Effects of task-oriented training on balance, gait, and fall in older adults: A systematic review and meta-analysis of randomized controlled trials

Abstract

Background

Task-oriented training(TOT) is a rehabilitation approach focused on practicing goal-oriented, functional tasks.

Aims

To investigate the effects of TOT compared with conventional training on improving balance, gait, and fall prevention in older adults(aged ≥ 60 years).

Methods

The study was registered in PROSPERO(ID: CRD42025635269). We searched PubMed, Web of Science, Embase, and ClinicalTrials.gov up to October 2025 for eligible randomized controlled trials. Methodological quality was assessed using the PEDro scale and the Cochrane risk of bias tool. Statistical analyses were performed using Stata 18 and RevMan 5.4. Minimal clinically important difference(MCID) values were determined by reviewing relevant literature.

Results

Twenty studies were included, fifteen of which with complete data were meta-analyzed. Compared to conventional training, TOT was more effective in improving Berg Balance Scale (BBS)(MD = 2.58, P < 0.001), Timed Up and Go test (TUG test)(MD = -0.55, P < 0.001), gait speed (GS)(MD = 0.14, P < 0.001), and Activities-specific Balance Confidence Scale(MD = 17.06, P < 0.001). Improvements in BBS, TUG test and GS for individuals aged ≥ 75 reached the MCID. Although a significant main effect was observed for the Falls Efficacy Scale-International(MD = -2.80, P = 0.03), the result was not robust. The evidence regarding the effect of TOT on fall incidence or Functional Gait Assessment scores was inconclusive.

Conclusion

TOT showed superior effects over conventional training in improving balance, gait, and balance confidence, with clinically meaningful improvements in balance and gait speed. However, its impact on reducing fear of falling and fall incidence requires further validation.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40520-025-03303-1.

Introduction

With the accelerating global population aging, the prevalence of falls among older adults has risen markedly, with recent studies reporting a global prevalence rate of 26.5% [1]. Falls can not only cause injuries, hospitalizations, and even death, but may also lead to long-term care needs, reduced quality of life, and increased healthcare costs [2–5]. Given the continuing demographic shift toward an older population, proactive fall prevention strategies have become imperative to mitigate these adverse outcomes and improve overall well-being in aging societies [6].

Among the multiple risk factors for falls, gait and balance impairments emerge as predominant risk factors in older adults [7–9]. Specific indicators, including decreased walking speed, lower Berg Balance Scale (BBS) scores, and prolonged Timed Up and Go (TUG) test duration, have been demonstrated to be associated with fall risk [10, 11]. Conventional training methods, such as strength and balance training [12–14], whole-body vibration [15], and gait training [16], have been proven to improve gait, balance, or reduce fall risk in older adults. These methods employ a bottom-up intervention paradigm, utilizing physical training to enhance motor function at the bodily level (bottom), thereby inducing adaptive changes in the nervous system (top), ultimately promoting functional recovery [17]. However, conventional training methods primarily focus on functional impairments and present several limitations: limited relevance to activities of daily living, poor generalization of functional gains, or limited translation to real-world functional improvements [12–15, 18, 19]. Furthermore, they often lack personalized design and explicit neuroplasticity-based mechanisms, with the absence of personalized design frequently resulting in suboptimal patient engagement [17, 20].

In contrast, top-down rehabilitation strategies leverage higher-order brain functions through active participation and task-specific training to promote neural plasticity and facilitate motor recovery [17]. One such strategy, task-oriented training (TOT), is a patient-centered method that operationalizes this principle by having patients repeatedly practice specific, functionally relevant tasks in real or simulated environments to restore function [18, 21–24]. TOT customizes tasks according to patients’ needs and goals, optimizing outcomes through timely feedback and progressive difficulty [18, 21–24]. It uses real tasks as therapy, enabling patients to naturally and efficiently reorganize motor patterns during task completion. The fundamental distinction from conventional training lies in TOT’s direct focus on “function,” using tasks themselves as treatment [21, 24]. This makes training more engaging and personalized, sustaining patient motivation [21, 25]. In contrast, conventional training often starts with isolated “single muscle or joint” exercises before combining them into functions, which can become monotonous, reduce engagement, and lack real-life relevance [25]. The use of virtual or augmented reality technologies for gait and balance training [19, 26], as well as balance training during daily tasks such as “walking while holding a cup” and “going up and down stairs” [27], are typical examples that distinguish TOT from conventional training methods. Furthermore, TOT can take various forms in clinical practice, such as constraint-induced movement therapy, circuit training, task-related training, and task-specific training, all reflecting its task-centered rehabilitation approach [17–19, 21, 24, 28]. Based on these principles and implementation characteristics, TOT has demonstrated application value across various clinical populations, though its efficacy varies by population and functional domain. Existing studies demonstrate that TOT surpasses conventional rehabilitation in improving motor function in children with cerebral palsy [18] and gait performance in stroke survivors [24], while its benefits for upper limb function in multiple sclerosis remain unclear [19]. A previous review found TOT significantly improves balance, mobility, and functional performance in older adults, making it an effective fall prevention intervention [29]. Another review found that compared to conventional training or health education, TOT moderately improved balance in older adults with mild cognitive impairment, but did not significantly improve gait [30]. However, progressive resistance training in older adults was more effective than TOT at increasing walking speed [31]. Therefore, there is currently inconsistent evidence regarding the comparative effectiveness of TOT versus conventional training in improving balance, gait, and fall prevention among older adults.

Given these gaps, this study aims to evaluate the effectiveness of TOT compared with conventional training in enhancing balance, gait, and reducing fall risk among older adults through a systematic review and meta-analysis. The findings will provide evidence-based support for clinical practice and community-based rehabilitation, thereby optimizing exercise interventions for older adults and establishing a theoretical foundation for developing scientifically validated fall prevention strategies.

Methods

We strictly followed the guidelines of the PRISMA [32] and registered in the PROSPERO platform. The registration ID is CRD42025635269. This systematic review meets the high-quality criteria of AMSTAR 2 (Supplementary File 1) [33]. Ethical approval is not required because the information used in this study is obtained from published randomized controlled trials (RCTs).

