Authors: Shikun Zhang, Guoping Qian, Hong Xu, Meiling Gu, Sujie Mao, Yan’an Wang, Yuqing Zhang, Wensheng Zhou
Categories: Systematic Review, Balance, Gait speed, Berg balance scale test, One-leg stance with eyes open, Timed up and go test
Source: BMC Geriatrics
Authors: Shikun Zhang, Guoping Qian, Hong Xu, Meiling Gu, Sujie Mao, Yan’an Wang, Yuqing Zhang, Wensheng Zhou
Exercise has been proven to significantly enhance balance performance in healthy older adults, particularly those living independently in community settings. However, it remains unclear which specific types of exercise are the most effective and which modality yields the best intervention outcomes. We aimed to evaluate the effects of different exercise modalities on four balance-related outcomes—Berg Balance Scale (BBS), Timed Up and Go Test (TUG), gait speed (GS), and one-leg stance with eyes open (OLSEO)—in healthy older adults, using a systematic review and network meta-analysis.
We searched five Web of Science (WOS), EMBASE, EBSCO, Cochrane Library, and PubMed, with a search cutoff date of March 25, 2024. The search was restricted to randomized controlled trials (RCTs) published in English that investigated the impact of exercise interventions on balance in healthy older adults. Quality was assessed using the PEDro scale, and both traditional and network meta-analyses were performed using Stata 15.1. Surface under the cumulative ranking curve (SUCRA) values were calculated to rank the intervention effects of different exercise modalities.
A total of 93 articles comprising 103 studies with 5,114 participants were included. Overall, the methodological quality of the included studies was relatively high. Pairwise meta-analysis revealed that exercise significantly improved BBS, TUG, OLSEO, and GS (p < 0.05). The network meta-analysis showed that, for improving BBS, augmented reality-based training (AR), aquatic exercise, balance training (BT), comprehensive training (CT), Exergame, functional training, and Otago were significantly more effective than the control group. For improving GS, four exercise modalities—dual-task balance and gait training, balance and gait training, mind-body interventions (MBIs), and CT—were significantly more effective than the control group. For improving OLSEO, dance, Nintendo’s Wii Fit Game (Will Fit), and Otago were significantly more effective than the control group. For improving TUG, AR, BT, CT, MBIs, Will Fit, Otago, and resistance training (RT) were significantly more effective than the control group. According to the SUCRA rankings, Exergame was the most effective for improving BBS (SUCRA = 96.3%); CT was the most effective for improving GS (SUCRA = 85.8%); Otago was the most effective for improving OLSEO (SUCRA = 86.8%); and Will Fit was the most effective for reducing TUG test time (SUCRA = 88.5%).
This study confirms that exercise improves balance in healthy older adults. Traditional exercise like BT, MBIs, CT, and Otago are recommended, while newer approaches like Exergame, AR, and Will Fit can be cautiously considered.
PROSPERO (CRD42024591308).
The online version contains supplementary material available at 10.1186/s12877-025-06212-0.
The global phenomenon of population aging is becoming increasingly significant, with the proportion of older adults expected to rise from 11 to 22% between 2000 and 2050 [1]. As age increases, older adults experience gradual declines in balance performance, muscle strength, and other physical functions [2, 3], with more pronounced effects after the age of 60 [4]. This decline not only impacts daily living abilities but also directly leads to a decrease in balance, significantly increasing the risk of falls among older adults [5]. Among those aged 65 and older, approximately 30–40% experience falls each year, and half of these falls result in injury [6]. Falls among older adults not only increase the burden on social healthcare systems and medical costs [7], but also restrict the mobility [8]. To prevent further falls [9], older adults often reduce their daily activities, which in turn leads to a sedentary lifestyle, further may ultimately heighten the likehood of future falls [10]. Therefore, improving balance performance in older adults is crucial not only for fall prevention but also improves overall quality of life.
The American College of Sports Medicine (ACSM) and the American Heart Association (AHA) recommend that older adults at risk of falling should regularly engage in balance training [11], as improving balance can help reduce the risk of falls [12]. Howe et al. and Sun et al. suggest that regular exercise can enhance muscle strength, improve gait [13, 14], and enhance balance performance in older adults [15]. Hunter et al. found that older adults who engage in low-intensity high-velocity resistance training at least once a week can increase muscle mass and strength, thereby reducing the difficulty of daily activities and work [16]. Naczk et al. further supported this view, showing that after six weeks of resistance training, older adults experienced significant improvements in upper and lower limb strength and balance performance [17]. A meta-analysis indicated that balance in older adults can be effectively improved through various forms of exercise, including resistance training, aerobic exercise, balance training, and activities such as T-Bow and balance board training [3].
Multiple randomized controlled trials and meta-analyses have demonstrated that exercise can improve balance in older adults and reduce the risk of falls.However, their findings have not always been consistent. Some studies showed significant improvements following aerobic or mind-body exercises (p<0.05), while others found no statistically significant effects (p > 0.05) [18–21]. In recent years, several meta-analyses have been to assess the impact of exercise interventions on balance in older adults, each with a different focus [22–25]. Most of these reviews have concentrated on a single training, a narrow subgroup of the population, or only one dimension of balance performance (e.g., reactivity or dynamic balance).
