Effects of motor-cognitive training on cognitive function and gait performance in older adults with dementia: a systematic review and meta-analysis

Abstract

Motor-cognitive training combines motor and cognitive tasks during the training. So far, its effectiveness on cognitive function and gait in people with dementia remains unknown, and whether it is superior to single physical or cognitive training has yet to be investigated. Therefore, this meta-analysis aimed to explore the effects of motor-cognitive training on cognitive function and gait in people with dementia. Randomized controlled trials comparing motor-cognitive training with cognitive intervention alone, physical exercise alone, or other control group programs were included. Outcomes included cognitive functions and single/dual task gait performance. We conducted subgroup analysis based on the type of intervention applied in the control group. The pooled meta-analysis showed significant improvements following motor-cognitive training compared to control interventions in global cognition (SMD = 1.00, 95% CI 0.75, 1.26, p < 0.00001), single gait speed (SMD = 0.4, 95% CI 0.19, 0.61, p = 0.0002), and dual-task gait speed (SMD = 0.28, 95% CI 0.01, 0.55, p = 0.05). In the subgroup analysis, motor-cognitive training exerted significantly more improvement in global cognition and single gait speed when compared to either physical or cognitive training alone, or other control. Our results demonstrated the positive effects of motor-cognitive training on global cognition and gait speed in people with dementia. However, no significant improvements were observed in memory, attention, or executive function.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-09582-y.

Introduction

Dementia is a leading cause of rising dependency and disability among older population. Dementia is a general term that includes various conditions characterized by impaired cognitive functions, such as memory loss and decreased attention¹. These impairments are also linked to decreased physical performance, such as slower walking speed². Alarmingly, the global number of individuals with dementia is expected to rise significantly, rising from 57 million in 2019 to 153 million by 2050³. The anticipated rise in dementia prevalence is largely attributed to increased life expectancy, population aging, and the growing burden of chronic diseases associated with cognitive decline. As a result, effective intervention is needed to preserve both cognitive and physical function and help maintain functional independence in individuals with dementia.

Pharmacological interventions have been found to offer only mild effects to slow down cognitive deterioration in people with cognitive impairment⁴. Alternatively, non-pharmacological strategies are regarded as the primary strategy for supporting brain health in this population⁴. Among these approaches, physical and cognitive training are commonly recommended. A meta-analysis revealed that while cognitive training has shown to improve several cognitive domains for people with mild cognitive impairment (MCI), there is lacking evidence to confirm its effectiveness in individuals with dementia⁵. Another meta study reported that computerized cognitive training was only beneficial for improving verbal memory in individuals with dementia⁶. A recent meta-analysis found that physical intervention program had only a minor effect on enhancing cognition in individuals with dementia⁷. According to a meta-analysis by Borges-Machado et al., multicomponent physical training improved activities of daily living in older adults with dementia, however, the evidence for cognitive function enhancement was not robust⁸. Therefore, in addition to cognitive or physical training alone, more effective training strategies are needed for improving the cognitive function for people with dementia.

Motor-cognitive intervention combines motor and cognitive tasks simultaneously during training^9,10. It can be applied through two different approaches: (1) motor training combined with additional cognitive tasks, and (2) motor training integrated with cognitive tasks^9,11. The first approach resembles traditional dual-task training, where the cognitive task serves as a secondary element that may act as a distractor during motor training^12,13. On the other hand, when the cognitive task is integrated into the training (e.g., exergame or dance), it means that cognitive task is integral and necessary for successfully completing the motor task¹⁴.

According to the “guided plasticity facilitation” framework, it is suggested that motor-cognitive concept may specifically stimulate neuroplasticity, leading to greater and more additive benefits than cognitive or physical training alone^15,16. One previous review article indicated that motor-cognitive training had a greater impact on global cognition and inhibitory control compared to single physical or cognitive interventions for healthy older adults¹⁷. Moreover, one meta-analysis reported that motor-cognitive intervention significantly enhanced memory and executive function compared with single motor or cognitive interventions in MCI participants¹⁸. However, the effectiveness of motor-cognitive training on cognitive function and gait performance in older adults with dementia has yet to be established, and whether it is better than physical or cognitive training alone remains to be investigated.

Therefore, we aimed to investigate the evidence-based effects of motor-cognitive training on cognitive function and gait performance in older adults with dementia through systematic review and meta-analysis of RCTs. We also wanted to examine the varying effect of motor-cognitive intervention by comparing it to different controls, such as motor or cognitive training alone, through a subgroup analysis.

Methods

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines¹⁹. This systematic review and meta-analysis was registered in PROSPERO under the registration number CRD42024588024.

Search strategy

Literature searches were conducted in the following five electronic databases: PubMed, CINAHL, Cochrane, Medline, and Wiley. We performed initial search in September 2024 and updated in February 2025. The following keywords were used: (“dementia” OR “demented” OR “cognitive disorder” OR “cognitive impairment” OR “cognitive dysfunction” OR “Alzheimer” OR “Alzheimer’s” OR “Alzheimer Disease” OR “Lewy body” OR “Lewy bodies” OR “frontotemporal dementia” OR “fronto-temporal dementia” OR “vascular dementia”) AND (“motor-cognitive” OR “cognitive-motor” OR “dual task” OR “physical-cognitive” OR “cognitive-physical” OR “combined” OR “combining” OR “combination” OR “multi-model” OR “multi-component” OR “exergame” OR “dance” OR “virtual reality” OR “video game” OR “Wii” OR “Xbox” OR “computerized”) AND (“cognition” OR “cognitive function” OR “attention” OR “executive function” OR “memory” OR “learning” OR “language” OR “gait” OR “walking” OR “walk” OR “mobility”). Only studies published in English were selected. Furthermore, we reviewed the reference lists of relevant articles to identify any that were not captured by the initial search strategy.