Search strategy

We systematically searched multiple major databases from their inception to October 2, 2025, including PubMed, Web of Science, Embase, as well as one clinical trial registry (http://www.ClinicalTrials.gov). Language restrictions were applied to English during the retrieval process. Two authors independently screened the articles to determine their eligibility based on the predefined inclusion criteria. Based on a review of relevant literature [17–19, 21, 24, 28], we have formulated the following search terms related to TOT: “task-oriented,” “task-specific,” “repetitive task,” “constrained-induced,” “task-related,” “circuit-based exercise,” “circuit training,” “circuit class,” “motor learning,” “game-based,” “virtual reality,” and “robotics.” Additional search terms included: “fall,” “drop,” “trip,” “imbalance,” “fracture,” “balance,” “gait,” “elderly,” “older adult,” “aging,” “aged,” “geriatric,” “senior,” “old,” “randomized controlled trial,” “controlled clinical trial,” “randomized,” “placebo,” “randomly,” “trial.” The detailed search strategy is provided in Supplementary File 2.

Eligibility criteria

This review included studies involving older adults aged 60 years and above. The experimental group belonged to TOT. A training program is deemed qualified as TOT only when it meets all of the following criteria: clear and specific task objectives, relevance to daily life, a focus on functional tasks, active client participation, repetitive practice, provision of immediate feedback, meaningfulness to the client, and a realistic or simulated real-world training environment. In contrast, the control group received conventional training. This included any exercise programs other than TOT (e.g., strength training, balance training, gait training, or any other types of exercise training that do not comply with TOT principles), as well as conventional non-exercise interventions (such as fall prevention education or general health education sessions). Included studies were required to report at least one outcome measure related to balance, gait, or fall risk. This review included only RCTs.

Studies were excluded under the following conditions: the experimental group received group-based training; the study lacked a control group; the control group received no intervention; the study lacked the relevant outcome measures; the trial protocols did not report final results; the study was a non-RCT (e.g., reviews, conference abstracts, cell or animal experiments) or had a crossover design. Furthermore, studies involving older adults with orthopedic, cardiovascular, or neurological disorders were also excluded. Notably, no restrictions were applied regarding the dosage parameters of the interventions, including weekly frequency and total duration.

For RCTs involving more than two intervention groups, the following procedures were adopted to ensure comparability of interventions and consistency in analysis. The suitability of each intervention group as an “experimental group” or “control group” was independently assessed based on the inclusion and exclusion criteria. Only the intervention groups meeting TOT criteria were included in the experimental group, and only those meeting conventional training criteria were included in the control group. All other groups were excluded. If multiple intervention groups within a study meet the inclusion criteria for the experimental group (or for the control group), we followed the recommendation of the Cochrane Handbook by combining these multiple groups into a single group [34].

Outcome measures

The outcome measures were categorized into four domains: (1) balance performance, assessed with the Berg Balance Scale (BBS) and Timed Up and Go test (TUG test); (2) gait function, evaluated by the Functional Gait Assessment (FGA) and gait speed (GS); (3) fall-related psychological measures, including the Activities-specific Balance Confidence Scale (ABC) and the Falls Efficacy Scale–International (FES-I); and (4) fall incidence.

Study screening and data extraction

All records were imported into reference management software (EndNote 21), and duplicate records from the same trial were subsequently removed. Two reviewers independently screened the titles and abstracts, and then proceeded to read the full texts to select articles that fully met the inclusion and exclusion criteria. Subsequently, they independently extracted and summarized data on the following parameters: author, publication year, population type, age, gender, number of participants, interventions (control and experimental groups), training frequency, treatment duration, and outcome measures. Outcome data from the immediate post-intervention assessment were extracted for inclusion in the meta-analysis to evaluate the direct effects of the interventions. After completion, the extracted data were cross-checked. Any disagreements were resolved through discussion or by consulting a third reviewer.

Assessment of methodological quality and risk of bias

The methodological quality and risk of bias of the included studies were evaluated using two complementary tools. The PEDro scale was applied to provide a broad assessment of internal validity and statistical reporting, yielding an overall quality score. This scale consists of 11 items used to rate the methodological quality of the studies, with each item scored as either 1 (indicating “Yes”/present) or 0 (indicating “No”/absent), except for the first item, which is not assigned a score. The total quality scores were categorized as follows: 6–10 (high quality), 4–5 (moderate quality), and 0–3 (low quality) [35–37]. For a more detailed, domain-specific evaluation of bias in the RCTs, the Cochrane Risk of Bias tool was used, thereby ensuring a comprehensive appraisal [38]. The standards contain 7 aspects: random sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other bias. A grade of ‘high’, ‘low’, or ‘unclear’ was given for each item. To enhance objectivity and minimize potential bias, two reviewers independently conducted the assessments for methodological quality and risk of bias. Any discrepancies between their evaluations were resolved through discussion or, when necessary, by consulting a third author.

Statistical analysis

Two statistical software programs, Stata18 and RevMan 5.4, were used for the meta-analysis. The mean difference (MD) with 95% confidence intervals (CI) was selected as the effect measure for continuous variables. A p-value < 0.05 was considered statistically significant. Higgins et al. suggested that I² values of 25%, 50%, and 75% stand for low, moderate, and high thresholds of heterogeneity, respectively [39]. Moreover, a meta-analysis with a low heterogeneity (25% ≤ I² < 50%) implied that the variability had only a small effect. Therefore, a random-effect model was used for outcomes that displayed significant heterogeneity with I² values > 50%; otherwise, the fixed-effect model was used. If heterogeneity existed, subgroup analysis or sensitivity analysis was required to find the source of heterogeneity. Sensitivity analysis was conducted by sequentially excluding each individual study to assess the robustness of the pooled results. Egger’s test [40] was employed to assess publication bias. The absence of publication bias was confirmed by Egger’s test (P > 0.05). The Minimal Clinically Important Difference (MCID) refers to the smallest change in a treatment outcome that a patient or clinician would identify as being clinically meaningful or practically valuable [41]. We first reviewed the relevant literature to obtain the MCID values for each outcome measure and then compared the effect sizes from our meta-analysis with these MCID values to evaluate the clinical significance of the intervention [42]. The MCID for the outcome measures was defined a priori as follows: 1.9 points for the BBS [42, 43], 1.6 s for the TUG test [41, 42], and 0.05 m/s for self-selected gait speed [42, 44]. No reported MCID value was identified for the FES-I [42, 45].