While numerous studies have examined the impact of different exercise interventions on the balance performance of older adults, no network meta-analysis has yet systematically compared the effects of various forms of exercise on balance performance specifically in healthy older adults.Network meta-analysis (NMA) is a statistical method that allows for the comparison of multiple interventions simultaneously by combining both direct and indirect evidence across studies sharing common comparators [26]. Compared to traditional pairwise meta-analysis, NMA enables estimation of the relative effects between all treatment options [27]. Therefore, this study aims to perform a systematic review and network meta-analysis of randomized controlled trials (RCTs) on exercise interventions and their effects on balance performance in healthy older adults, specifically excluding participants with clinical conditions such as Parkinson’s disease, stroke, or musculoskeletal disorders, with the following (1) to comprehensively evaluate the impact of exercise interventions on balance performance in healthy older adults; (2) to analyze the effects of different exercise interventions on various balance performance indicators; and (3) to establish a probability ranking of the effectiveness of different interventions in improving balance performance. The results of this study will provide important evidence for exercise rehabilitation and clinical practice.
This systematic review and network meta-analysis was implemented according to preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines and extension statement for network meta-analyses [28, 29]. In addition, the protocol of this study has been registered with PROSPERO (CRD42024591308).
The literature search strategy was developed based on PRISMA guidelines [30]. Two co-authors (MSJ, XH) formulated the search strategy and independently searched five Web of Science (WOS) Core Collection, EMBASE, EBSCO, Cochrane Library, and PubMed. The search covered the period from the inception of each database until March 25, 2024, and was conducted in English. Disagreements were resolved through discussion. If consensus could not be reached, a third co-author (ZSK) was consulted to make the final decision. The search terms were derived from Medical Subject Headings (MeSH) terms, exercise-related terms, and keywords from relevant literature. The following search terms and their combinations were “aged” “senior” “elder” “old people” “elderly” “old aged” “balance” “balancing” “core stability” “physical performance” “physical function” “Posture Balance” “physical activity” “fitness” “exercise” “sport” “training” and “randomized controlled trial”. To ensure comprehensiveness, the references of included studies and related systematic reviews were manually searched. In Table 1, the WOS database is shown as an example of the specific literature retrieval strategy.
Table 1The searching strategy for WOSSetSearch Query#1TOPIC: Aged OR elder OR aging OR senior OR older adults OR older patients OR older women OR older people OR older persons OR older subjects OR old age OR older age OR old aged OR elder OR elderly OR old people OR geriatric OR late life OR postmenopausal women OR Aged, 65and over Age Groups OR old#2TOPIC: Exercise OR Physical Activity OR Exercise, Physical OR Physical Exercise OR Acute Exercise OR Acute Exercises OR Exercise, Acute OR Exercises, Acute OR Exercise, Isometric OR Isometric Exercises OR Exercise, Aerobic OR Aerobic Exercise OR Exercise Training OR swim OR walk OR dance OR jog OR run OR cycle OR bicycle OR hiking OR tai ji OR tai chi OR yoga OR qigong OR sport OR physical training OR strength training OR weight training OR resistance training OR balance training OR aerobic training OR anaerobic training OR endurance training OR muscle training OR exergame OR active video game OR Wii OR Kinect OR pilates OR motor activity OR cardiac rehabilitation OR fitness OR train OR deep water running OR resistance exercise OR core stability OR core control OR core strength OR sling exercise OR stretching exercise OR pressure biofeedback OR stability ball OR proprioceptive neuromuscular facilitation OR muscle energy technique OR aquatic exercise OR high-intensity interval training OR breathing exercise OR virtual reality exercise OR combined exercise OR yijinjing OR wuqinxi OR Liu Zi Jue OR Baduanjin OR eight section brocade exercise#3TOPIC: Postural Balance OR Posture Equilibrium OR Equilibrium, Posture OR Posture Equilibriums OR Balance, Postural OR Postural Equilibrium OR Equilibrium, Postural OR Posture Balance OR Balance, Posture OR Posture Balances OR Musculoskeletal Equilibrium OR Equilibrium, Musculoskeletal OR Postural Control OR Control, Postural OR Postural Controls OR Posture Control OR Control, Posture OR Posture Controls OR core stability OR balance OR balancing OR TUG OR Timed Up and Go Test OR BBS OR Berg Balance Scale test OR one leg standing with eyes open OR gait speed#4#1 AND #2 AND #3
The inclusion criteria strictly followed the PICOS (population, interventions, comparators, and outcomes) model established by Cochrane for evidence-based medicine [29]. The specific inclusion criteria (1) the minimum age of participants was ≥ 65 years, and they were healthy older adults. There were no restrictions on gender, nationality, or race; (2) the experimental group received single-exercise training or a combination of exercises (including two or more types of exercise), while the control group either maintained their usual daily activities or received a different type of exercise intervention than the experimental group; (3) according to the research of Granacher et al. and Shumway-Cook and Woollacott, balance can be divided into static/dynamic steady-state balance, proactive balance, and reactive balance [31, 32]. Based on these findings, this study included the following Static steady-state balance (one-leg stance with eyes open, OLSEO), Dynamic steady-state balance (gait speed, GS), Proactive balance (Timed Up and Go Test, TUG), and Balance Test Battery (Berg Balance Scale, BBS). Studies had to include at least one of these balance assessment indicators; (4) study only randomized controlled trials (RCTs) were included.