Eligibility criteria

Studies were selected in this meta-analysis based on the following criteria: (1) participants’ age over 60 years old; (2) RCT study design; (3) participants with a clinical diagnosis of dementia, regardless of severity; (4) the experimental group received both motor and cognitive training concurrently (motor-cognitive training) as an intervention, such as dual-task exercises and exergames etc.; (5) the control group for comparison received either physical intervention alone, cognitive intervention alone, or other control programs (e.g., usual care or waitlist); (6) global cognition, attention, memory, executive function or single/dual task gait performance were measured; (7) Physiotherapy Evidence Database (PEDro) score ≥ 4²⁰. The exclusion criteria included: (1) participants included in the study were with diagnosis of other severe medical issues; (2) conference abstract, letters, case report, or systematic review.

Study selection

Two authors independently screened the studies identified through database searches based on their titles and abstracts. Following this initial screening, the full texts of potentially relevant articles were assessed for eligibility. Any disagreements were resolved through discussion, and if consensus could not be reached, a third author was consulted to determine the final decision.

Quality assessment

We assessed the methodological quality of each included study using the PEDro scale, which ranges from 0 to 10, with higher scores reflecting better quality. Two authors independently rated each study, resolving any discrepancies via discussion or consultation with a third author. We only included studies with a PEDro score of 4 or higher in the current study^20,21.

Risk of bias assessment and quality of evidence

The risk of bias in the RCTs was assessed using the Cochrane risk of bias tool, which covers six domains of bias: selection, performance, detection, attrition, reporting, and other biases. Each item was evaluated as having a high (red), low (green), or unclear (yellow) risk of bias²².

The quality of evidence for each outcome was assessed using the Grades of Recommendation, Assessment, Development, and Evaluation (GRADE) criteria. The level of evidence quality was downgraded based on five factors: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Conversely, the level of evidence could be upgraded if there was a large effect size, the presence of plausible confounding factors, and evidence of a dose–response relationship²³. The GRADE categorizes quality of evidence into 4 levels: high, moderate, low, and very low. We applied GRADEpro GDT software to generate the result tables²⁴.

Data extraction

We extracted four types of characteristics data, including the study characteristics (authors, study design, country, publication year, and sample size), participants characteristics (age, diagnosis, and diagnostic criteria used to define dementia), intervention program (motor-cognitive program, session length, session frequency, intervention duration, adherence rate, setting, and control group intervention), and interested outcomes.

Data analysis

We conducted this meta-analysis using Review Manager 5.4. To estimate the intervention effect, we calculated the pooled standardized mean difference (SMD) with a 95% confidence interval (CI). The SMD provides a point estimate of the intervention effect and is classified as small (0.2), moderate (0.5), or large (0.8)²⁵. For studies with multiple control groups, the participants in the experimental group were evenly distributed for the comparisons, following the recommendations from Cochrane. Heterogeneity among the selected studies was assessed using the I² statistic, with thresholds defined as small (25%), moderate (50%), or large (75%) based on the I² value. If I² > 50%, a random-effects model was used instead of a fixed-effects model, and a sensitivity analysis was conducted. Due to the limited number of studies for each outcome, funnel plot asymmetry was not assessed for publication bias. Because the included studies used different types of control groups, we conducted subgroup analysis according to the control group intervention. Active interventions, including aerobic exercise or cognitive stimulation therapy, were classified as the cognitive or physical training alone group, while inactive interventions, such as usual care, were classified as other control group. A p-value < 0.05 was considered statistically significant.

Results

Search results

The literature search identified 9,002 potentially relevant articles. After removing the duplicates, 4430 articles remained for screening. Then, 4309 articles were excluded after screening titles and abstracts, leaving 121 studies for full-text reviewing. After full-text screening, 106 of the 121 articles were excluded for some reasons: young participants (n = 1), without raw data (n = 3), without full text (n = 5), non-RCT (n = 5), interventions did not match (n = 26), participants did not match (n = 35), conference abstract (n = 6), control group did not match (n = 6), without interested outcomes (n = 9), and study protocol (n = 10). Finally, 15 articles met the criteria and were selected in present review. Study flow chart was presented as Fig. 1.

Fig. 1.

Flow chart of study inclusion in the systematic review and meta-analysis.

Quality assessment and risk of bias

Table 1 illustrates the methodological quality of the included studies. All of the included studies met the three items: random allocation, between group difference reported, and point estimate and variability reported. Most studies achieved scores for concealed allocation and baseline group similarity. Some studies achieved scores for assessor blinding, < 15% dropouts, and intention-to-treat analysis. However, none of the studies reached participant blinding or therapist blinding. The average PEDro scores of all the included articles was 6.3.

Table 1.

Methodological quality indicated by PEDro criteria of the included studies.

