Effect of dual task-based training on motor and cognitive function in stroke patients: a systematic review and meta-analysis of randomized controlled trails

Abstract

Background

The clinical application of dual-task training based on movement and cognition in stroke population is still controversial. This study used a systematic review and meta-analysis to compare the effects of dual-task exercise versus single-task training including cognitive-only, exercise-only, and usual rehabilitation for motor function and cognitive function in stroke patients.

Methods

Extensive electronic database search (from inception to November 27, 2024) was conducted in 8 databases to identify randomized controlled trials that investigated the effects of dual task-based training on motor and cognitive function in stroke patients.

Results

30 RCTs involving 1,588 people were included in the analysis. The study found that compared with the control group, dual-task cognitive motor training can improve the walking performance of stroke patients (WMD = 3.19, 95%CI: 2.26, 4.12), the recovery of lower limb motor function (WMD = 2.78, 95% CI: 1.38, 4.18), cognitive function (WMD = 2.93, 95% CI: 0.95, 4.91) and mental state (WMD = 3.39, 95% CI: 0.06, 6.72), and the functional state of activities of daily living (WMD = 7.47, 95% CI: 3.97, 10.96). Subgroup analyses showed that cognitive-motor dual-task training was more likely to have a clinical effect after at least 3 weeks of intervention.

Conclusions

Dual-task training significantly improves walking ability, lower limb motor function, cognitive function, mental status, and activities of daily living in stroke patients. No significant effects were found for basic mobility and gait speed. These findings support its clinical application, with personalized programs recommended based on patient needs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12883-025-04305-2.

Introduction

According to the World Stroke Organization (WSO), the global economic cost of stroke is more than $891 billion [1, 2]. Similarly, stroke has become one of the leading causes of disability in adults today, placing a huge burden on individual families and severely affecting cognitive and motor function of patients [3]. Furthermore, stroke is the fifth most common cause of death in the updated stroke statistics for 2022 from the American Heart Association in collaboration with the National Institutes of Health concerning [4]. The latest figures show that 795,000 people experience new or recurrent strokes each year [5]. Stroke-related cognitive deficits have been shown to be associated with death [6]. Not to mention the title that stroke is known as one of the leading causes of long-term disability [7, 8]. Besides, up to 47% of all hospitalized stroke patients may face a very common and serious problem called falls [9]. Falls tend to occur when lower limb weakness, loss of balance and reduced walking speed occur during stroke development [10]. Especially when people with stroke have cognitive impairment, falls are often more likely to occur [11]. In this case, in addition to the longer recovery time cost and heavier economic cost of stroke patients, it may also result in secondary complications such as fractures and depression in stroke patients, exacerbating the survival dilemma of patients [12]. Therefore, it is necessary to pay attention to motor and mental function in the treatment of stroke.

Dual task training can provide stroke patients with the combination of motor tasks and cognitive functions needed to restore functional independence in daily living activities after stroke, such as maintaining posture control, standing or walking [13]. In view of this condition, dual-task exercise can be a good choice for patients in stroke. The existing research aims to explore whether dual-task walking training can effectively improve balance and walking function and cognitive function in old patients [14–16]. Zhou et al.‘s study found that dual-task exercise can improve balance, gait, and upper limb function in patients with chronic stroke [16]. Ye et al.‘s study aimed to explore the application of dual-task exercise in the elderly group of cognitive impairment, and found that dual-task training can effectively improve cognitive function, physical function and depression in older adults with cognitive dysfunction [15]. Although studies have been published on the use of dual-task training in stroke populations, this view is still controversial, and existing meta-analyses suggest that dual-task training does not appear to be superior to individual exercise training or cognitive training in the corresponding motor function and cognitive function [17].

In addition, a large number of new clinical studies have been published since 2020, particularly in Asian countries, which were underrepresented in earlier meta-analyses. More importantly, previous reviews have rarely conducted detailed subgroup analyses based on key clinical factors, such as intervention duration, stroke stage, type of control group (cognitive vs. physical vs. conventional rehabilitation), or regional differences. These elements may influence the effectiveness of dual-task training, yet have not been systematically explored. Therefore, the present study aims to update the current evidence base by incorporating recently published randomized controlled trials and, more critically, to conduct comprehensive subgroup analyses. These analyses seek to clarify under what conditions dual-task training is most beneficial and for which patient subgroups, thereby enhancing its clinical utility and guiding personalized rehabilitation strategies for stroke patients.