The exclusion criteria (1) participants younger than 65 years old; (2) non-RCT articles; (3) studies in which the experimental or control groups did not involve exercise interventions; (4) studies that did not include the above indicators; (5) targeted participants with clinical diagnoses (e.g., stroke, Parkinson’s disease, osteoarthritis, or other neurological or musculoskeletal disorders).
The relevant data were extracted independently by two co-authors (ZWS, GML) using Excel. In case of disagreements during data extraction, a third co-author (ZSK) was consulted to make a decision. The primary data extracted (1) author first author and year of publication; (2) participant gender, age, and number of participants (male/female); (3) study intervention details (experimental group/control group) and exercise parameters (duration, time, and frequency); (4) outcome BBS, GS, TUG, and OLSEO. Baseline values and post-intervention values were extracted from each study for the relevant indicators, including the means and standard deviations.
In this study, the methodological quality of the included studies was evaluated using the Physiotherapy Evidence Database (PEDro) scale [33, 34]. Research by Morton (2009) and others has demonstrated the reliability and validity of the PEDro scale [35]. The scale consists of 11 items to assess the methodological quality of RCTs, with each item scoring 1 point (the first item is not scored), resulting in a total score of 10 points. A score of 6 or higher is considered indicative of high-quality studies [34, 36]. To ensure the quality of the included studies, two co-authors (MSJ, ZWS) independently assessed the studies. In case of any disagreement, a third author (ZSK) was consulted for evaluation.To evaluate inter-rater reliability, Cohen’s kappa statistic was calculated to assess the level of agreement between reviewers [37].
In this study, the Revised Cochrane Collaboration Risk of Bias Tool for Randomized Trials (RoB2) was adopted to evaluate the quality of included RCT studies [38]. The risk of bias map and the summary of risk of bias were completed through the Robvis tool [39]. Quality evaluation was conducted independently by two co-authors (ZSK and QGP), and any disagreement was resolved through a third author (ZWS).
PMA was used to directly compare the effect sizes between each intervention group and the control group [40]. The analysis was performed using Stata software (version 15.1). The four indicators included in this study were continuous variables. If the measurement methods and units were consistent, the weighted mean difference (WMD) and 95% confidence interval (CI) were used as the effect size indicators. Otherwise, the standardized mean difference (SMD) and 95% CI were used. The heterogeneity between studies was assessed using the Q test and I-squared statistic (I^2^) in the forest plot. p<0.05 was considered statistically significant. The I^2^ statistic was used to measure the degree of heterogeneity, with 0–29%, 30–59%, 60–89%, and > 89% indicating low, moderate, substantial, and high heterogeneity, respectively [41].
NMA relies on four assumptions—connectivity, homogeneity, transitivity, and consistency—to ensure the reliability of the results [42]. Stata software (version 15.1) was used to conduct the NMA, which directly and indirectly compared the effects of the interventions and control groups. A network relationship diagram was used to describe the impact of different types of exercise on balance-related indicators in healthy older adults. The size of the nodes and the thickness of the lines in the diagram were adjusted according to the types of exercise and the sample sizes of the study participants [43]. SMD and 95% CI were used as the effect size indicators. Loop inconsistency (LI) was used to test the inconsistency in the closed loops of the network relationship diagram. If the lower limit of the inconsistency factor (IF) value’s 95% CI was 0 or close to 0, it indicated good consistency between direct and indirect evidence [44]. The surface under the cumulative ranking (SUCRA) in the cumulative probability plot was used to rank the interventions, where 0 ≤ SUCRA ≤ 100%. A SUCRA value of 100% indicated the best intervention, while 0 represented the worst; the larger the value, the better the intervention effect [45]. A funnel plot generated using Stata software was used to assess publication bias in the included studies.
A total of 33,404 studies were obtained by searching five databases. After removing 8,674 duplicates using EndNote20, 24,200 irrelevant studies were excluded based on the titles and abstracts. After reviewing the remaining 530 full-text articles, 440 were excluded for reasons such as not meeting the age criteria, lacking relevant indicators, or not being RCTs. Ultimately, 90 studies were included. By reviewing the references of the included studies, 3 additional studies were identified, resulting in a total of 93 studies being included in the systematic review and network meta-analysis. The study selection process is shown in Fig. 1.