Article	Random allocation	Concealed allocation	Groups similar at baseline	Participant blinding	Therapist blinding	Assessor blinding	< 15% dropouts	Intention-to- treat analysis	Between group difference reported	Point estimate and variability reported	Total score
Ho et al., 2020	Y	Y	Y	N	N	Y	Y	Y	Y	Y	8
Bracco et al., 2023	Y	N	Y	N	N	N	Y	N	Y	Y	5
Wu et al., 2023	Y	Y	Y	N	N	Y	N	N	Y	Y	6
Karssemeijer et al., 2019a	Y	Y	Y	N	N	Y	Y	Y	Y	Y	8
Karssemeijer et al., 2019b	Y	Y	Y	N	N	Y	Y	Y	Y	Y	8
Okamura et al., 2018	Y	Y	Y	N	N	Y	Y	Y	Y	Y	8
Menengi et al., 2022	Y	Y	Y	N	N	N	Y	Y	Y	Y	7
Lemke et al., 2019	Y	Y	Y	N	N	Y	N	Y	Y	Y	7
Padala et al., 2017	Y	Y	Y	N	N	N	N	Y	Y	Y	6
Yoon et al., 2013	Y	Y	Y	N	N	N	N	N	Y	Y	5
Swinnen et al., 2023	Y	Y	Y	N	N	N	N	N	Y	Y	5
Zheng et al., 2022	Y	Y	Y	N	N	Y	N	N	Y	Y	6
Trautwein et al., 2020	Y	N	Y	N	N	N	N	Y	Y	Y	5
Swinnen et al., 2021	Y	Y	Y	N	N	Y	N	N	Y	Y	6
Binns et al., 2020	Y	N	N	N	N	Y	Y	N	Y	Y	5

PEDro, physiotherapy evidence database; Y, yes, meets the criteria; N, no, does not meet the criteria.

Figure 2 summarizes the risk of bias assessment results. All included articles had a low risk of bias in the random sequence generation (selection bias). Most studies also had a low risk of bias in allocation concealment (selection bias), blinding of outcome assessment (detection bias), and selective reporting (reporting bias). On the other hand, some studies had unclear or high risk of incomplete outcome data (attrition bias) due to high percentages of participants lost to follow-up or treatment withdrawals. Given the nature of the intervention, participants and personnel blinding were not feasible, leading to a high risk of bias across all studies.

Fig. 2.

Cochrane Risk-of-Bias assessment of the included studies. + : low risk of bias; − : high risk of bias; ?: unclear risk of bias.

Study characteristics

Table 2 shows the included studies characteristics. 910 participants (control: 498; experimental: 412) with a mean age ranged from 70.1 to 87.5 years old were included. Fifteen studies were conducted in Hong Kong (n = 1)²⁶, France (n = 1)²⁷, South Korea (n = 2)^28,29, Netherlands (n = 2)^30,31, Japan (n = 1)³², Turkey (n = 1)³³, Germany (n = 2)^34,35, United States of America (n = 1)³⁶, Belgium (n = 2)^37,38, China (n = 1)³⁹, and New Zealand (n = 1)⁴⁰. All participants were diagnosed with dementia, Alzheimer’s disease, or major neurocognitive disorders, a term introduced in the Diagnostic and Statistical Manual of Mental Disorders (DSM V) criteria to replace dementia⁴¹. Regarding the diagnostic criteria, most of the included studies used the DSM IV or DSM V criteria^{26,30–32,36–38}, one study used the ICD-10 criteria³⁵, and one study used the National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer’s Disease (NINCDS-ADRDA) criteria³³. Some of the included studies did not specify the diagnostic criteria, but they noted that participants had been diagnosed with dementia by physicians or qualified specialists^28,34,39,40. Two studies identified their participants as residents of facilities for older adults with dementia^27,29, with one study specifically reporting that participants had mild to moderate dementia²⁷.

Table 2.

Characteristics of the included studies.