Method

This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement [18]. It was preregistered in the International Prospective Register of Systematic Reviews(Prospero Number: CRD42025634746).

Literature search and eligibility criteria

Searches were conducted by information specialists in the following databases up to November 27th, 2024, four English databases (PubMed, Cochrane, Embase, Web of Science) and four Chinese databases (the China National Knowledge Infrastructure, China Science and Technology Journal Database, the Chinese Biomedical Literature Database and Wanfang Data). Reference lists of all studies that were found to be eligible for this review were hand searched for further eligible trials. Language was restricted to English, Chinese, and Korean. No restrictions were set for publication date. Additionally, in order to obtain more studies for meta-analysis, the literatures were searched manually, which included relevant reviews and reference.

Search strategy

A detailed search strategy was developed based on three core concepts:

1. Stroke (“stroke”, “cerebral infarction”, “brain ischemia”, “cerebrovascular accident”);

2. Dual-task training (“dual task”, “cognitive-motor training”, “dual-task exercise”, “motor-cognitive rehabilitation”);

3. Study design (“randomized controlled trial”, “RCT”, “clinical trial”).

Boolean operators such as “AND” and “OR” were used to combine these terms appropriately across databases. For Chinese databases, equivalent search terms were translated and adapted by native-speaking researchers using standardized terminology (“脑卒中”, “双任务训练”, “随机对照试验”). Korean terms were translated similarly using professional medical equivalents, though no eligible Korean-language articles were ultimately included.

The exclusion criteria of this study were strictly in accordance with PICOS principles of evidence-based medicine (population, intervention, control, outcome, type of study).

Inclusion criteria

• P: Patients diagnosed with stroke, regardless of stroke type, lesion location, or time since onset.

• I: Dual-task training interventions involving the simultaneous execution of motor and cognitive tasks. Motor components typically included gait training, balance exercises, or functional mobility training, while cognitive components included tasks such as serial subtraction, attention switching, verbal fluency, or working memory tasks.

• C: Control groups received either conventional rehabilitation, single-task motor training (without cognitive components), or single-task cognitive training (without motor activity).

• O: Primary outcomes included motor function (e.g., lower limb recovery, walking performance, mobility, balance) and cognitive function (e.g., executive function, global cognition), assessed using validated tools such as the Fugl-Meyer Assessment, Berg Balance Scale, TUG test, MMSE, and MoCA.

• S: Randomized controlled trials (RCTs).

Exclusion Criteria

1. Duplicate literature.

2. Non-original research articles, including reviews, guidelines, case reports, protocols, trial registrations, letters, and conference abstracts.

3. Basic research includes animal experiments, cell-related studies, and in vitro studies.

4. Protocols, Registers, Reviews.

The following definitions of the outcomes included in this review are as follows:

In terms of physical function, the degree of walking improvement was assessed by Berg balance scale (BBS) [19, 20], the basic activity ability was assessed by the Timed Up and Go (TUG) test [21, 22], which reflects a person’s capacity to stand up, walk, turn, and sit down-core components of independent functional mobility in daily life. The TUG test is a widely used and validated tool for assessing basic mobility and transfer ability, particularly in stroke rehabilitation. Gait function was assessed by the 10-meter walk test [23]; Recovery of lower limb motor function in the patient was evaluated by the Fugl-Meyer Lower limb extremities (FMA-LE) [24]; Cognitive function and mental status were assessed by the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MOCA), respectively; The Barthel Index (BI) was used to evaluate the function status of activities of daily living [25]. All of the outcome measures selected in this review-BBS, TUG, FMA-LE, MoCA, MMSE, and BI-are standard, validated instruments that are widely used in stroke rehabilitation research. They comprehensively capture core aspects of physical and cognitive recovery, and are endorsed by numerous clinical studies as reliable tools for evaluating rehabilitation effectiveness in stroke populations.

Literature screening and data extraction

The two review authors (Y.J., C.M.) conducted independent screening of the retrieved literatures. The screening process included removing duplicate literatures, eliminating literatures that did not meet the exclusion criteria, and screening the literatures that met the initial screening by reading the title abstract. In addition, the re-screening of the full text was conducted in strict accordance with the principles of evidence-based medicine PICOS to screen the population, intervention, control, outcome, and study type. Controversy in the period are determined by a third-party adjudication.