Results

Selection and inclusion of studies

The PRISMA flowchart illustrated the study selection process (Fig. 1). The initial search identified 4341 records, of which 939 were duplicates. After removing duplicate articles, 3402 records remained. Based on the PICOS principles, titles and abstracts were screened, leading to the exclusion of an additional 3189 articles. The remaining 213 articles underwent full-text review according to the inclusion and exclusion criteria. Ultimately, 20 studies were included [26, 27, 46–63], with 15 studies included in the meta-analysis [26, 27, 47–49, 51, 53, 55–57, 59–63]. 5 studies were unable to be included in the meta-analysis and were synthesized in narrative form instead because they did not report sufficient data to calculate effect sizes [46, 50, 52, 54, 58].

Fig. 1.

Literature search of the included studies. The term ‘records’ refers to initial database entries screened by title and abstract, while ‘reports’ refers to full-text articles retrieved for detailed assessment

Characteristics of the studies

The main characteristics of selected studies were summarized in Table 1. The participants comprised older adults aged ≥ 60 years (n = 927). Among the 20 studies, 8 assessed BBS [27, 46, 48–51, 54, 57], 2 examined ABC [50, 59], 12 explored TUG test [26, 27, 46, 48–50, 52–56, 61], 2 included FGA [46, 58], 6 measured GS [47, 50, 51, 53, 60, 62], 4 evaluated FES-I [27, 48, 54, 63], and 2 reported fall incidence [46, 51]. Regarding the types of TOT, 13 studies employed virtual reality-based TOT [46, 48, 49, 52–58, 61, 63, 64], while the remaining 7 studies utilized non-virtual reality forms of TOT (functional circuit-training program [59, 60], task-oriented motor learning exercise [47, 62], progressive balance-specific exercises [51], task-oriented balance training [27] and task-oriented dynamic balance training program integrated with interactive computer games [50]). As for the types of conventional training, 17 studies adopted conventional exercise training [48, 49, 51, 57], and 3 studies employed conventional non-exercise training (health education meetings [59, 60] and fall prevention education [27]). Regarding the total intervention duration, 7 studies implemented a 6-week intervention [26, 27, 48, 49, 51, 52, 56], 1 study conducted a 7-week intervention [58], 6 studies employed an 8-week intervention [50, 53–55, 57, 63], 1 study carried out a 9-week intervention [61], 4 studies implemented a 12-week intervention [47, 59, 60, 62], and 1 study performed a 12-month intervention [46].

Table 1.

Characteristics of the studies included in the systematic review and meta-analysis