Fig. 1The PRISMA Flowchart for Included RCTs
All included studies were RCTs, with a total of 93 papers. Among them, 83 were two-arm studies, and 10 were three-arm studies, resulting in 103 comparison groups and a total of 5,114 participants, including 3,382 women and 1,169 men. Of these participants, 2,633 were in the experimental group and 2,481 in the control group. Twenty-two studies (24.2%) targeted only women, 58 studies (62.1%) had no gender restrictions, and 13 studies (13.7%) did not report gender information. Among the 93 included papers, 33 studies assessed BBS, 67 studies assessed TUG, 16 studies assessed OLSEO, and 26 studies assessed GSs. The exercise intervention duration ranged from 2 to 96 weeks (with most concentrated around 12 weeks), with intervention frequency ranging from 1 to 7 times per week, and each session lasting 15 to 90 min. The studies included a total of 16 types of exercise (1) comprehensive training (CT), (2) resistance training (RT), (3) Otago exercise, (4) balance training (BT), (5) mind-body interventions (MBIs), (6) exergame, (7) Nintendo’s Wii Fit Game (Will Fit), (8) dual-task balance and gait training (DT-BGT), (9) aerobic exercise (AE), (10) dance, 11) balance and gait training (BGT), 12) aquatic exercise (AUE), 13) augmented reality-based training (AR), 14) virtual reality-based training (VR), 15) flexibility, and 16) functional training (FT). Details can be found in Appendix 1.
The PEDro scores for the included studies are shown in Appendix2. The total methodological quality scores ranged from 4 to 9 points, with an average PEDro scores of approximately 6.57/10. Specifically, there was 1 paper with 4 points, 9 papers with 5 points, 35 papers with 6 points, 33 papers with 7 points, 14 papers with 8 points, and 1 paper with 9 points. Overall, the methodological quality of the included papers was high. The inter-rater agreement for methodological quality assessment was evaluated using Cohen’s kappa (κ = 0.73), indicating substantial agreement between the two reviewers.
Among 93 included papers, 4 mentioned that the overall assessment was carried out at “high risk;” 72 had “some concerns,” and 17 were at “low risk”. Overall, most studies were rated as having low risk or some concerns. Detailed Risk of bias domains and summary plots are presented in Appendix 4.
Thirty-three papers (35 comparison groups) included BBS, consisting of 32 two-arm studies and 1 three-arm study, involving 16 types of exercise interventions. Meta-analysis results showed that exercise interventions significantly improved BBS [SMD = 0.83,95% CI (0.52;1.14)]. Specifically, BT, Otago, Will Fit, and CT showed significantly better improvements in BBS compared to the control group (p < 0.05). See Table 2 for details.
Table 2Meta-analysis results for the BBS outcome measureInterventionNumber of StudiesSMD (95% CI)I² ValueZ Valuep-valueIncluded in Pooled Meta-analysisBT VS. CON51.65(0.80, 2.49)77.3%3.830YesOtago VS. CON20.85(0.46, 1.25)04.210YesDance VS. CON11.05(0.36, 1.74)-2.980.003No(single study)MBIs VS. CON3−0.07(−0.76, 0.61)80.1%0.210.836YesWill Fit VS. CON20.56(0.19, 0.93)13.4%2.980.003YesFT VS. CON10.44(−0.12, 1.00)-1.550.121No(single study)CT VS. CON31.07(0.30, 1.84)72.4%2.740.006YesRT VS. CON10.03(−0.56, 0.62)-0.090.93No(single study)AUE VS. CON11.33(0.74, 1.91)-4.480No(single study)Overall190.83(0.52, 1.14)78.6%5.220-Only interventions supported by ≥ 2 studies were included in the pooled meta-analysis. Comparisons based on a single study are presented for descriptive purposes only
Twenty-six papers (30 comparison groups) included GS, comprising 24 two-arm studies and 2 three-arm studies, with 13 types of exercise interventions. Traditional meta-analysis results indicated that exercise interventions significantly improved GS [SMD = 0.63,95% CI (0.4;0.87)]. Among them, BT, CT, and DT_BGT significantly improved GS compared to the control group (p < 0.05). See Table 3 for details.
Table 3Meta-analysis results for the GS outcome measureInterventionNumber of StudiesSMD (95% CI)I² ValueZ Valuep-valueIncluded in Pooled Meta-analysisBT VS. CON30.46(0.06, 0.85)02.260.024YesOtago VS. CON20.28(−0.06, 1.22)60.9%0.580.56YesCT VS. CON51.19(0.7, 1.67)60.9%4.80YesDT_BGT VS CON20.33(0.04, 0.62)02.20.028YesBGT VS CON10.72(0.12, 1.32)-2.370.018No(single study)Dance VS. CON1−0.27(−1.11, 0.58)-0.620.535No(single study)RT VS. CON20.43(−0.35, 1.31)48.7%0.960.3337YesMBIs VS. CON10.84(0.32, 1.37)-3.150.002No(single study)Overall170.63(0.4, 0.87)53.8%5.290-See Table 2
Sixteen papers (18 comparison groups) included OLSEO, consisting of 15 two-arm studies and 1 three-arm study, involving 11 types of exercise interventions. Meta-analysis results showed that exercise significantly improved OLSEO compared to the control group [SMD = 0.67,95% CI (0.31;1.03)]. See Table 4 for details.