A pre-piloted form was used to extract data from the included studies. Extracted information included: participant demographics, details of sample size, sample source, disease diagnostic criteria, age, gender, details of the dual-task training intervention and the control intervention including specific combinations of interventions and intervention period; details of the outcomes pre and post intervention pertaining to both control intervention and dual-task training intervention. Back-to-back data extraction was carried out by two review authors at the same time, and the third review author checked and dealt with relevant differences.

Data analyses

Stata15.1 was used for meta-analyses. The outcomes available for meta-analyses (FMA-LE, BI, BBS, the TUG, MMSE, MOCA) were expressed as weighted mean differences (WMDs) with 95% confidence intervals (CI). As we expected some heterogeneity in the trial designs, a random-effect model was used. Funnel plots and egger tests were used to detect publication bias. The robustness of the results was judged by sensitivity analysis. Heterogeneity testing was performed using Cochrane Q test and the I² test [26]. Subgroup analyses were used for heterogeneous source analyses.

Risk of bias assessment

Two reviewers independently (Y.J., C.M.) assessed the risk of bias in the included studies using the revised Cochrane Collaboration’s tools for assessing risk of bias (RoB 2), which covers five domains including Randomization process, Deviations from intended interventions, Missing outcome data, Measurement of the outcome, Selection of the reported result (each domain was scored as high, low, or unclear). The risk of bias is generated by an algorithm, based on answers to the signaling questions and is expressed either by ’low’ or ’high’ or by the intermediate option ’some concern of’ risk of bias [27]. Disagreements were resolved by discussion or, if necessary, by consultation with a third investigator.

Results

Study selection

A total of 2,155 records were initially identified through searches in eight databases, including four English-language databases (PubMed = 337, Embase = 482, Cochrane = 254, and Web of Science = 904) and four Chinese-language databases (Wanfang = 77, CNKI/Zhiwang = 41, VIP/Wepu = 33, and SinoMed = 27). After removing 862 duplicate records and excluding 761 records through automated screening tools, 531 records remained for initial screening. Following title and abstract screening, 256 articles were deemed potentially eligible and retrieved for full-text evaluation. Of these, 161 full-text articles were excluded for the following reasons: the intervention did not meet the inclusion criteria (n = 24), the outcomes were not relevant to the study objectives (n = 14), the study design was not a randomized controlled trial (n = 8), or the study data were unavailable (n = 19). Ultimately, 30 studies met all eligibility criteria and were included in the final systematic review and meta-analysis. The complete study selection process is presented in Fig. 1.

Fig. 1.

Flowchart of study selection process

Study characteristics

Population characteristics

All included studies were of random controlled trials design and had been conducted in the following countries: 1 from Portugal, 2 from Pakistan, 2 from America, 6 from Korea, 19 from China. The final 30 studies involving 1,588 people were included in the analysis, of which 620 (39%)were women and 945 (59.5%) were men, with a mean age of 58.65. Of the population included in this study, 14 studies reported the duration of stroke in patients, 10 studies actively reported patients with first onset, and 14 studies provided information on stroke classifications. The content of the motor-cognitive interventions mainly focus on gait training and exercise, The control group was mainly rehabilitation, with 2 studies for cognitive training alone and 14 studies for exercise training alone. The duration of the intervention ranged from 2 weeks to 12 weeks. 9 studies involved the FMA-LE, 16 studies utilized the BBS, nine studies included the BI, 11 studies employed the TUG, 6 studies used the MOCA, 6 studies involved the MMSE, and 8 studies incorporated the 10 m walk test. baseline information on the included studies is shown in Table 1.

Table 1.