Studies	Population type	Age (years)(Mean ±SD)	Sample Size (Male/ Female)	Intervention	Control	Duration, Frequency and Cycle	Outcomes
Szturm et al., 2011	Community dwelling and ambulatory	EG = 80.5 ± 6.0; CG = 81.0 ± 7.0	EG = 15(5/10); CG = 15(6/9)	Dynamic balance exercises coupled with computer games: Underwent a task-oriented dynamic balance training program integrated with interactive computer games. Games included single-axis and dual-axis movement challenges requiring weight shifting in multiple directions and varying speeds.	Typical rehabilitation program: Received the typical rehabilitation program consisting of strengthening exercises using Thera-Band and leg weights, endurance training on a cycle ergometer, standing balance exercises (e.g., hip flexion, side-leg raises, squats, sit-to-stand), gait re-education within parallel bars (e.g., heel-to-toe walking, lateral stepping, step-ups), and unsupervised walking practice with assistive devices.	EG and CG: 45 min, 2 times/wk, 8 wk	BBS ABC TUG GS (self-selected)
Tsang et al., 2016	Nursing home residents	EG = 82.3 ± 3.8; CG = 82.0 ± 4.3	EG = 39(16/23); CG = 40(15/25)	Wii Fit balance training: Utilized Nintendo Wii Fit balance games, including Soccer Heading, Table Tilt, and Balance Bubble, to deliver virtual reality-based balance training.	Conventional balance training: Conducted conventional balance training led by a physiotherapist, comprising leg strengthening, tandem standing and walking, sideways walking, turning, stepping, sit-to-stand, and mini-squat exercises.	EG and CG: 60 min, 3 times/wk, 6 wk	BBS TUG
Yesilyaprak et al., 2016	Nursing home residents	EG = 70.1 ± 4; CG = 73.1 ± 4.5	EG = 7(3/4); CG = 11(9/2)	Balance training with the BTS Nirvana virtual reality interactive system: Balance tasks involving static-dynamic balance, functional reach, and hand-eye coordination exercises projected onto walls or floors, with real-time visual and auditory feedback.	Conventional balance exercise: Performed exercises including marching in place, single-leg and tandem standing, sit-to-stand, heel raises, leg lifts, and multidirectional walking.	EG and CG: 35–45 min, 3 times/wk, 6wk	BBS TUG FES-I
Bacha et al. 2018	Community-dwelling older adults	EG = 71.0; CG = 66.5	EG = 23; CG = 23	Kinect Adventures training : Participants engaged in individualized training using the Xbox Kinect Adventures games, which required full-body movement and interaction with a virtual environment. The games included “20,000 Leaks,” “Space Pop,” “Reflex Ridge,” and “River Rush,” each demanding multidirectional weight shifting, rapid limb movements, jumping, squatting, and trunk inclination.	Conventional physical therapy : Performed therapy including warm-up walking, static and dynamic balance exercises on unstable surfaces, step-up/step-down endurance tasks, muscle strengthening (e.g., sit-to-stand, hip/shoulder adduction with ball squeezes), motor coordination circuits, global stretching, and cool-down respiratory and joint mobilization exercises.	EG and CG: 60 min/day, 2 times/wk, 7 wk	FGA
Htut et al. 2018	Nursing home residents	EG = 75.8 ± 4.89; CG = 75.95 ± 5.65	EG = 21(10/11); CG = 21(10/11)	Virtual reality-based exercise: Participants performed interactive exercises using Xbox 360 virtual reality games. Selected games involved stepping, trunk bending, weight shifting, upper and lower limb movements, and reaction-based tasks. These activities provided real-time visual feedback and cognitive engagement, promoting both motor and cognitive training in a dynamic, technology-enhanced environment.	Physical exercise: Participants engaged in conventional physical training consisting of warm-up and cool-down activities, along with strength and balance exercises. The program included upper and lower limb strengthening tasks (e.g., chest press, sit-to-stand squat, hip abduction), balance challenges (e.g., single-leg stance, tandem walk), and flexibility routines.	EG and CG: 30 min, 3 times/wk, 8wk	BBS TUG FES-I
Babadi et al. 2021	Nursing home residents	EG = 66.50 ± 3.80; CG = 67.50 ± 3.11	EG = 12; CG = 12	Virtual reality training: Participants engaged in balance training using Xbox Kinect exergames, including boxing, table tennis, soccer, golf, skiing, and American football. These games required full-body movements such as stepping, weight shifting, trunk bending, and coordinated limb actions.	Conventional balance training : Participants performed traditional balance exercises in a standing position, including double-leg stance, heel-to-toe raises, walking in place, neck and trunk rotations, one-leg stance, and sideways walking. The program progressively reduced the base of support, increased movement speed, and incorporated eyes-closed tasks.	EG and CG: 60 min, 3 times/wk, 9 wk	TUG
Sadeghi et al. 2021	Community-dwelling older men	EG = 74.1 ± 7.0; CG = 70.4 ± 4.3	EG = 16; CG = 16	Virtual reality training: Participants engaged in interactive Xbox Kinect exergames, including Light Race, Target Kick, and Goalkeeper. These games required stepping, jumping, crouching, and coordinated hand-eye/foot-eye movements, providing real-time visual feedback and cognitive-motor challenges.	Traditional balance training: Participants performed exercises including single-leg stance, heel-to-toe standing, tandem walking, forward/backward walking, and weight-shifting tasks.	EG and CG: 40 min, 3 times/wk, 8 wk	TUG GS (self-selected)
VanSwearingen et al. 2011	Community-dwelling older adults	EG = 77.2 ± 5.5; CG = 77.2 ± 5.5	EG = 23; CG = 24	Task-oriented, motor sequence learning exercise: Engaged in task-oriented, motor sequence learning exercises involving stepping and walking patterns. Activities included diagonal steps, oval/spiral path walking, variable-speed treadmill walking at preferred pace with brief speed bursts, and dual-task walking with upper-extremity challenges (e.g., carrying or tossing a ball).	Impairment-oriented, multicomponent exercise: Performed impairment-oriented, multicomponent exercises including lower-extremity strength training with progressive resistance, static and dynamic balance exercises, endurance training using recumbent stepper or stationary bike at moderate intensity, and gait training.	EG and CG: 2 times/wk, 12wk	GS (self-selected)
Zhao et al. 2023	Nursing home residents	EG = 72.16 ± 3.64; CG = 73.36 ± 3.25	EG = 25(13/12); CG = 25(8/17)	Virtual reality rehabilitation training: Underwent virtual reality-based rehabilitation training using a virtual reality system, participating in interactive sports games (skiing, diving, running) while wearing virtual reality headsets, incorporating warm-up and structured cool-down activities.	Traditional fall prevention exercise intervention: Received traditional fall-prevention exercises led by an instructor, including medium-intensity aerobic gymnastics and apparatus-based exercises targeting core and lower limb muscle strength, balance, and gait function.	EG and CG: 50 min, 3 times/wk, 12mo	BBS TUG FGA Fall incidence
Fidan et al. 2024	Nursing home residents	EG = 75.53 ± 8.43; CG = 75.57 ± 7.88	EG = 15; CG = 14	Virtual reality rehabilitation training: Engaged in individual virtual reality training using Xbox 360 Kinect, performing Tai Chi-inspired movements (e.g., weight shifting, arm lifting, squats, single-leg movements). Movement difficulty progressed based on individual performance, with feedback provided by the virtual coach and avatar system.	Traditional exercises: Received traditional group exercises led by a physiotherapist, including breathing control, postural training, general mobility exercises, walking practice, and whole-body stretching routines.	EG and CG: 45 min, 2 times/wk, 8 wk	TUG
Morone et al. 2016	Community-dwelling older adults(Women with bone loss condition	EG = 67.8 ± 2.98; CG = 70.05 ± 4.93	EG = 19(0/19); CG = 19(0/19)	Balance training with a Wii Fit: Engaged in supervised Wii Fit-based exergaming including center-of-gravity control tasks, yoga poses, single-leg and lateral leg extensions, squats, and interactive balance games (e.g., balance bubble, ski slalom, table tilt), with real-time visual and auditory feedback.	Conventional balance exercises: Received conventional balance training including stretching, antigravity muscle strengthening, balance exercises on unstable platforms, and postural control training with diaphragmatic breathing, all aimed at improving stability and axial control.	EG and CG: 60 min, 2 times/wk, 8 wk	BBS
Campelo et al. 2023	Community-dwelling older adults	EG = 72 ± 6; CG = 70 ± 5(male) EG = 73 ± 9; CG = 73 ± 5(female)	EG = 15(6/9); CG = 14(5/9)	Exergame training: Participants used the Nintendo Wii-U console (Wii Fit-U games) to perform interactive exercises in four categories: aerobics, strength, balance, and flexibility. Each session included warm-up and training segments using the Wii Balance Board and Wii Remote.	Conventional training: Participants engaged in a structured exercise program including aerobic, strength, balance, and flexibility activities. Exercises were performed seated, standing, or on a mat, with resistance adjusted using elastic bands or ankle weights. Instructors individualized intensity and progression every two weeks to maintain moderate exercise load.	EG and CG: 3 times/wk, 6 wk	TUG
Sápi et al. 2019	Community-dwelling older adults	EG = 69.57 ± 4.66; CG = 69.12 ± 4.19	EG = 30(1/29); CG = 23(1/22)	Kinect training: Participants performed full-body, gesture-controlled exergames using Microsoft Xbox 360 Kinect, including games like bowling, football, skiing, and dance, as well as cognitive-motor games (e.g., 20,000 Leaks, Space Pop, Reflex Ridge, River Rush). Movements involved weight shifting, stepping, squatting, lunging, reaching, and hopping.	Conventional balance training: Participants engaged in balance exercises, including static and dynamic tasks such as weight shifting, forward/backward/side stepping, obstacle negotiation, reaching, squatting, lunging, and hopping.	EG and CG: 3 times/wk, 6 wk	TUG
Steadman et al. 2003	Community-dwelling older adult(Elderly patients with balance problems)	EG = 82; CG = 82	EG = 82; CG = 84	Enhanced balance training: Participants performed a structured program of progressive balance-specific exercises integrated into functional tasks. Activities included repetitive sit-to-stand, lateral reaching, 360-degree turning, tandem standing, one-leg stance, step-ups, balance board training, and walking practice with timed feedback.	Conventional therapy: Participants received standard physiotherapy focused on general mobility and functional rehabilitation. Interventions included assisted walking, transfer training, stair practice, bed mobility, and gait training with mobility aids as needed.	EG: 45 min, 2 times/wk, 6 wk CG: 45 min, 2 times/wk, 4 wk + 2 wk (follow-up)	BBS Fall incidence GS (self-selected)
Giné-Garriga et al. 2010• o	Primary health care center	EG = 83.9 ± 2.8; CG = 84.1 ± 3	EG = 22(9/13); CG = 19(7/12)	Functional circuit-training program: Balance training included static and dynamic tasks such as single-leg and tandem standing, walking on varied surfaces, obstacle negotiation, and dual-task activities (e.g., walking while catching or carrying objects). Strength exercises included chair rises, stair climbing, lunges, squats, and leg extensions using ankle weights.	Health education meetings: Participants continued their usual daily activities and received standard care from their primary-care providers. They attended weekly social meetings with researchers and participated in health education sessions covering topics such as nutrition, medication use, foot care, and sleep hygiene.	EG: 45 min, 2 times/wk, 12wk CG: 1 time/wk, 12 wk.	GS (self-selected)
Sik Oh et al. 2021	Community-dwelling older adults	EG = 78.70 ± 2.62; CG = 77.20 ± 8.29	EG = 11(3/8); CG = 12(4/8)	Task-oriented training: Participants performed balance training incorporating real-life tasks. Activities included weight-shifting while seated on a Swiss ball, maintaining balance on an air cushion, walking along straight or curved lines, obstacle avoidance, stair climbing, and object manipulation (e.g., moving cups, replacing towels).	Fall prevention education: Participants received fall prevention education through audio-visual materials. Sessions covered general health topics relevant to older adults, with no physical or cognitive training components.	EG and CG: 40 min, 3 times/wk, 6wk	BBS TUG FES-I
Gine´-Garriga et al. 2013	Primary health care center in the Barcelona area	EG = 83.9 ± 2.8; CG = 84.1 ± 3	EG = 22(9/13); CG = 19(7/12)	Functional circuit training program: Balance training included static tasks (e.g., single-leg and tandem standing on varied surfaces with eyes open/closed) and dynamic tasks (e.g., walking on uneven surfaces, narrow walking, obstacle avoidance, dual-task activities). Strength training included chair rises, stair climbing, lunges, squats, and leg extensions and calf raises using ankle weights.	Routine activities and health education meetings: Participants continued their usual daily activities and received standard care from their primary-care providers. They attended weekly social meetings with researchers and participated in health education sessions covering general wellness topics.	EG: 2 times/wk, 12wk CG: 1 times/wk, 12wk	ABC
Parmak et al. 2025	Community-dwelling older adults	EG = 71.14 ± 4.83; CG = 75.36 ± 9.16	EG = 22(4/18); CG = 22(4/18)	Immersive virtual reality training: Home-based immersive virtual reality training using Oculus Quest 2 and FIT-XR apps (Box & Slam) with real-time difficulty adjustment, under on-site supervision.	Conventional exercise: Home-based conventional exercise program including balance, strength, flexibility, and relaxation with illustrated brochure and weekly phone follow-up.	EG and CG: 35 min, 3 times/wk, 8wk	FES-I
Abd El-Kafy et al. 2024	Community-dwelling, healthy elderly at risk of falls	EG = 66.53 ± 3.82; CG = 65.43 ± 4.28	EG = 29(23/6); CG = 28(21/7)	C- Mill virtual reality/augmented reality treadmill training: Participants walked on the C-Mill treadmill while interacting with projected visual obstacles and stepping targets (augmented reality) and with immersive virtual scenes displayed on a front screen (virtual reality), then immediately performed conventional postural-reaction drills on balance boards/beams and overground gait exercises involving cones, steps, and uneven surfaces.	Conventional exercise: Participants walked on the same C-Mill treadmill without any visual projections or virtual reality/augmented reality content, then performed the identical conventional postural-reaction and overground gait exercises.	EG and CG: 60 min, 3 times/wk, 6wk	TUG
Brach et al. 2013	Community-dwelling older adults	EG = 75.7 ± 5.5; CG = 78.5 ± 6.2	EG = 18(8/10); CG = 20(7/13)	Motor learning group: Goal-oriented stepping and walking exercises to improve timing, coordination, and skill. Progressed by varying speed, amplitude, and task complexity.	Standard training: Endurance training was performed as treadmill walking at a self-selected pace.	EG and CG: 60 min, 2 times/wk, 12wk	GS (self-selected)