Study	Country or area	Sample			Motor-cognitive intervention			Control intervention			Interested Outcomes
		Participant	Age	Diagnosis	Description	Dose	Setting	Adherence Levels	Description	Dose
Ho et al., 2020	Hong Kong	EG (n = 69)	EG: 79.4 ± 7.6 Y/O	Mild dementia (DSM IV)	Dance-movement program including simple group dance, movement games, improvisational dance movement, and movement interactions among members	1 h/session, 2 sessions/wk for 12 wks	Hospital & community; Group-based	Not mentioned	Stretching and towel exercise/ Waitlist control	1 h/session, 2 sessions/wk for 12 wks	DST
		Exercise (n = 67)	Exercise: 79.3 ± 8.1 Y/O								TMT
		Waitlist control (n = 68)	Waitlist control: 78.3 ± 8.4 Y/O								VFT
Bracco et al., 2023	France	EG (n = 13)	EG: 81 ± 8 Y/O	Dementia	Therapeutic tango program including warm-up, dance standing or sitting, cool down, and feedback	1 h/session, 2 sessions/wk for 12 wks	Nursing home; Group-based	90%	Strength, balance, and stretching exercises	1 h/session, 2 sessions/wk for 12 wks	Single gait
		Exercise (n = 13)	Exercise: 85 ± 5 Y/O
Wu et al., 2023	South Korea	EG (n = 13)	EG: 78.8 ± 4.8 Y/O	Mild or moderate dementia	Exergame. During the game, players run with the avatar, avoiding obstacles, and win items while running or jumping at speed on the mat	30–50 min/session, 3 sessions/wk for 12 wks	Daycare centers; Individualized	Not mentioned	Cycling exercise	30–50 min/session, 3 sessions/wk for 12 wks	Eriksen flanker test
		Exercise (n = 11)	Exercise: 81.2 ± 4.4 Y/O
Karssemeijer et al., 2019 a	Netherlands	EG (n = 38)	EG: 79 ± 6.9 Y/O	Dementia (DSM IV)	Exergame. Cognitive–bicycle training, Participants followed a route through a digital environment and simultaneously performed cognitive tasks	30–50 min/session, 3 sessions/wk for 12 wks	Community; Individualized	87.3%	Aerobic cycling exercise/ Relaxation and flexibility exercises	30–50 min/session, 3 sessions/wk for 12 wks	TMT
		Aerobic exercise (n = 38)	Aerobic exercise: 80.9 ± 6.1 Y/O								Stroop test
Karssemeijer et al., 2019 b		Flexibility exercise (n = 39)	Flexibility exercise: 79.8 ± 6.5 Y/O								VFT
											DST
											Single gait
Okamura et al., 2018	Japan	EG (n = 50)	EG: 82.4 ± 8 Y/O	Dementia (DSM V)	Combined exercise and cognitive training. Patient have to drive an arm ergometer to achieve the target speed shown by the moving mark on the display	5 min/session, 1 session or more/wk for 6 wks	Day care or residential care facilities; Individualized	Not mentioned	Arm ergometer exercise	5 min/session, 1 session or more/wk for 6 wks	MMSE
		Exercise (n = 50)	Exercise: 79.2 ± 9.8 Y/O								TMTA
Menengi et al., 2022	Turkey	EG (n = 10)	EG: 77.7 ± 5.29 Y/O	Alzheimer’s disease (NINCDS-ADRDA criteria)	Dual-task exercise. Simple chair-based exercises with cognitive tasks	15 min/session, 5 days/wk, and gradually progressed to 40 min/session, 4 days/wk. totally 25 sessions over 6 wks	Home (online); Individualized	High adherence	Usual care	N/A	MMSE
		Usual care (n = 10)	Usual care: 80.6 ± 6.11 Y/O
Lemke et al., 2019	Germany	EG (n = 44)	EG: 82.7 ± 6.2 Y/O	Dementia	Dementia-specific motor-cognitive training. 15–20 min of dual task training (walking and counting) + game-based motor-cognitive training + motor learning exercise program	1.5 h twice a wk for 10 wks	Geriatric Center; Group-based	Not mentioned	Strength and flexibility exercises	1 h twice a wk for 10 wks	Dual task gait
		Exercise (n = 43)	Exercise: 82.6 ± 5.8 Y/O
Padala et al., 2017	United States of America	EG (n = 15)	EG: 72.1 ± 5.3 Y/O	Mild Alzheimer’s disease (DSM IV)	Wii-Fit. Yoga, strength training, aerobics, balance games, and training plus, which includes more complex tasks	30 min/session, 5 sessions/wk for 8 wks	Home; Individualized	95%	Walking program	30 min/session, 5 sessions/wk for 8 wks	MMSE
		Exercise (n = 15)	Exercise: 73.9 ± 7.1 Y/O
Yoon et al., 2013	South Korea	EG (n = 11)	EG: 77.9 ± 7.5 Y/O	Dementia	Cycling exercise during cognitive activity. Patients rhythmicallyrepeated simple actions to build up reactionary movement. Cognitive activity included sequential memory recall tasks	30 min/session, 3 sessions/wk for 12 wks	Long-term care facility; Individualized	Not mentioned	Cognitive training	30 min/session, 3 sessions/wk for 12 wks	DST
		Cognitive training (n = 9)	Cognitive training: 70.1 ± 12.2 Y/O
Swinnen et al., 2023	Belgium	EG (n = 7)	EG: 81.9 ± 8.2 Y/O	Major neurocogn-itive disorder (DSM V)	Exergame. Strength exercises, Tai-Chi movements, and balance training, combining with cognitively tasks	30 min/session, 3 sessions/wk for 12 wks	Long-term care facility; Individualized	61%	Walking, squatting and stepping exercises	30 min/session, 3 sessions/wk for 12 wks	MMSE
		Exercise (n = 11)	Exercise: 84.2 ± 5.9 Y/O
Zheng et al., 2022	China	EG (n = 18)	EG: 81.7 ± 5.8 Y/O	Dementia	Active game (box Kinect system). Person’s hand–eye coordination and motor skills were trained	1 h/session, 5 sessions/wk for 8 wks	Senior day care center and a nursing home; Group-based	100%	Usual care	NA	MMSE
		Usual care (n = 20)	Usual care: 84.3 ± 5.5 Y/O
Trautwein et al., 2020	Germany	EG (n = 90)	EG: 85 ± 7 Y/O	Dementia (ICD-10 criteria)	Multimodal exercise program. Participants underwent a program combining motor (i.e. strength, balance, endurance, and flexibility) and cognitive tasks (i.e. memory, attention, language, and executive function)	1 h/session, twice a wk for 16 wks	Care facilities; Group-based	62%	Usual care	NA	Single gait
		Usual care (n = 73)	Usual care: 86 ± 5 Y/O								Dual task gait
Swinnen et al., 2021	Belgium	EG (n = 23)	EG: 84.7 ± 5.6 Y/O	Major neurocogn-itive disorder (DSM V)	Exergame. Participants interacted with the game interface by pushing one foot on one of the four arrows. Attention, flexibility, postural control, and visuospatial memory were trained	15 min/session, 3 sessions/wk for 8 wks	Long-term care facility; Individualized	82.9%	Watching and listening to music videos	15 min/session, 3 sessions/wk for 8 wks	MoCA
		Control (n = 22)	Control: 85.3 ± 6.5 Y/O								Single gait
Binns et al., 2020	New Zealand	EG (n = 11)	EG: 83.6 Y/O	Mild to moderate dementia	Combing fall prevention exercises with CST. The program was CST with aerobic, progressive strength and balance exercises embedded	1 h twice a wk for 7 wks	Residential aged care; Group-based	55%	CST	1 h twice a wk for 7 wks	MoCA
		Cognitive training (n = 9)	Cognitive training: 87.5 Y/O								Single gait