Study characteristics

Study	Intervention	Control	Sample size	Female	Male	Mean age	First stroke	Post-stroke Phase	Stroke type
li2024	Dual-task training	motion traing	68	26	42	73.43	na	< 6months	na
wang2024	Dual-task training	motion traing	100	21	71	62	na	na	na
shi2024	Dual-task training	rehabilitation	90	28	62	56.76	na	na	na
huang2024	Dual-task training	rehabilitation	30	10	18	48.54	yes	≥ 6months	Chronic stage stroke patients
Zafar2024	Dual-task training	cognitive training	30	16	14	52.6	na	na	Chronic and subacute
zhou2023	Dual-task training	rehabilitation	108	43	65	67.44	yes	< 6months	na
yang2023	Dual-task training	na	40	12	28	56.75	na	na	na
pei2023	Dual-task training	rehabilitation	60	29	31	51.8	yes	na	na
ma2023	Dual-task training	rehabilitation	32	19	13	62.47	yes	< 3months	Acute ischemic stroke
Yeh2023	Dual-task training	motion traing	42	7	35	62.22	na	≥ 6months	na
yang2022(2)	Dual-task training	na	40	12	28	56.75	na	na	na
yang2022(1)	Dual-task training	rehabilitation	260	122	138	58.29	na	na	na
zheng2021	Dual-task training	rehabilitation	96	51	45	61.91	yes	na	na
yang2021	Dual-task training	motion traing	67	29	38	60.17	na	> 6months	na
wang2021	Dual-task training	rehabilitation	31	8	23	60.9	yes	na	chronic stroke
song2021	Dual-task training	motion traing	40	10	30	48.01	yes	≥2weeks	na
liu2021	Dual-task training	rehabilitation	60	27	33	67.2	yes	≤ 6months	na
fu2020	Dual-task training	motion traing	30	8	22	51.5	yes	1 month ≤ Duration < 6months	stroke in the subacute phase
Iqbal2020	Dual-task training	motion traing	64	30	34	58.58	na	na	chronic stroke
zhuang2017	Dual-task training	rehabilitation	20	6	14	60.9	na	<3 months	na
Song2015	Dual-task training	motion traing	40	19	21	56.24	na	na	na
lee2017	Dual-task training	rehabilitation	22	5	17	57.85	na	> 6months	chronic stroke patients
kim2016	Dual-task training	motion traing	14	3	11	59.7	na	na	chronic stroke patients
Bang2012	Dual-task training	motion traing	14	6	8	69.8	na	≥ 6months	chronic stroke patients
kim2011	Dual-task training	motion traing	24	11	13	53.75	na	na	na
Maeneja2023	Dual-task training	motion traing	34	15	19	56.06	na	na	subacute to chronic ischemic stroke
sun2022	Dual-task training	cognitive training	40	8	25	57.25	yes	na	na
Plummer2022	Dual-task training	motion traing	36	17	19	57	na	na	na
Kannan2019	Dual-task training	rehabilitation	30	11	13	59.25	na	> 6months	chronic stroke
kim2018	Dual-task training	motion traing	26	11	15	54.39	na	≥ 6months	chronic hemiparesis

In terms of dual-task training, although all included studies combined motor and cognitive components, the specific formats varied. The motor components primarily involved gait training, lower limb strengthening, and balance exercises. Cognitive components included tasks such as serial subtraction, verbal recall, auditory stimulus response, and dual-response coordination. According to intervention structure, we broadly categorized the dual-task protocols into three types: (1) “Simultaneous dual-task execution: motor and cognitive tasks performed at the same time (walking while performing calculations).” (2) “Motor tasks with cognitive interference: cognitive distractions embedded during physical tasks (stepping while responding to verbal cues).” (3) “Progressive dual-task training: increasing difficulty of cognitive tasks in parallel with motor progression.” These classifications offer insight into the heterogeneity of dual-task interventions and provide context for the subgroup analyses performed.

Results meta-analysis

Physical functions

The degree of walking improvement

16 studies were included in the meta-analysis for the outcome BBS, see Fig. 2a. In terms of overall risk of bias, 9 studies were assessed as some concerns, and 7 studies were considered to have low risk of bias. The random-effects model showed a significant mean difference in BBS of 3.19 (95% CI 2.26, 4.12) in favor of the motor-cognitive training in contrast to control groups.

Fig. 2 .

Flowchart of study selection process. A Berg Balance Scale (BBS) for walking improvement. B Timed Up & Go test (TUG) for basic mobility. C 10-min test for gait function

The basic mobility

11 studies were included in the meta-analysis for the TUG, see Fig. 2b. In terms of overall risk of bias, 4 studies were assessed as some concerns, and 7 studies were considered to have low risk of bias. The random-effects model showed no significant mean difference − 1.64 (95% CI −4.24, 0.96) in TUG between motor–cognitive training and control.

The gait function

8 studies were included in the meta-analysis for the 10-min test, see Fig. 2c.In terms of overall risk of bias, 4 study were assessed as some concerns, and 4 studies were considered to have low risk of bias. The random-effects model showed no significant mean difference in 10-mintest of 0.09 (95% CI −0.36, 0.53) in favor of the motor–cognitive training in contrast to control groups.