EG(Experimental group), CG(Control group)Timed Up and Go Test (TUG test), Berg Balance Scale (BBS), and number of falls. The secondary outcomes comprised the Falls Efficacy Scale-International (FES-I), Gait speed (GS), Functional Gait Assessment (FGA), Activities-specific Balance Confidence Scale (ABC)

Characteristics of the studies included in the systematic review and meta-analysis

Methodological quality of the included studies

A total of 19 studies were assessed as high quality [26, 27, 47–63], while the remaining one was of moderate quality [46]. All included studies clearly defined the eligibility criteria and conducted random allocation. Supplementary File 3 provides detailed information, including the baseline and post-training values for each outcome measure, as well as the corresponding p-values from the original studies. The baseline characteristics of the two groups were similar across all studies for the different outcome measures. In all studies, at least one primary outcome was measured in ≥ 85% of initially allocated participants. All studies reported between-group comparisons, with point estimates and measures of variability provided for key outcomes in all studies. Six studies had allocation concealment [47, 50, 55–57, 60]. Only one study employed participant blinding [52], and no studies employed therapist blinding. Except for 6 studies [46, 48, 49, 52, 61, 63], all utilized outcome assessor blinding. Eight studies were judged as not having conducted an intention-to-treat analysis because they explicitly stated in their methods that they analyzed data only from participants who completed the trial, thereby excluding dropouts from the analysis [26, 46, 47, 50, 52, 53, 56, 62]. Table 2 provides detailed methodological information on the included studies.