EG, experimental group; DSM, Diagnostic and Statistical Manual of Mental Disorders; Y/O, years old; wk, week; DST, digit span test; TMT, trail making test; VFT, verbal fluency test, min, minute; MMSE, Mini-Mental State Exam; NINCDS-ADRDA, National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer’s Disease; N/A, not applicable; MoCA, Montreal Cognitive Assessment; CST, cognitive stimulation therapy.

The content of the motor-cognitive training varied across the included studies. Two studies conducted dance-based program^26,27, 7 studies identified their training programs as exergame-based motor-cognitive training^{28,30,31,36–39}, and 1 study designed a dementia-specific training program, including traditional dual-task gait training and gamed-based motor-cognitive training³⁴. In addition, 1 study conducted dual-task cycling training²⁹, and the others conducted dual-task exercise intervention (e.g., combining cognitive tasks and strength exercises)^32,33,35,40. As for the control group, programs in cognitive or physical alone group included multicomponent exercises, aerobic exercise, mobility and walking exercises, memory training, and cognitive stimulation therapy, while programs in other control group included usual care, a waitlist, and music listening. The motor-cognitive intervention durations ranged from 5 to 90 min per day, with a frequency of 2–5 days per week for 6–16 weeks except 1 study did not provide the exact frequency of its intervention³². The adherence rate, measured as the proportion of training sessions completed, ranged from 55 to 100% across the included studies. Regarding the settings in which the interventions were conducted, 2 took place at home^33,36, 3 in community settings^26,30,31, and the others in facilities such as nursing homes and daycare centers^{27–29,32,34,35,37–40}. Six studies emphasized that their interventions were group-based training^{26,27,34,35,39,40}, while the others implemented individualized training programs^{28–33,36–38}. A dance-movement therapist executed the training program in one study²⁶. Staff at facilities performed the training programs in 4 studies^27,32,38,40. Trained students or researchers supervised the programs in 3 studies^30,31,39. One study was conducted online with real-time supervision by a physical therapist³³. Trained movement scientists executed the training program in one study³⁴. Caregiver supervised the training program in 1 study³⁶. Therapists performed the training program in 1 study³⁷, trained instructors executed the training program in the other study³⁵, while this information was not mentioned in two studies^28,29.

Cognitive function

Global cognition

Seven studies, involving a total of 271 participants examined the effects of motor-cognitive training on global cognition in people with dementia. Among these studies, 5 studies applied MMSE^{32,33,36,38,39} and 2 studies used Montreal Cognitive Assessment (MoCA)^37,40 to assess global cognition. The pooled data showed a beneficial effect of motor-cognitive training on improving global cognition (SMD = 1.00, 95% CI 0.75, 1.26, p < 0.00001) (Fig. 3). As for the subgroup analysis, motor-cognitive intervention exerted significant improvement either comparing with the physical or cognitive training alone (SMD = 1.13, 95% CI 0.80, 1.46, p < 0.00001)^32,36,38,40 or other types of control group (SMD = 0.81, 95% CI 0.40, 1.22, p = 0.0001)^33,37,39. The certainty of evidence according to GRADE evaluation was moderate (Table 3).

Fig. 3.

Forest plot. Meta-analysis of the effects of motor-cognitive training compared to control on global cognition in individuals with dementia. SD, standard deviation; Std., standard; CI, confidence interval.

Table 3.

Quality of evidence based on the GRADE system.

No. of studies	Certainty assessment						No. of patients		Effect	Certainty
	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Intervention	Control	Absolute
									(95% CI)
Global cognition
7	Randomised trials	Seriousa	Not serious	Not serious	Seriousb	Strong association	134	137	SMD 1.00 SD higher	⨁⨁⨁◯
									(0.75 higher to 1.26 higher)	Moderate
Memory
2	Randomised trials	Seriousa	Not serious	Not serious	Seriousb	None	107	212	SMD 0.02 SD higher	⨁⨁◯◯
									(0.21 lower to 0.25 higher)	Low
Attention
3	Randomised trials	Seriousa	Not serious	Not serious	Not serious	None	157	262	SMD 0.05 SD lower	⨁⨁⨁◯
									(0.41 lower to 0.3 higher)	Moderate
Executive function-working memory
3	Randomised trials	Seriousa	Not serious	Not serious	Seriousb	None	118	221	SMD 0.14 SD lower	⨁⨁◯◯
									(0.36 lower to 0.09 higher)	Low
Executive function-mental flexibility
2	Randomised trials	Seriousa	Not serious	Not serious	Seriousb	None	107	212	SMD 0.06 SD higher	⨁⨁◯◯
									(0.17 lower to 0.3 higher)	Low
Executive function-inhibitory control
2	Randomised trials	Seriousa	Not serious	Seriousc	Seriousb	None	51	88	SMD 0.01 SD lower	⨁◯◯◯
									(0.36 lower to 0.34 higher)	Very low
Single gait speed
5	Randomised trials	Seriousa	Not serious	Not serious	Seriousc	None	174	194	SMD 0.4 SD higher	⨁⨁◯◯
									(0.19 higher to 0.61 higher)	Low
Dual task gait speed
2	Randomised trials	Seriousa	Not serious	Not serious	Seriousc	None	105	102	SMD 0.28 SD higher	⨁⨁◯◯
									(0.01 higher to 0.55 higher)	Low