The recovery of lower limb motor function

9 studies were included in the meta-analysis for the FMA-LE, see Fig. 3. In terms of overall risk of bias, 3 studies were assessed as some concerns, and 6x studies were considered to have low risk of bias. The random-effects model showed a significant mean difference in FMA-LE of 2.78 (95% CI 1.38, 4.18) in favor of the motor-cognitive training in contrast to control groups.

Fig. 3.

Meta-analysis results of the Recovery of lower limb motor function (FMA-LE)

Cognitive function and mental status

Cognitive function

6 studies were included in the meta-analysis for the outcome MOCA, see Fig. 4a. In terms of overall risk of bias, all 6 studies were considered to have low risk of bias. The random-effects model showed a significant mean difference in MOCA of 2.93 (95% CI 0.95, 4.91) in favor of the motor-cognitive training in contrast to control groups.

Fig. 4.

Meta-analysis results of cognitive function and mental status. A Montreal Cognitive Assessment (MOCA) for cognitive function; B Mini-Mental State Examination (MMSE) for mental status

Cognitive status

6 studies were included in the meta-analysis for the MMSE, see Fig. 4b. In terms of overall risk of bias, 1 study was assessed as some concerns, and 6 studies were considered to have low risk of bias. The random-effects model showed a significant mean difference in MMSE of 3.39 (95% CI 0.06, 6.72) in favor of the motor-cognitive training in contrast to control groups.

The function status of activities of daily living

9 studies were included in the meta-analysis for the outcome BI, see Fig. 5. In terms of overall risk of bias, 3 studies were assessed as some concerns, and 6 studies were considered to have low risk of bias. The random-effects model showed a significant mean difference in BI of 7.47 (95% CI 3.97, 10.96)in favor of the motor-cognitive training in contrast to control groups.

Fig. 5.

Meta-analysis results of the function status of activities of daily living (Barthel Index, BI)

Sensitivity analysis and publish bias

Sensitivity analysis of all results showed robust results. The P-value of the Egger test was greater than 0.05 for all outcomes, suggesting no publication bias.

Risk of bias

We used the RoB2 tool to assess risk of bias for each of the included reports. The results of risk assessment of bias for the 30 included studies are shown in Supplementary Fig. 1. Among the bias generated during randomization, 9 studies were assessed as medium risk because they were not randomly assigned or group concealed, and the remaining 21 studies were assessed as low risk. In terms of risk bias in deviation from established interventions and measurement outcomes, 1 study each was assessed as likely risk due to lack of description of blindness, and the remaining studies were assessed as low risk. All studies were assessed as having a low risk for missing outcome data and other source bias assessments. A summary of the assessments is provided in Supplementary Fig. 2. 66.7% of the reports (10/30) were considered to have some concerns in terms of overall risk of bias, and 33.3% (20/30) reports were assessed as having some concerns. Taken together, the included literature had a low risk of bias.

Subgroup analysis

We performed subgroup analyses to examine heterogeneity by analyzing differences in duration of intervention, country, type of comparator for FMA-LE, BBS, BI and the TUG. In the following subgroup analyses, “dual-task training” refers to the intervention group combining motor and cognitive tasks. The comparator group interventions are categorized as “cognitive training” (single-task cognitive intervention) and “rehabilitation training” (single-task motor intervention). To avoid confusion, “training” is replaced with “dual-task training” throughout this section.

Subgroup of FMA-LE

The examine heterogeneity by analyzing differences in duration of intervention was categorized into four subgroups: 3weeks (WMD = 3.40, 95% CI:1.36, 5.44), 4weeks (WMD = 1.72, 95% CI: −0.88, 4.31), 6weeks (WMD = 3.38, 95% CI: 1.89, 4.87), and 8weeks (WMD = 3.82, 95% CI: 1.82, 5.81), no heterogeneity in the effect size difference in the four subgroups was observed (I² = 0%). The examine heterogeneity by analyzing differences in country was categorized into China (WMD = 2.85, 95% CI: 1.40, 4.31) and korea (WMD = 1.43, 95% CI: −3.43, 6.29), no heterogeneity in the effect size difference in the two groups was observed (I² = 0%). The examine heterogeneity by analyzing differences in the type of comparator was categorized into training (WMD = 1.77, 95% CI: −0.18, 3.71) and rehabilitation (WMD = 3.49, 95% CI: 2.12, 4.85), with high heterogeneity in the effect size difference in the two subgroups was observed (I² = 75.4%). The results showed that motor cognitive dual-task training could significantly improve the motor function of stroke patients in the intervention cycle subgroup (3weeks, 6weeks, 8weeks), national subgroup (China) and control subgroup (rehabilitation).