Table 4Meta-analysis results for the OLSEO outcome measureInterventionNumber of StudiesSMD (95% CI)I² ValueZ Valuep-valueIncluded in Pooled Meta-analysisRT VS. CON20.07(−0376, 0.52)6.6%0.320.75YesCT VS. CON40.44(−0.06, 0.95)64.9%1.710.087YesOtago VS. CON11.31(0.7, 1.92)-4.210No(single study)MBIs VS. CON10.78(0.05, 1.51)-2.080.037No(single study)Will Fit VS CON11.33(0.76, 1.91)-4.540No(single study)Dance VS. CON11.02(−0.02, 2.06)81.4%1.910.056No(single study)Overall100.67(0.31, 1.03)73.3%3.640-See Table 2
Sixty-seven papers (87 comparison groups) included TUG, comprising 57 two-arm studies and 10 three-arm studies, involving 16 types of exercise interventions. Meta-analysis results showed that exercise significantly improved TUG [SMD=−0.62,95% CI (−0.82; −0.42)]. Specifically, BT, Otago, dance, Will Fit, CT, and AE demonstrated significantly better effects on TUG compared to the control group (p < 0.05). See Table 5 for details.
Table 5Meta-analysis results for the TUG outcome measureInterventionNumber of StudiesSMD (95% CI)I² ValueZ Valuep-valueIncluded in Pooled Meta-analysisCT VS. CON12−0.69(−0.99, −0.40)61.4%4.60YesRT VS. CON5−0.54(−1.17, 0.08)74.7%1.70.089YesBT VS. CON6−1.36(−2.17, −0.54)85.8%3.270.001YesOtago VS. CON4−0.87(−1.38, −0.36)54.6%3.340.001YesBGT VS. CON3−0.24(−0.91, 0.43)69%0.710.476YesDT_BGT VS. CON6−0.13(−1.03, 0.77)94.2%0.290.773YesVR VS. CON1−0.78(−0.62, 0.05)-1.830.067No(single study)Will Fit VS. CON3−0.83(−1.47, −0.19)70.2%2.550.011YesMBIs VS. CON3−0.09(−0.41, 0.24)00.530.599YesExergame VS. CON3−0.09(−0.48, 0.29)38.8%0.480.631YesDance VS. CON2−1.14(−1.56, −0.73)05.390YesAE VS. CON1−0.84(−1.57, −0.10)-2.230.026No(single study)Overall49−0.62(−0.82, −0.42)80.8%6.070-See Table 2
Figures 2a, b, c, and d represent the network diagrams for BBS, GS, OLSEO, and TUG, respectively. In these diagrams, the size of the circles represents the sample size, and the thickness of the lines between circles represents the number of direct comparison studies between two interventions.Fig. 2Network Diagrams for the BBS, GS, OLSEO, and TUG Outcome Measures Across Different Exercise modalities.The sizes of the nodes in the network diagram represent the total number of participants in each intervention, and the thicknesses of the connecting lines between two points represent the number of direct comparisons between the two interventions being connected.
The loop inconsistency test for the included studies generated 8 closed loops, with IF values ranging from 0.04 to 1.54, and the lower limit of the 95% CI being 0, indicating no significant inconsistency across the closed loops. The network meta-analysis results showed that compared to the control group, four types of exercise effectively improved BBS (p < 0.05): AR [SMD = 1.06,95% CI (0.01;2.11)], BT [SMD = 1.18,95% CI (0.58;1.78)], Exergame [SMD = 2.07,95% CI (0.80;3.34)], and Otago [SMD = 1.07,95% CI (0.30;1.85)]. Comparing different exercise interventions, AR, AUE, BT, CT, Dance, Exergame, and FT showed significant differences compared to Flexibility (p < 0.05); BT and Exergame were significantly different from MBIs (p < 0.05); BT and Exergame also showed significant differences compared to RT (p < 0.05); Exergame was significantly different from Will Fit (p < 0.05), while no significant differences were found among the rest (p > 0.05). Based on SUCRA rankings, the top three exercises for improving BBS were Exergame (SUCRA = 95.7%), BT (SUCRA = 78.6%), and Otago (SUCRA = 73.3%). The rankings for other exercises are shown in Table 6.
The loop inconsistency test produced 6 closed loops, with IF values ranging from 0.05 to 5, and the lower limit of the 95% CI being 0, indicating no significant inconsistency across the closed loops. The network meta-analysis showed that compared to the control group, four types of exercise significantly improved GS (p < 0.05): DT_BGT [SMD = 0.45,95% CI (0.06;0.83)], BGT [SMD = 0.81,95% CI (0.37;1.26)], MBIs [SMD = 0.92,95% CI (0.25;1.59)], and CT [SMD = 0.96,95% CI (0.65;1.27)]. Pairwise comparisons revealed that CT was significantly different from DT_BGT, Dance, FT, and Flexibility (p < 0.05), while no significant differences were found between the other interventions (p > 0.05). According to SUCRA rankings, the top three exercises for improving GS were CT (SUCRA = 85.8%), MBIs (SUCRA = 79.4%), and AR (SUCRA = 76.1%), with the remaining rankings shown in Table 6.