Table 2.

Methodological quality assessment of included studies (PEDro scale)

Inclusion study	1	2	3	4	5	6	7	8	9	10	11	Total
Szturm et al. 2011	1	1	1	1	0	0	1	1	0	1	1	7
Tsang et al. 2016	1	1	0	1	0	0	0	1	1	1	1	6
Yesilyaprak et al. 2016	1	1	0	1	0	0	0	1	1	1	1	6
Bacha et al. 2018	1	1	0	1	0	0	1	1	1	1	1	7
Htut et al. 2018	1	1	0	1	0	0	1	1	1	1	1	7
Babadi et al. 2021	1	1	0	1	0	0	0	1	1	1	1	6
Sadeghi et al. 2021	1	1	0	1	0	0	1	1	0	1	1	6
VanSwearingen et al. 2011	1	1	1	1	0	0	1	1	0	1	1	7
Zhao et al. 2023	1	1	0	1	0	0	0	1	0	1	1	5
Fidan et al. 2024	1	1	1	1	0	0	1	1	1	1	1	8
Morone et al. 2016	1	1	1	1	0	0	1	1	1	1	1	8
Campelo et al. 2023	1	1	1	1	0	0	1	1	0	1	1	7
Sápi et al. 2019	1	1	0	1	1	0	0	1	0	1	1	6
Steadman et al. 2003	1	1	0	1	0	0	1	1	1	1	1	7
Giné-Garriga et al. 2010	1	1	1	1	0	0	1	1	1	1	1	8
Sik Oh et al. 2021	1	1	0	1	0	0	1	1	1	1	1	7
Gine´-Garriga et al. 2013	1	1	0	1	0	0	1	1	1	1	1	7
Parmak et al. 2025	1	1	0	1	0	0	0	1	1	1	1	6
Abd El-Kafy et al. 2024	1	1	0	1	0	0	1	1	0	1	1	6
Brach et al. 2013	1	1	0	1	0	0	1	1	0	1	1	6

1:eligibility criteria; 2:randomly allocated; 3:assigning concealment; 4:similar at baseline; 5:blinding of all subjects;6:blinding of all therapists; 7:blinding of all assessors; 8:Measures of at least one key outcome were obtained from more than 85% of the subjects initially allocated to groups; 9:intention to treat;10:comparison between groups; 11:point measures and measures of variability

Risk of bias in studies

Four of the included studies were rated as ‘unclear’ because they only stated that participants were randomly assigned but did not describe the specific randomization method [47, 49, 52, 61]; 1 study exhibited a high risk of selection bias due to allocation concealment [51]; Allocation concealment was unclear in 13 studies [26, 27, 46–49, 52–54, 58, 59, 61, 62]; All studies carried a high risk of performance bias, as blinding of participants and personnel was not feasible; 3 studies had a high risk of detection bias because outcome assessors were not blinded [46, 49, 63];3 studies did not report blinding of outcome assessors [48, 52, 61]. All studies were at low risk of attrition bias, reporting bias, and other bias (Supplementary File 4). Furthermore, among the 20 included studies, 11 reported clinical trial registration information [26, 47, 50, 52–55, 58, 61–63], while the remaining 9 did not mention it in their manuscripts [27, 46, 48, 49, 51, 56, 57, 59, 60].

The effect of TOT on balance function

Berg balance scale (BBS)

A total of 8 studies reported post-intervention BBS outcomes [27, 46, 48–51, 54, 57], with 5 studies [27, 48, 49, 51, 57] included in the meta-analysis due to the availability of complete numerical data. The analysis involved 324 participants (TOT group: 158; Conventional training group: 166). Among the five studies [27, 48, 49, 51, 57], one study implemented an 8-week intervention [57], while the other four studies employed 6-week interventions [27, 48, 49, 51]. Three implemented task-oriented training via virtual reality for the experimental group [48, 49, 57], while the remaining two employed non-virtual reality forms of TOT (progressive balance-specific exercises [51] and task-oriented balance training [27]). In terms of the control groups, one received a non-exercise intervention (fall prevention education) as part of conventional training [27], and the other four involved exercise-based interventions [48, 49, 51, 57].

The results demonstrated statistically significant differences (MD = 2.58, 95% CI 1.80 to 3.36, P < 0.001, I² = 0%), indicating that TOT was more effective than conventional training in improving BBS scores among older adults (Fig. 2). The results reached the MCID of the BBS(2.58 > 1.9). Moreover, narrative synthesis of the three studies, which provided insufficient data for meta-analysis, still showed significant findings [46, 50, 54].

Fig. 2.

Forest plot for the BBS

Timed up and go test (TUG test)

A total of 12 studies recorded TUG test scores [26, 27, 46, 48–50, 52–56, 61], with 8 studies involving 288 participants (TOT group: 143; Conventional training group: 145) included in the meta-analysis due to the availability of complete numerical data [26, 27, 48, 49, 53, 55, 56, 61]. Furthermore, regarding the intervention duration, among the eight studies, one had an intervention period of 9 weeks [61], two studies implemented 8-week interventions [53, 55], and the remaining five studies all adopted a 6-week intervention period [26, 27, 48, 49, 56]. For the intervention in the experimental groups, seven out of the eight studies implemented TOT via virtual reality [26, 48, 49, 53, 55, 56, 61], while only one utilized non-virtual reality-based TOT [27]. As for the control group interventions, one study utilized a non-exercise intervention (fall prevention education) [27], while the other seven studies all employed exercise-based interventions [26, 48, 49, 53, 55, 56, 61].