Subgroup of BBS

The examine heterogeneity by analyzing differences in duration of intervention was categorized into five subgroups: 3weeks (WMD = 9.07, 95% CI: 5.89, 12.25), 4weeks (WMD = 3.20, 95% CI: 1.57, 4.84), 6weeks (WMD = 2.47, 95% CI: 0.51, 4.42), 8weeks(WMD = 3.53, 95% CI: 1.62, 5.45), 12weeks (WMD = 1.70, 95% CI: 0.68, 2.72), no heterogeneity in the effect size difference in the five subgroups was observed (I² = 0%). The examine heterogeneity by analyzing differences in country was categorized into China (WMD = 3.96, 95% CI: 2.73, 5.19), Pakistan (WMD = 2.09, 95% CI: 1.13, 3.05), Korea (WMD = 2.18, 95% CI: −0.51, 4.87), Portugal (WMD = 1.70, 95% CI: 0.68, 2.72), America (WMD=−1.44, 95% CI: −5.43, 2.55), no heterogeneity in the effect size difference among the five subgroups was observed (I² = 0%). The examine heterogeneity by analyzing differences in the type of comparator was categorized into training (WMD = 2.75, 95%CI: 1.17, 4.34), cognition (WMD = 2.09, 95%CI: 1.13, 3.05), and rehabilitation (WMD = 3.84, 95% CI: 2.77, 4.91), with high heterogeneity in the effect size difference in the three subgroups was observed (I² = 72.3%). Results showed that dual-task training in motor cognition was observed in the intervention cycle subgroup (3weeks, 4weeks, 6weeks, 8weeks, 12weeks), national subgroups (China, Pakistan, Portugal), control subgroups (training, cognition, and rehabilitation) can make walking stroke patients to improve the level of improved significantly.

Subgroup of BI

The examine heterogeneity by analyzing differences in duration of intervention was categorized into five subgroups: 2weeks (WMD = 7.50, 95%CI: −5.74, 20.74), 3weeks (WMD = 9.53, 95%CI: 4.64, 14.42), 4weeks (WMD = 7.47, 95%CI: −0.24, 15.17), 6weeks (WMD = 4.54, 95%CI: 2.38, 6.71), 8weeks (WMD = 9.50, 95%CI: −0.07, 19.06); no heterogeneity in the effect size difference among the five subgroups was observed (I² = 0%).

The examine heterogeneity by analyzing differences in the type of comparator was categorized into training (WMD = 8.01, 95% CI: 3.27, 12.75) and rehabilitation (WMD = 6.87, 95% CI: 0.66, 13.07); with high heterogeneity in the effect size difference in the three subgroups was observed (I² = 89.7%). The results showed that dual-task training of motor cognition could improve the function status of activities of daily living in stroke patients in the intervention cycle subgroup (3weeks,6weeks) and the control subgroup (training, rehabilitation).

Subgroup of the TUG

The examine heterogeneity by analyzing differences in duration of intervention was categorized into four subgroups: 4weeks − 0.72, 95% CI: −4.96, 3.52, 6weeks − 2.09, 95% CI: −7.96, 3.78, 8weeks − 2.17, 95% CI: −3.68, −0.66, 12weeks − 4.84, 95% CI:−8.02, −1.66; no heterogeneity in the effect size difference in the five subgroups was observed (I² = 0%). The examine heterogeneity by analyzing differences in country was categorized into China − 0.61, 95% CI: −0.94, −0.28, Pakistan 0.47, 95% CI: −3.10, 4.04, Korea − 0.99, 95% CI: −2.22, 0.24, America 0.41, 95% CI: −0.19, 1.01; with heterogeneity results of each subgroup were (I² = 40.2%), (I² = 98.0%), (I² = 56.3%) and (I² = 0.0%). The examine heterogeneity by analyzing differences in the type of comparator was categorized into training 0.96, 95% CI: −3.76, 5.69, rehabilitation − 3.99, 95% CI: −5.46, −2.51, cognition − 1.68, 95% CI: −2.56, −0.80; with high heterogeneity in the effect size difference in the three subgroups was observed (I² = 95.3%). Dual-task motor cognition training can effectively improve the basic activity ability of stroke patients in the intervention cycle subgroup (8weeks,12weeks), national subgroup (China), and control subgroup (cognition, rehabilitation). Detailed results of subgroup analysis are shown in supplements.