The loop inconsistency test generated 1 closed loop with an IF value of 0.11, and a 95% CI of (0;0.78), indicating no significant inconsistency across the closed loops. The network meta-analysis showed that compared to the control group, three types of exercise significantly improved OLSEO (p < 0.05): Dance [SMD = 0.93,95% CI (0.25;1.61)], Will Fit [SMD = 1.25,95% CI (0.33;2.16)], and Otago [SMD = 1.36,95% CI (0.41;2.31)]. Pairwise comparisons revealed significant differences between Will Fit, Otago, and RT (p < 0.05), while no significant differences were found among the other interventions (p > 0.05). Based on SUCRA rankings, the top three exercises for improving OLSEO were Otago (SUCRA = 86.8%), Will Fit (SUCRA = 84.2%), and Dance (SUCRA = 74.1%). The rankings for other exercises are shown in Table 6.
The loop inconsistency test produced 39 closed loops, with IF values ranging from 0.01 to 3.28, and the lower limit of the 95% CI being 0, indicating no significant inconsistency across the closed loops. The network meta-analysis showed that compared to the control group, eight types of exercise effectively improved TUG (p < 0.05): AR [SMD=−1.64,95% CI (−3.11;−0.17)], AUE [SMD=−1.45,95% CI (−2.88;−0.03)], BT [SMD=−0.54,95% CI (−0.95;−0.13)], CT [SMD=−0.62,95% CI (−0.92;−0.33)], MBIs [SMD=−0.77,95% CI (−1.23;−0.31)], Will Fit [SMD=−1.57,95% CI (−2.21;−0.92)], Otago [SMD=−0.96,95% CI (−1.53;−0.39)], and RT [SMD=−0.58,95% CI (−0.99;−0.16)]. Pairwise comparisons showed significant differences between AR, AUE, BT, and CT compared to Flexibility, and Will Fit compared to RT and VR, while no significant differences were found between other interventions (p > 0.05). Based on SUCRA rankings, the top three exercises for improving TUG were AE (SUCRA = 91.9%), Flexibility (SUCRA = 90.2%), and BGT (SUCRA = 75.6%), with the remaining rankings shown in Table 6.
Table 6Probability rankings of different exercise interventionsRankExercise Type (SUCRA)BBSGSOLSEOTUG1Exergame (95.7)CT (85.8)Otago (86.8)AE (91.2)2BT (78.6)MBIs (79.4)Will Fit (84.2)Flexibility (90.2)3Otago (73.3)AR (76.1)Dance (74.1)CON (85.4)4AUE (72.6)BGT (75.6)VR (72.2)BGT (75.6)5AR (71.4)VR (64.8)BT (59.4)Exergame (73.5)6FT (5.93)AE (58.3)CT (47.4)DT_BGT (68.0)7VR (55.4)RT (50.6)MBIs (44.6)VR (50.3)8CT (50.0)Otago (48.7)RT (37.3)BT (49.9)9BGT (49.1)DT_BGT (47.5)BGT (29.6)Dance (48.3)10DT_BGT (47.5)BT (39.1)CON (25.2)RT (47.7)11Will Fit (47.3)FT (24.1)DT_BGT (23.4)CT (44.0)12Dance (44.4)Dance (23.6)Flexibility (15.8)MBIs (34.9)13RT (34.5)CON (16.0)-Otago (26.5)14AE (30.6)Flexibility (10.3)-FT (25.0)15MBIs (20.6)--AUE (17.2)16CON (17.4)--AR (13.3)17Flexibility (2.2)--Will Fit (8.9)
We conducted publication bias assessments for the four outcome BBS, GS, OLSEO, and TUG. The results showed that the funnel plots for these measures were generally symmetrical, suggesting a low likelihood of publication bias. However, not all studies were symmetrically distributed on both sides of the vertical line, indicating a potential risk of publication bias. Therefore, the results of this study should be interpreted with caution. See Fig. 3 for details.Fig. 3Funnel plot of publication bias
This study performed a systematic review and network meta-analysis to compare the effects of different exercise modalities on balance performance in healthy older adults aged 65 and above. A total of 93 studies, including 103 comparison groups and 5,114 participants, were included in the analysis. The PMA results indicated that various exercise modalities could significantly improve the BBS, TUG, GS, and OLSEO balance outcomes in healthy older adults. According to the NMA results, in terms of improving the BBS outcome, Exergame showed the most prominent effect, followed by BT, with Otago ranking third. For TUG improvement, AE was the most effective intervention, followed by Flexibility, with BGT ranking third. In terms of GS improvement, AE performed the best, with VR coming in second and CT ranking third. As for the OLSEO outcome, the interventions ranked in effectiveness as Otago, Will Fit, and Dance.