CI, confidence interval; SMD, standardised mean difference.

^aIncluded studies lack of blinding of participants and personnel; ^bSmall sample size; ^cDifferent outcomes measures.

Memory

Two studies, involving a total of 319 participants examined the effects of motor-cognitive intervention on memory in people with dementia. Both studies used verbal fluency test (VFT) to evaluate memory function^26,30. The pooled data showed no significant difference between motor-cognitive intervention and the control in improving the memory function (SMD = 0.02, 95% CI -0.21, 0.25, p = 0.87) (Supplementary Fig. 1). No significant results were found in the subgroup analysis. The certainty of evidence based on GRADE evaluation was low (Table 3).

Attention

Three studies, involving a total of 419 participants investigated the effects of motor-cognitive intervention on attention in people with dementia. Trail Making Test A (TMT-A) was used to assess attention and processing speed^26,30,32. The pooled data showed no significant difference between motor-cognitive intervention and the control in improving attention (SMD = -0.05, 95% CI -0.41, 0.30, p = 0.76) (Supplementary Fig. 2). No significant results were found in the subgroup analysis. The sensitivity analysis was conducted because of the high heterogeneity observed among the included studies (I² = 67%), and the results remained the same. (Supplementary Table 1). The certainty of evidence according to GRADE evaluation was moderate (Table 3).

Executive function—working memory

Three studies, involving a total of 339 participants investigated the effects of motor-cognitive intervention on working memory in people with dementia. The Digit Span Test (DST) was used in these studies to indicate working memory function^26,29,30. The pooled data showed no significant difference between motor-cognitive intervention and the control in improving working memory (SMD = -0.14, 95% CI -0.36, 0.09, p = 0.23) (Supplementary Fig. 3a). No significant results were found in subgroup analysis. The certainty of evidence based on GRADE evaluation was low (Table 3).

Executive function-mental flexibility

Two studies, involving a total of 319 participants investigated the effects of motor-cognitive intervention on mental flexibility in people with dementia. The Trail Making Test B (TMT-B) was used in these studies to represent mental flexibility function^26,30. The pooled data showed no significant difference between motor-cognitive intervention and the control in improving mental flexibility (SMD = 0.06, 95% CI -0.17, 0.30, p = 0.59) (Supplementary Fig. 3b). No significant results were found in subgroup analysis. The certainty of evidence based on GRADE evaluation was low (Table 3).

Executive function-inhibitory control

Two studies, involving a total of 139 participants investigated the effects of motor-cognitive intervention on inhibitory control in people with dementia. One study used the Stroop Color Word Test (SCWT)³⁰, while the other used the Eriksen Flanker Test²⁸. We chose the response time of the incongruent task to represent inhibitory control. The pooled data showed no significant difference between motor-cognitive intervention and the control in improving inhibitory control (SMD = -0.01, 95% CI -0.36, 0.34, p = 0.96) (Supplementary Fig. 3c). The certainty of evidence according to GRADE evaluation was very low (Table 3).

Single gait speed

Five studies, involving a total of 368 participants investigated the effects of motor-cognitive intervention on single gait speed in people with dementia. Among these studies, three used the Four-Meter Walk Test^27,37,40, one used the Ten-Meter Walk Test³¹, and the other used the electronic gait analysis system to evaluate single gait speed³⁵. The pooled data showed a more positive effect of motor-cognitive intervention than the control to improve single gait speed (SMD = 0.4, 95% CI 0.19, 0.61, p = 0.0002) (Fig. 4). As for the subgroup analysis, motor-cognitive intervention exerted significantly more improvement either comparing with the physical or cognitive training alone (SMD = 0.38, 95% CI 0.05, 0.71, p = 0.02)^27,30,40 or other types of control group (SMD = 0.41, 95% CI 0.13, 0.69, p = 0.004)^35,37. The certainty of evidence based on GRADE evaluation was low (Table 3).

Fig. 4.

Forest plot. Meta-analysis of the effects of motor-cognitive training compared to control on single gait speed in individuals with dementia. SD, standard deviation; Std., standard; CI, confidence interval.