Discussion

Our study found that, compared with the control group, dual-task cognitive motor training can improve walking improvement, the recovery of lower limb motor function, cognitive function and mental state, and the functional status of activities of daily living in stroke patients.

Through subgroup analysis, we also found that the motor function in patients with dual-task cognitive motor training was significantly improved when the intervention period lasted 3 weeks or more, compared with the control group. Dual task cognitive motor training significantly improved lower limb partial motor function, walking improvement degree, functional status of activities of daily living, and the basic activity ability compared with the conventional rehabilitation group; dual task cognitive motor training can significantly improve the degree of walking improvement and the functional status of activities of daily living compared with the exercise group alone; dual task cognitive motor training can improve the degree of walking improvement and the basic activity ability compared with the cognitive training group.

These findings help to clarify some of the inconsistencies reported in prior studies. Our subgroup analysis indicated that dual-task training may only yield measurable benefits when the intervention duration reaches at least 3 weeks, suggesting that training intensity and exposure time are critical for achieving neurological and functional gains. Moreover, although detailed categorization of cognitive task types was limited by the reporting quality of the included trials, the majority involved cognitive components such as attention, verbal fluency, calculation, or working memory-domains known to engage executive control and prefrontal circuits. These findings suggest that dual-task training is not universally effective, but its impact depends on factors such as training duration, control group type, and the relevance of the cognitive load to functional recovery targets. This may help explain previously conflicting results in the literature and supports a more condition-specific application of dual-task interventions in clinical practice.

Our subgroup analyses strongly suggest that training duration is a key determinant of dual-task training efficacy. We observed that interventions lasting 3 weeks or more were associated with significantly better outcomes in lower limb motor function (FMA-LE), balance (BBS), and activities of daily living (BI), whereas shorter interventions yielded limited improvements. This finding is consistent with previous meta-analyses and clinical trials. For instance, Zhou et al. (2021) reported that stroke patients required at least 3–4 weeks of cognitive-motor training to achieve meaningful improvements in functional mobility. Similarly, Ye et al. (2024) found that dual-task training of six weeks or more produced significantly greater cognitive and physical benefits in older adults with cognitive impairment. These results support the hypothesis that neuroplastic adaptations and functional gains from dual-task interventions require a minimum threshold of intensity and exposure time.

The results of this study are consistent with the results of Embrechts et al. [14]. Dual-task cognitive motor training may not be better than single motor training in some cases. In our study, when compared with single motor training in the Fugl-Meyer subgroup, dual-task cognitive motor training did not show better effects. Similarly, in the subgroup analysis of the TUG, there was no statistical difference between dual-task cognitive motor training and motor training alone. On the contrary, when we evaluated the BBS of dual task cognitive-motor training compared to exercise or cognitive training alone, dual task cognitive-motor training seemed to perform better. Not only that, dual task cognitive-motor training also scored better on the TUG than dual task cognitive-motor training compared to cognitive training alone. In addition, dual-task cognitive-motor training appears to be more desirable in the BI compared to exercise alone.

About 88% of stroke patients experience severe sequelae of lower limb motor dysfunction, with a long course of disease and poor recovery [28]. In this study, Fugl-Meyer was used to assess the lower limb motor function of the patients [24]. The intervention mode of exercise and cognitive training is more closely related to the content involved in daily life, such as maintaining standing balance or even walking freely for daily activities. However, the recovery of lower limb motor function is the result of some basic limb states of the body or temporary test results, which may need a longer duration of the rehabilitation program to show significant results in short tests, so this may be the reason why the results are not significant. Likewise, this also explains the significant results of dual-task training in the outcome of BBS or BI in stroke populations. The results in the TUG may be due to the fact that exercise participation involved is an advantage in assessing the desired outcome. Therefore, there was no statistically significant difference when dual-task training was compared to exercise alone, but there was a statistically significant difference when compared to cognition alone.

Cognitive symptoms caused by stroke include neglect, aphasia, apraxia, executive function, and memory, and cognitive impairment remains the leading cause of post-stroke morbidity [29]. The cognitive impairment of stroke is related to the loss of brain tissue and structure and the condition of neurons [30, 31], such as damage or loss of neurons [32]. Brain function is related to neuroplasticity, changes in neural structure and function in response to experiential or environmental stimuli. Cognitive training can also continue to promote changes through neuroplasticity, thereby improving cognitive function [33]. The reconnection of surviving neurons is a basic process of functional recovery [34]. Besides, the mechanisms associated with stroke-related injuries include excitotoxicity, oxidative stress, and inflammation [35–40]. Motor training can reduce glial activation and neuroinflammation in response to oxidative stress and improve related cognitive function, but this hypothesis still needs more research to prove [41].