Berg Balance Scale (BBS) is a commonly used clinical scale to assess balance performance in older adults [46]. The total score ranges from 0 to 56, with higher scores indicating better balance performance [47]. The results of this study showed that exercise interventions significantly improved the BBS scores of healthy older adults [SMD = 0.83, 95% CI (0.52;1.14)], indicating a large effect of exercise on improving BBS. Among the interventions, Exergame was found to be the best exercise form for enhancing functional balance performance. Karahan et al.found that after 6 weeks of Exergame and balance training, both the experimental and control groups showed significant improvements, but the effect in the experimental group was significantly better than in the control group (p < 0.05) [48]. Similarly, Szturm et al. confirmed that Exergame had a significantly better improvement effect compared to dynamic balance training (p < 0.05) [49]. The authors suggested that this could be due to Exergame’s ability to provide immediate feedback, which enhances participants’ motivation and self-efficacy, thereby improving the training effect. However, Duclos et al. and Karahan et al. pointed out that for older adults lacking regular exercise habits, replacing traditional exercises with Exergame might be challenging [48, 50]. This study further validated previous findings, confirming that Will Fit could improve BBS [51], and that augmented reality-based Otago was more effective in improving BBS than traditional Otago [52]. Lesinski, Hortobágyi et al. [23] conducted a meta-analysis which showed that balance training significantly improved BBS in healthy older adults compared to the control group [SMD = 1.52,95% CI (0.65;2.39)]. Additionally, the study by Mohammed et al.found that FT was significantly more effective than RT in improving BBS scores, although both groups showed improvements pre- and post-experiment [53]. This finding contrasts with the results of the present study. The authors believed that RT might help increase lower limb strength, which could indirectly enhance balance performance [54]. Yu et al. also noted that although RT was conducted in the experimental group, no significant changes in BBS scores were observed in either group before or after the test [55].
Gait speed is a key indicator of older adults’ health, reflecting walking stability and the ability to prevent falls. It is crucial for maintaining healthy aging and independent living [56, 57]. GS is related to balance performance and is one of the important factors in fall prevention among older adults [58]. It plays a vital role in body balance control [59, 60]. Age-related declines in physical function often impair gait [61, 62]. The results of this study show that exercise intervention significantly improved GS in healthy older adults [SMD = 0.63,95% CI (0.4;0.87)], indicating that exercise intervention has a moderate effect on improving GS in older adults, with combined training (CT) being the most effective. This finding demonstrates that comprehensive interventions can effectively enhance gait speed, thus improving gait stability and fall prevention abilities. These results are consistent with those of Zou et al., who found that after 8 weeks of mind-body exercise training focused on Tai Chi, the gait speed of older adults in the experimental group was significantly better than that of the control group [63]. A similar effect was observed in Otago training. Yoo et al. reported that Otago training significantly improved the lower limb strength and gait stability of older adults through diverse gait exercises such as backward walking, turning, heel-to-toe walking, and stair walking, thus improving GS [52]. This result further supports the positive effects of various exercise forms on GS as demonstrated in this study. However, Zhang et al. found no significant difference in GS after 23 weeks of moderate-intensity aerobic training in sedentary older adults [64]. The authors speculated that this discrepancy may be due to differences in the type and intensity of exercise programs across studies or that sedentary older adults may need a longer intervention time to show significant gait improvements. Other studies have similarly demonstrated the positive effects of multi-component training on GS. Wang et al. reported significant improvements in GS after 12 weeks of combined training (resistance, endurance, and balance training) [65]. The studies of Bohrer et al. and Cho et al. also showed that multi-component training (resistance, agility, and coordination exercises) can significantly improve functional indicators such as balance, gait speed, and muscle strength in older adults [66, 67].
One-leg standing with eyes open (OLSEO) is commonly used to assess the static stability of older adults’ limbs [68], and this test method has received widespread clinical support [69]. The longer the one-leg standing time, the better the balance performance [70]. The results of this study show that exercise intervention significantly improved the OLSEO score in healthy older adults [SMD = 0.67,95% CI (0.31;1.03)], indicating that exercise intervention has a moderate effect on improving OLSEO, with Otago training being the most effective. This result further supports existing literature. Granacher et al. and Nicholson et al. pointed out that exercise intervention can significantly improve the static balance performance of older adults, and the American College of Sports Medicine has recommended OLSEO as one of the training programs to improve balance in older adults [71, 72]. This study supports previous findings. Lee indicated that gait training based on virtual reality was more effective in improving older adults’ single-leg support ability than traditional gait training (p < 0.05) [73]. Additionally, Yao and Tseng found that 12 weeks of yoga training significantly improved the static balance performance of older adults [74]. Yoga effectively enhances body flexibility and muscle strength, improving stability in the one-leg standing test. Similarly, Kamide et al. also confirmed the effectiveness of comprehensive training (stretching and lower limb strength training) in improving one-leg standing performance [75]. This study further confirmed the advantage of Otago training in improving OLSEO performance in older adults. The Otago Exercise Program is specifically designed for older individuals and includes all the aforementioned training elements, such as leg stretches, toe raises, and turning exercises [52, 76], making it particularly effective in enhancing overall balance performance in older adults. Will Fit involves four basic movements, including two-legged balance, forward and backward leaning, side tilts, hip rotation, and body movements, which effectively strengthen body stability during the game [77]. Dance primarily maintains rhythmic movements to increase flexibility, mobility, and agility in participants [78]. This aligns with the findings of this study, where Will Fit and Dance ranked among the top three in improving OLSEO performance. Additionally, Freiberger et al. found that comprehensive training, including flexibility, balance, strength, and endurance, can significantly improve balance and reduce the risk of falls in older adults [79]. In conclusion, body weight during exercise places a certain load on the lower limb muscles, and this load effectively increases lower limb muscle strength [74].