There was a significant difference between the two groups in improving TUG test performance in older adults (MD = −0.55, 95%CI −0.81 to −0.29, P < 0.001, I² = 48%) (Fig. 3a). The results did not reach the MCID for the TUG test (0.55 < 1.6). Narrative synthesis of the four studies, which provided insufficient data for meta-analysis, suggested inconsistent results. Szturm et al. found no significant difference in TUG test scores between the TOT group and the conventional training group [50]. Two studies demonstrated that TOT was more effective than conventional training in improving TUG test performance among older adults [46, 52]. However, Htut et al. reported opposite findings [54].

Fig. 3.

(a) Forest plot for the TUG test, (b) Subgroup analysis of the TUG test, stratified by mean age, (c) Subgroup analysis of the TUG test, stratified by the type of control intervention, (d) Subgroup analysis of the TUG test, stratified by intervention duration. 

The subgroup analysis results indicated that TOT was more effective than conventional training in improving the TUG test performance in older adults with a mean age ≥ 75 years (MD = −1.61, 95% CI −2.60 to −0.63, P = 0.001, I² = 0%) and 60–74 years (MD = −0.47, 95% CI −0.74 to −0.20, P < 0.001, I² = 40%) (Fig. 3b). Significant difference was observed between the two age subgroups (test for subgroup differences: χ² = 4.76, P = 0.03, I² = 79.0%), suggesting that age may be a major moderator of the intervention effect. Notably, the magnitude of improvement in TUG test scores induced by TOT reached the MCID threshold in the aged over 75 years subgroup (1.61 > 1.6).

Subgroup analyses based on the type of control intervention (exercise and non-exercise control groups) demonstrated that TOT was more effective than the exercise control group in improving TUG test performance (MD = −0.52, 95% CI −0.78 to −0.25, P < 0.001, I² = 29%) (Fig. 3c). Significant difference was observed between the two subgroups (test for subgroup differences: χ² = 4.96, P = 0.03, I² = 79.8%).

Subgroup analysis based on intervention duration revealed a statistically significant effect for the 6-week intervention group (MD = −1.26, 95% CI −2.18 to −0.33, P = 0.008, I² = 66%) (Fig. 3d). However, effects for the 8-week and 9-week intervention groups did not reach statistical significance. Due to the limited number of studies included, these non-significant results should be interpreted with caution. The between-subgroup differences were not statistically significant (test for subgroup differences: χ² = 3.18, P = 0.20, I² = 37.0%).

The impact of TOT on gait function

Functional gait assessment (FGA)

A total of 2 RCTs reported on the FGA outcomes in older adults after intervention [46, 58]. One study found that TOT had lower FGA results compared to conventional training [58]. However, one study found that TOT led to significantly greater improvements in FGA than conventional training [46].

Gait speed(GS)

A total of 6 studies reported changes in gait speed post-intervention [47, 50, 51, 53, 60, 62], with 4 studies involving 145 participants (TOT group: 72; Conventional training group: 73) included in the meta-analysis due to the availability of complete numerical data [47, 53, 60, 62]. Among the four studies, one employed an 8-week intervention [53], while the other three lasted 12 weeks [47, 60, 62]. Among the intervention measures for the experimental group, only one of the four studies implemented TOT in a virtual reality format [53], while the other three studies involved a functional circuit-training program (including balance and lower-body strength exercises) [60], task-oriented motor sequence learning exercises incorporating stepping and walking patterns [47], and goal-oriented stepping and walking exercises [62], respectively. In terms of the control groups, one received a non-exercise intervention (health education meetings) as part of conventional training [60], and the other three involved exercise-based interventions [47, 53, 62].

A significant difference was observed between the TOT group and the conventional training group in improving gait speed (MD = 0.14, 95% CI 0.12 to 0.16, P < 0.001, I² = 0%) (Fig. 4). The results reached the MCID for GS (0.14 > 0.05). Narrative synthesis of the two studies, which provided insufficient data for meta-analysis, suggested that neither study found significant differences [50, 51].

Fig. 4.

Forest plot for the GS

The effect of TOT on the fall-related psychological measures

Activities-specific balance confidence scale (ABC)

A total of 2 RCTs reported on the ABC outcomes in older adults after intervention [50, 59]. One study involving 41 participants (TOT group: 22; Conventional training group: 19) was included in the meta-analysis due to the availability of complete numerical data [59]. A statistically significant difference was observed between the TOT group and conventional training group (MD = 17.06, 95% CI 7.71 to 26.41, P < 0.001). A narrative analysis of the other study also found that the TOT group was more effective in improving ABC scores [50].

Falls efficacy Scale-International (FES-I)

A total of 4 RCTs reported FES-I scores in older adults post-intervention [27, 48, 54, 63]. 3 studies involving 85 participants (TOT group: 40; Conventional training group: 45) were included in the meta-analysis due to the availability of complete numerical data [27, 48, 63]. Among the three studies, one employed an 8-week intervention [63], while the other two lasted 6 weeks [27, 48]. Of the three studies investigating the experimental intervention, two implemented TOT using a virtual reality format [48, 63], while the remaining study employed a non-virtual reality format of TOT (task-oriented balance training) [27]. In terms of the control groups, one received a non-exercise intervention (fall prevention education) as part of conventional training [27], and the other two involved exercise-based interventions [48, 63].

A statistically significant difference was observed between TOT group and conventional training group in reducing fear of falling (MD = −2.80, 95%CI −5.30 to −0.29, P = 0.03, I² = 47%) (Fig. 5a). One study, which was excluded from the meta-analysis due to insufficient data, was included in the narrative syntheses and also found that TOT was more effective than conventional training in reducing the fear of falling [54].

Fig. 5.

(a) Forest plot for the FES-I, (b) Sensitivity analysis for the FES-I.

Sensitivity analysis conducted by systematically excluding individual studies demonstrated that the study by Sik Oh was the primary source of heterogeneity [27]. This study differs from the other two in that the control group received a non-exercise intervention (fall prevention education). After excluding this study, the heterogeneity disappeared, but the result was not statistically significant (MD = −2.08, 95% CI −4.70 to 0.54, P = 0.12, I² = 0%) (Fig. 5b).

The effect of T OT on fall incidence

The two RCTs reported fall incidence in older adults, with one finding no statistically significant differences between the groups at any time point after the intervention [51], and the other also reporting no significant differences during the 12-month follow-up period [46].

