117], but with delayed and threshold effects. Studies have shown that neurotransmitter levels, including norepinephrine, peak within 30–90 min after exercise [118]. Some research indicates that even 20 min of exercise can significantly improve cognitive function in older adults [119]. However, excessively prolonged or high-intensity exercise requires greater activation of the premotor and supplementary motor areas of the brain, which may affect prefrontal cortex activation and impair cognitive performance [120].Additionally, high-intensity exercise may cause fatigue, dehydration, and excessive depletion of self-regulatory resources, all of which can temporarily impair subsequent cognitive performance [121, 122]. According to the arousal theory hypothesis, moderate-intensity exercise optimally stimulates the release of catecholamines in older adults, enhances biological arousal in the central nervous system, and facilitates more efficient cognitive resource allocation. This makes exercise more effective in improving cognition than low- or high-intensity exercise [123].
This study examined the effects of three participant characteristics (gender, mean age, and mean years of education) and one study design characteristic (type of control group) on executive function outcomes. The results showed that exercise significantly improved executive function in cognitively healthy older adults aged 60–69 and ≥ 70 years, with the effect size increasing with age. Exercise produced a larger effect size in cognitively healthy older adults aged ≥ 70 years (g = 0.203) compared to those aged 60–69 years (g = 0.134), which aligns with the findings of Boucard et al. [124]. This may reflect a compensatory neuroplastic mechanism [125], where the brain’s sensitivity to external stimuli significantly increases as it approaches a critical threshold of synaptic integrity loss due to aging. This study also found that exercise significantly improved executive function in studies with both ≥ 50% and < 50% proportion of female participants (both p < 0.05), but a larger effect size was observed in studies with < 50% proportion of female participants (g = 0.359 vs. 0.137). However, previous research has indicated that proportion of females tend to derive greater cognitive benefits from exercise [54]. The discrepancy with the present findings might be attributed to the relatively small number of effect sizes (k = 6) from studies with < 50% proportion of female participants, potentially leading to an overestimation of the effect size in this subgroup. It is worth noting that exercise yielded a small-to-medium effect size (g = 0.322) on executive function in older adults with average education ≤ 12 years. For cognitively healthy older adults with > 12 years of education, the baseline level of executive function decline might be lower, and a ceiling effect could limit the potential for further improvement through exercise. In contrast, a more pronounced compensatory effect might be observed in those with ≤ 12 years of education. The research by Foubert-Samier et al. [126] demonstrated that educational experience primarily influences the accumulation of cognitive reserve. Other findings indicate that higher educational attainment serves as a protective factor against MCI and DE [127]. Individuals with longer education typically possess higher cognitive reserve and more developed neural networks. Higher education levels are associated with slower rates of cognitive decline [128], thereby helping to preserve cognitive function [129]. Furthermore, Fyock et al. [130] found that subjective memory complaints in older adults with advanced age, low education levels, and depressive tendencies are more likely to be indicative of underlying executive function deficits.
This study also revealed that when the control group involved an active intervention, it may have indirectly produced cognitive benefits similar to those of exercise by enhancing participant compliance and promoting health-related behavior change, thereby reducing the effect size difference between the intervention and control groups. This finding aligns with that of Zhang et al. [131]. Future studies should adopt prospective designs and treat the control group type as a prespecified stratification variable to assess the clinical value of exercise interventions more accurately.
This study employed a three-level meta-analytic approach to systematically evaluate the effects of exercise on executive function in cognitively healthy older adults. By incorporating within-study variance, it accounts for the dependency among multiple effect sizes from the same study, thus addressing the issue of nonindependence. With this method, effective exercise protocols were identified, and how age, education level, and sex moderate the effects of exercise on executive function was explored. However, several limitations remain. Despite setting objective inclusion and exclusion criteria, Egger’s test indicated the presence of publication bias. Some included studies failed to report key variables such as exercise intensity or participants’ educational background, resulting in missing data that may compromise the robustness of the conclusions. Moreover, owing to the nature of exercise interventions, the blinding of participants and implementers is often not feasible, potentially affecting the objectivity of outcome assessments. In addition, the use of active control groups involving balance or stretching interventions may have led to an underestimation of the true effects of exercise. The limitation pertains to the classification methods for mean age and average years of education. We adopted a dichotomous grouping for age (60–69 years vs. ≥70 years) and education (using 12 years as the cutoff). While this approach helps distinguish older adults at different life stages and explore differences in cognitive reserve levels, such classification criteria lack a unified consensus in the academic field. This artificial dichotomization may oversimplify the complex influences of age and education as continuous variables and potentially mask heterogeneity within the groups.
Exercise is an effective means of enhancing global executive function, working memory, and cognitive flexibility in cognitively healthy older adults. Compared with the usual control conditions, both aerobic and resistance training at moderate intensity ≥ 3 times/week, for ≤ 60 min/session, over a duration of ≤ 26 weeks—significantly improved these cognitive domains. The beneficial effects were particularly pronounced among older individuals and those with fewer years of education. Nonetheless, future interventions should tailor exercise protocols to the specific needs and physical conditions of older adults to optimize their cognitive outcomes.
Below is the link to the electronic supplementary material.
Supplementary Material 1.