The American and British Geriatrics Societies recommend the use of the Timed Up and Go (TUG) test to assess gait and balance in older adults [80, 81]. The results of this study show that exercise interventions significantly improved TUG scores in healthy older adults [SMD=−0.62,95% CI (−0.82; −0.42)], indicating a moderate effect. According to the SUCRA rankings, AE demonstrated the greatest improvement in TUG scores, followed by flexibility training, with the BGT also showing relatively higher effectiveness.
Bhuyan and Kumar found that both the experimental group, which received Swiss ball training, and the control group, which received strength training, showed significant improvements pre- and post-intervention, but there were no significant differences between the groups [82]. Kyrdalen et al. further supported the effectiveness of Otago training in improving TUG performance, with significant improvements observed in the experimental group after 12 weeks [83]. However, Binns and Taylor found inconsistent results, where after six months of Otago training, the experimental group did not show significant differences in the TUG test compared to the control group [84]. The authors suggested that this could be due to the high level of physical activity maintained by participants in their daily lives, which might have diminished the intervention’s effect. Similar findings were reported in Beling and Roller’s study, where no significant differences were found between the experimental group that underwent 12 weeks of balance training and the control group, or even within the experimental group before and after the intervention (p > 0.05) [85].
Although some individual studies, such as Campelo et al., found significant improvements in TUG performance after 6 weeks of Will Fit training (p < 0.05) [86], our network meta-analysis ranks Will Fit as the lowest among all evaluated interventions for TUG (SUCRA = 8.9). A systematic review suggested that compared to traditional exercises, exergames or Will Fit are more competitive and interactive, which can motivate older adults to engage more actively in the exercises and thus enhance their effectiveness [87]. However, it is noteworthy that only Exergames achieved a moderate SUCRA score (73.5, ranked 5th) in our analysis. These findings indicate that despite their motivational appeal, current pooled evidence indicates their overall impact on TUG performance remains limited. As Campelo et al. recommend generalizing these findings cautiously to the entire older population, and future research should further explore the applicability of such interventions [86].
However, this study has several (1) in terms of included literature, while we aimed to incorporate all relevant studies, the limitations of literature screening and databases may have resulted in some studies being overlooked; (2) regarding research indicators, there are numerous testing methods for balance function, and this study only selected four balance assessment metrics, which may introduce selection bias; (3) concerning the forms of intervention, the categorization of exercises was based on existing literature, but certain subjective judgments may have introduced bias; (4) with respect to the study subjects, some studies did not clearly specify the health status of older adults, which may affect the consistency of the results;5) this study was limited to English-language studies, may exclude high-quality research published in other languages. Future research should implement stricter controls over intervention design and sample selection criteria to enhance the reliability of the findings. Additionally, some methodological limitations inherent to NMA should be acknowledged. NMA is based on the assumption of transitivity, which requires sufficient similarity across studies in terms of populations, interventions, and outcomes. If this assumption is violated, the validity of indirect comparisons may be compromised.
This study, through systematic evaluation and network meta-analysis, further demonstrates the significant effects of exercise interventions on balance abilities (BBS, TUG, GS, and OLSEO) in healthy older adults. The results indicate that different exercise modalities and assessment indicators yield varying effects on balance improvement.
For BBS (balance test battery), Exergames performed the best (SUCRA = 95.7%), 30 to 45 min three times a week is recommended as a core program for improving balance and fall prevention, combined with virtual reality-based tasks (e.g., balance games from Wii Fit) with conventional balance training exercises (such as single-leg stance and turning drills); For GS scores (dynamic steady-state balance), CT was the most effective(SUCRA = 85.8%), it should be include multimodal exercises such as resistance training, balance and endurance for at least 30–45 min (3–5 times per week); For OLSEO (static steady-state balance), the Otago Exercise Program was the most effective(SUCRA = 85.8%), 10–15 min of lower limb strengthening and stability training is recommended daily; For TUG test, Will Fit performed best(SUCRA = 8.9%), and it is recommended to use moderate intensity (50–70% of heart rate reserve), 2–3 times a week, 20–30 min each training session.
In summary, it is advisable to use traditional training methods (such as BT, MBIs, CT, and Otago) combined with emerging exercise methods (like exergames, AR, and Will Fit)to enhance balance abilities in healthy older adults.While emerging exercise methods may demonstrate high efficacy, their implementation in clinical or community settings can be constrained by limited access to technology, equipment, or digital literacy. In contrast, traditional programs are cost-effective, require minimal resources, and are broadly applicable across various settings Future research should investigate the long-term effects of different exercise modalities and optimal combination strategies to create more personalized balance enhancement programs for older adults.
Supplementary Material 1.
Supplementary Material 2.
Supplementary Material 3.
Supplementary Material 4.
Supplementary Material 5.