Risk of publication bias

We assessed potential publication bias in our meta-analysis of the TUG test and BBS. Egger’s test was conducted, indicating the absence of publication bias. The P-values for the TUG test (0.782) and BBS (0.200) indicated no significant publication bias. However, due to the limited number of studies included in the other meta-analyses, publication bias assessment was not feasible.

Discussion

This study demonstrated that TOT was more effective than conventional rehabilitation methods (e.g., conventional gait and balance training, traditional fall prevention exercises, health education, etc.) in improving balance, gait function, and balance confidence related to falls among older adults. The effect of TOT on reducing fear of falling was no longer significant after sensitivity analysis, indicating that its benefit is not robust. However, the limited number of studies precluded definitive conclusions on FGA improvement or fall incidence reduction, warranting further investigation in higher-quality studies.

TOT was more effective than conventional rehabilitation training in improving balance among older adults. This finding aligns with previous studies on stroke patients [24, 28] and children with cerebral palsy [18]. The improvement in the BBS reached the MCID threshold, which suggests that TOT may have clinical significance for balance. The superior efficacy of TOT in balance may be attributed to TOT’s core mechanism of inducing neuroplasticity through task-specific practice, which efficiently optimizes postural control strategies [18, 21–24]. Interestingly, although the improvement in the TUG test did not reach the MCID threshold, subgroup analysis by age revealed that the improvement in adults aged ≥ 75 years reached the MCID threshold. This suggests that TOT may have higher clinical value in enhancing mobility and functional balance in older adults over 75 years of age. This could stem from the more pronounced functional decline observed in the advanced age group [65, 66]. Furthermore, subgroup analyses for the TUG test indicated that the type of control group may influence the effect of the intervention. However, this finding should be interpreted with caution, as the non-exercise control subgroup included only one study (Sik Oh et al. [27]), limiting the reliability and generalizability of this comparison.

This study demonstrates that TOT significantly improves gait speed in older adults, consistent with previous findings [24, 28, 67], with the improvement reaching MCID thresholds. The improvement in gait function could potentially reduce the risk of falls [68]. Based on current research, the mechanisms by which TOT improves gait speed may involve multiple factors. First, TOT activates higher brain functions, boosting motor cortex and spinal cord adaptability to facilitate neural remodeling, thereby refining motor control for more precise and efficient movements [17]. Second, by emphasizing repetitive functional tasks, TOT reinforces correct movement patterns, enhancing gait coordination and fluidity while reducing energy expenditure, enabling older adults to walk more effortlessly and swiftly [21, 47]. Third, improved balance confidence and better static/dynamic stability collectively increase gait speed [69–72].

TOT is more effective than conventional training in enhancing balance confidence, which can be attributed to its high relevance to daily functional tasks, repetitive reinforcement, and timely feedback [18, 21–24]. By simulating activities of daily living, TOT enables patients to accumulate successful experiences, thereby boosting their self-efficacy [18, 21–24]. For fear of falling, while the primary analysis indicated that TOT significantly reduced FES-I scores, sensitivity analyses revealed that this effect was not robust and was primarily driven by heterogeneity from Sik Oh’s study [27], after the exclusion of which the result became non-significant. Further analysis revealed that the control group in Sik Oh’s study received fall prevention education without any physical training, which may have amplified the intervention effect of TOT. This finding suggests that TOT is likely significantly more effective than no exercise or education-only controls in reducing fear of falling; however, its advantage may be limited when compared to other conventional exercise therapies (such as gait and balance training). Given the limited number of studies currently included, more high-quality, large-sample randomized controlled trials are urgently needed to verify the effect of TOT on fear of falling.

This study could not conclusively determine whether TOT reduces actual fall incidence or improves FGA scores in older adults, possibly due to the limited number of included studies. Moreover, while TOT enhances key fall prevention factors like balance capacity, gait speed, and confidence [68], falls represent a multifactorial geriatric syndrome influenced by various elements, including fall history, muscle strength, psychological factors, and medication effects [68]. Future studies should employ larger samples and comprehensive assessments to better evaluate TOT’s preventive efficacy.

By enhancing balance, gait function, and balance confidence, TOT may represent a more effective intervention than conventional training for reducing fall risk in older adults. Healthcare professionals (e.g., clinicians and physical therapists) should identify and select TOT programs tailored to individual patient needs. Integrating virtual reality or robotic technology with task-oriented training could further enhance patient engagement and effectiveness. To optimize outcomes, TOT protocols should emphasize therapeutic dosage, intensity, and repetition to promote neuroplasticity and functional recovery [67]. Future research should design TOT in greater detail to determine the optimal treatment dosage, intensity, and repetition for different populations. Additionally, follow-up periods should be strengthened to explore the long-term maintenance of TOT effects.

However, this study has several limitations. First, the limited number of included studies made subgroup analyses difficult to interpret, and some outcomes did not yield conclusive results. Second, due to the lack of standardized search terms for TOT, we only used a restricted set of keywords to identify RCTs. Third, since few studies reported follow-up data, we were unable to assess the long-term effects of TOT. Fourth, the literature search was conducted in only a limited number of databases. Finally, while this review focuses on clinical functional outcomes such as balance scales, gait speed, and fall incidence, it does not include biomechanical parameters (e.g., center of pressure). Integrating biomechanical indicators with clinical functional assessments could provide deeper insights into the intrinsic mechanisms underlying the therapeutic effects of TOT.

Conclusion

TOT demonstrates unique advantages compared to conventional training in improving older adults’ balance, gait, and balance confidence related to falls, while also exhibiting potential for reducing fall risk. The improvements in BBS and GS reached the MCID, as did the improvement in the TUG test for older people aged ≥ 75, confirming the TOT’s practical clinical value. However, whether TOT can more effectively reduce the fear of falling, improve FGA scores, and reduce the incidence of falls remains to be confirmed by high-quality research.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: WPZ, LW; methodol ogy: WJZ, JL; database search: WJZ, YJ; study quality assessment: MY, JL; formal analysis and investigation: WPZ, WJZ; writing—original draft preparation: WPZ; writing—review and editing: WJZ, MY; supervision: JT, LW, GZ.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

The data supporting this study are included in the manuscript and supplementary materials.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.