Dual task gait speed

Two studies, involving a total of 207 participants examined the effects of motor-cognitive intervention on dual task gait speed in people with dementia. The performance of walking while naming animals was used for the present meta-analysis because this dual-task condition was evaluated in both studies.^34,35. Both studies asked their participants to walk on the electronic gait analysis system to evaluate gait speed. The pooled data showed a more positive effect of motor-cognitive intervention than the control to improve dual task gait speed (SMD = 0.28, 95% CI 0.01, 0.55, p = 0.05) (Fig. 5). The results of the subgroup analysis showed a significantly more effect of motor-cognitive training when compared with physical or cognitive training alone (SMD = 0.46, 95% CI = 0.03, 0.88, p = 0.03)³⁴. However, no significant effect was observed when comparing motor-cognitive training with other control groups (SMD = 0.15, 95% CI = -0.21, 0.51, p = 0.4)³⁵. The certainty of evidence according to GRADE evaluation was low (Table 3).

Fig. 5.

Forest plot. Meta-analysis of the effects of motor-cognitive training compared to control on dual gait speed in individuals with dementia. SD, standard deviation; Std., standard; CI, confidence interval.

Discussion

To our knowledge, this is the first systematic review and meta-analysis of RCTs to explore the effects of simultaneous motor-cognitive training on cognitive function and gait performance in older adults with dementia. Our results showed significantly more improvements in global cognition, single gait speed, and dual task gait speed in the intervention group compared to the control group among dementia population, with effect sizes ranging from small to large. However, no significant improvements were observed in memory, attention, or executive function.

Cognitive decline is a significant concern for individuals with dementia, as it seriously impacts their quality of life⁴². The findings of the present study demonstrated motor-cognitive training effectively improved global cognitive function in individuals with dementia. This aligns with recent systematic reviews that also reported significant benefits of motor-cognitive interventions for enhancing global cognition in populations with MCI and dementia^43,44. However, one of these reviews combined participants with MCI and dementia in the meta-analysis⁴³, limiting the ability to discern dementia-specific effects. The other review included only four studies in the dementia subgroup analysis for global cognition, and just one study each for attention and executive function⁴⁴. Therefore, our study contributes more robust and dementia-specific evidence by incorporating a larger number of studies focused on individuals with dementia.

Moreover, our result further demonstrated that motor-cognitive intervention was more effective than single cognitive or physical training in enhancing global cognition. Previous study also reported beneficial effects of motor-cognitive training on enhancing global cognition in older population compared to cognitive or physical training alone¹⁷. These enhanced effects may be explained by neural plasticity. According to previous studies, physical training facilitated neurophysiological mechanisms, such as increasing the expression of brain-derived neurotrophic factor (BDNF) to support the plastic changes^16,45,46. However, the increase in BDNF levels after exercise is sometimes to be transient and time-constrained. On the other hand, cognitive intervention was believed to promote the survival of the newly generated synapses or neurons, guide the neuroplastic processes, and further maintain the positive plastic changes⁴⁷. Thus, combining these two training approaches exerted better effects than either physical or cognitive training alone for enhancing global cognition in demented individuals.

Seven studies were included in current meta-analysis of global cognition. Most of the included studies applied individualized training program^{32,33,36–38}. Previous studies suggested that the individualized training format, designed based on individuals’ abilities, needs, priorities, and learning styles, can specifically benefit cognitive function in older adults^48,49. We speculate that individualized motor-cognitive program may be a better option for the demented population to improve global cognition, as it offers a higher level of adaptability and flexibility compared to group-based formats, making it more suitable for individuals with cognitive impairments. With regard to training type, six of seven included studies used motor training integrated with cognitive tasks^{12,32,36–38,40}. Integrating cognitive tasks into motor tasks appears to be more effective at stabilizing neuroplasticity than using cognitive tasks merely as distractors. This approach engages multiple sensory systems and more closely mimics real-life scenarios^9,50. However, a study that implemented individualized exergame training did not find significant improvements in global cognition, possibly due to a low adherence rate (61%)³⁸. Taken together, we suggest that individualized motor training integrated with cognitive tasks should be prioritized as a motor-cognitive approach for people with dementia. During the training period, adherence levels should be closely monitored to ensure effectiveness⁵¹. It should also be noted that dementia severity may be a critical factor influencing the effectiveness of motor-cognitive training. Individuals with more advanced cognitive impairment may have greater difficulty fully engaging in training programs, which could contribute to reduced intervention efficacy.

In terms of domain-specific effects, our results did not show significantly more improvements in memory, attention, and executive function in the intervention group compared to control group in people with dementia. As for results of subgroup analysis, the motor-cognitive training did not exert better effects on such domain-specific cognitive function as compared with other control or physical/cognitive training alone. The lack of significant effects may be attributed to several factors, including the training complexity, the severity of cognitive impairment among participants, and the relatively short training duration. One possible explanation for the insignificant result might be the complexity of the training. In Ho et al.’s study, their dance training program emphasized both individual movements and movement interactions among group members²⁶. This increased the challenge and difficulty, potentially causing excessive stress for individuals with dementia, which may have affected the cognitive benefits. In addition, the comparisons with other control were based on single study for each specific domain. Therefore, further investigation into the effects of motor-cognitive training on specific cognitive functions is needed.