Dual-task cognitive motor training is the simultaneous processing of motor (such as gait, gait primes, balance, or physical exercise) and cognitive (such as attention, decision-making, or working memory) activities [42, 43] This interaction between cognitive and motor abilities has been observed between some cognitive functions (such as attention and executive functions) and motor activities (such as gait balance and motor control) [43].

The beneficial effects of dual-task training on both functional and cognitive outcomes may be further explained by several theoretical models. The capacity-sharing model suggests that dual-task conditions train the brain to allocate limited cognitive resources more efficiently between motor and cognitive tasks. Over time, this leads to improved attentional control, working memory, and executive functioning-all of which are essential for post-stroke recovery. Another important framework is the bottleneck theory, which proposes that certain cognitive processes, such as decision-making and motor planning, compete for shared neural pathways. Dual-task training may help streamline these processes through repeated co-activation, improving both motor response efficiency and cognitive flexibility. Empirical findings from neuroimaging studies support these models, showing increased prefrontal cortex activation and enhanced functional connectivity between motor and cognitive control regions following dual-task interventions.

These mechanisms may be differentially engaged depending on the specific design of the dual-task training. Interventions combining gait training with serial subtraction likely target attentional control and working memory circuits, whereas balance exercises combined with auditory Stroop tasks may place greater demands on response inhibition and executive switching. Tailoring dual-task components to stimulate specific cognitive domains could therefore enhance training efficiency and facilitate more targeted neuroplastic adaptation.

It is also worth noting that only 12 of the included studies reported cognitive outcomes using standardized tools. This relatively limited sample size may affect the robustness of our conclusions regarding cognitive improvements. Moreover, the effectiveness of dual-task interventions may vary depending on the difficulty of the cognitive component and the patient’s baseline cognitive capacity. High task difficulty may lead to cognitive overload in some patients, while under-challenging tasks may limit training effects. Unfortunately, most studies did not stratify by baseline cognitive status or report detailed cognitive task characteristics, which limits further analysis. Furthermore, it is important to consider that stroke-related characteristics, such as time since stroke onset, lesion side, stroke type, and affected brain region, may also influence the effects of dual-task training. These factors can affect both motor recovery potential and cognitive capacity, which in turn modulate the responsiveness to dual-task interventions. However, most of the included studies did not provide stratified data on these clinical variables, limiting further exploration. Future studies should aim to include and report such characteristics to better personalize rehabilitation protocols.

Although the underlying mechanisms of dual-task cognitive motor training remain unclear, three main theories have been proposed: (1) competition for attentional resources, (2) competition for neural pathways involved in information processing, and (3) attentional control models [44]. Neuroimaging studies suggest that the prefrontal cortex plays a central role in dual-task processing. Some evidence also indicates that distinct neural modules may be involved, and that incorporating neuroimaging indicators into individualized intervention plans may help optimize clinical outcomes [45].

There are some limitations to this study. First, despite our comprehensive and systematic search, the number of included studies is still limited. Secondly, the included studies were mainly small sample studies. In addition, although we discussed the heterogeneity of the results of the subgroup analysis, the data of the subgroup was still limited, so the results should be interpreted with caution.

Conclusion

Nevertheless, our study results still suggest the potential of motor cognitive dual task exercise in stroke rehabilitation. Clinicians should consider patients’ specific physical functions and cognitive states when adopting the dual task program to adjust the cycle and corresponding intensity of the dual task, and design personalized programs based on the imaging related brain region features of patients and the rehabilitation goals that patients need to achieve, such as the emphasis on exercise or cognitive function. In the future, large-scale, multicenter randomized controlled trials are warranted to validate the generalizability of these findings across diverse clinical settings and patient populations.

Authors’ contributions

All authors contributed to the study’s conception. Yundi Jiang designed the study; Chuan Mou wrote the main manuscript text; Yundi Jiang, Chuan Mou: Literature survery; Yundi Jiang: review and comments, and all the authors approved the manuscript.

Funding

This research was supported by the Fundamental Research Funds for the Central Universities in Sichuan University.

Data availability

All data relevant to the study are included in the article or uploaded as online supplemental information. The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.