The insignificant difference between motor-cognitive training and cognitive/ physical alone was also noted in a previous meta-analysis which assessed the effects of combining cognitive and physical training on cognitive functions in older adults with MCI or dementia⁵². On the other hand, another meta-analysis indicated that combined cognitive and motor training group significantly improved memory, attention, or executive function in individuals with MCI compared to cognitive or physical training alone¹⁸. Thus, this underscores the importance of early intervention for individuals with cognitive impairment. Previous studies highlighted that the earlier an intervention is introduced to individuals with cognitive impairment, the more likely it is to effectively improve their daily function and cognitive abilities^53,54. Furthermore, the relatively short training duration of the included studies in our study may affect the training effects on cognitive subdomains. The training duration of the studies included in current review ranged from 6 to 12 weeks. In contrast, a meta-analysis found that dance movement interventions had a beneficial effect on cognitive functions in individuals with cognitive impairment, particularly in the subgroup with a training period exceeding 12 weeks⁴³. Consequently, the length of training duration could be a crucial factor influencing the effects of motor-cognitive intervention.

Gait speed is a reliable and sensitive measure of physical function in older adults, and it serves as a predictor of hospitalization, functional decline, and mortality^55–57. Furthermore, higher gait speed is linked to improved community mobility and a better quality of life. A previous review reported positive effects of motor-cognitive training on improving gait speed in older adults with MCI⁵⁸. Our results further reported significant benefits on gait speed in older adults with dementia after motor-cognitive training. A previous study indicated that gait speed is strongly associated with global cognition in individuals with cognitive impairment⁵⁹. The improvement in global cognition after motor-cognitive training may lead to an increase in gait speed in people with dementia. According to the subgroup analysis, motor-cognitive training significantly improved single gait speed compared to other control group. Among included studies, Swinnen et al.’s study demonstrated a large effect size (SMD = 1.15)³⁷ by applying the stepping exergaming program (15 min/session, 3 sessions/week for 8 weeks). During the game, participants were asked to step in different directions according to the prompts provided by the game interface to possibly stimulate cognitive function and postural control. Meanwhile, the game device provided visual, auditory and somatosensory feedback, creating conditions that was close to real world walking. Therefore, such motor-cognitive intervention could be considered as an effective training program to increase gait speed in demented population. We also noted significantly more improvement in gait speed following motor-cognitive training compared to single physical or cognitive training. This suggests that motor-cognitive training could be a prioritized strategy to enhance single gait speed in people with dementia. However, it should be noticed that studies included in this subgroup analysis did not report positive effects. More rigorous studies are needed to draw definitive conclusions. Moreover, future studies incorporating biomechanical analyses, such as joint angle trajectories and coordination patterns, may provide a more comprehensive understanding of the underlying mechanisms by which motor-cognitive training influences gait performance.

Regarding the dual task gait speed, we found a significant improvement in older adults with dementia after motor-cognitive training. Among the two included studies, Lemke et al.’s study applied dementia-specific motor-cognitive training (1.5 h/session, 2 sessions/week for 10 weeks) and found significant improvements compared to physical training³⁴. Their program focused on dual task walking training under low- and high-demand dual task conditions, such as 2-forward and 3-backward calculations³⁴. Based on the task specific concept, a key principle in motor learning, frequent repetitions of specific task exercises may help to improve the specific task performances^60,61. Therefore, we suggested that the above motor-cognitive training programs may be adopted to improve dual task walking speed in older adults with dementia. Trautwein et al. conducted dual-task exercise training (e.g., combining strength exercises and cognitive tasks), and found no significant improvement in dual-task gait speed compared to usual care³⁵. The low adherence rate (62%) may have contributed to the insignificant result, as previous studies indicated that adherence is an important factor for the success of training in older population^51,62.

Regarding the GRADE score, the certainty of evidence was rated as moderate for global cognition and attention, while the other outcomes were classified as low to very low quality evidence. The most common limitations among the included articles were lack of blinding, small sample size, and different outcomes measures.

This study has several limitations that should be acknowledged. First, variations in the diagnostic criteria and severity of cognitive impairment across the included studies make it challenging to interpret the findings with precision. Second, the number of included studies was relatively small. Therefore, more studies are encouraged to explore the effect of motor-cognitive training on cognitive function and gait function in people with dementia. Third, we acknowledge that publication or language bias may exist, particularly as our search was limited to English publications. Fourth, while variability in motor-cognitive training protocols, intervention durations, and outcome measures may have contributed to differences in study findings, the heterogeneity in our analyses remained predominantly low to moderate, suggesting a consistent effect across the included studies. Finally, the optimal dosage and intensity for motor-cognitive intervention remains unclear, as the training protocols differed across studies and few provided details on intensity adjustments.

Conclusion

The finding of this systematic review and meta-analysis demonstrated that motor-cognitive training can be considered a valid and preferred strategy for improving global cognition in demented population. However, no significant effects on specific cognitive domains were found after motor-cognitive training compared to the control group. Thus, further evidence is needed before motor-cognitive training can be recommended for improving specific cognitive domains in individuals with dementia. Our results also supported the positive effects of motor-cognitive training on enhancing gait speed compared to control group. Given the limited number of included studies and the low quality of evidence for most outcomes, more well-designed trials are needed to validate the findings of this meta-analysis.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

WWH wrote the main manuscript, participated in the literature screening, collected the data, and analyzed the data. YNC participated in the literature screening, collected the data and analyzed the data. YYR provided insights of the manuscript writing; WRY provided insights of the manuscript writing and participated in the literature screening. All authors reviewed the manuscript.

Funding

This study was supported by the National Science and Technology Council under Grant (NSTC 112-2314-B-A49-034-MY2) and (NSTC 113-2314-B-A49-061-).

Data availability

The datasets used or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Competing interest

The authors declare no competing interests.