We performed a systematic literature search in six bibliographic databases, i.e., PubMed, Pedro, Cinahl, SportDiscus, and Scopus. In addition, we performed a literature search on Google Scholar. The search strategy included only terms related to or describing the intervention. Terms were combined with the Cochrane MEDLINE filter for controlled trials of interventions. Our search syntax was: (“motor-cognitive intervention” OR “dual-task” OR “motor-cognitive training” OR “physical-cognitive intervention” OR “motor-cognitive exercise” OR “exergames” OR “serious game” OR “active video game”) AND (“gait” OR “walk” OR “walking” OR “mobility” OR “balance” OR “posture”) AND (“elderly” OR “old” OR “older” OR “older adult” OR “older adults” OR “aging” OR “elder adults” OR “elders” OR “old-olds”). Search terms were adapted for use with other bibliographic databases in combination with database-specific filters for controlled trials, where these are available. Database searches were supplemented by the review of the authors files. We additionally reviewed the reference list of each included article. We included only studies published in English. When searching for articles, we did not set a time frame for publication. Nevertheless, all articles that met our inclusion criteria were published between 2009 and 2021. The searches were performed again just before the final analysis and additional studies were selected for inclusion. Titles and abstracts that did not meet the inclusion criteria were excluded from the list. The remaining full texts were screened by 3 reviewers (UM, KT, and LŠ). Ultimately, only randomized controlled trials that met the listed inclusion criteria were included.

The strategy for the literature search followed the PRISMA (The Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. This is an evidence-based minimum set of items for reporting in systematic reviews and meta-analyses. PRISMA is used as the basis for reporting systematic reviews with objectives. In addition, we included the Problem/Population, Intervention, Comparison, and Outcome (PICO) framework; it can help formulate the search strategy with set of key questions to efficiently find high- quality evidence:

We included only randomized controlled trials (RCTs). The inclusion criteria were: (1) the type of outcome measure was gait speed under a single- and/or a dual-task, the Up and Go test and/or the Berg Balance Scale, (2) subjects who performed motor-cognitive interventions were compared with those who performed only motor interventions or subjects who were in the passive control group; (3) the mean age of subjects was greater than 60 years (4) studies in which the effect of interventions was of interest if data were available.

Initially, we included subjects classified as “healthy” older adults with no specific diseases diagnosed and individuals with different diseases such as Parkinson's disease, balance impairment, mild cognitive impairment, osteoporosis, dementia, diabetes mellitus, Alzheimer's disease, and studies, that included either patients after stroke or hospitalized patients or patients with a history of falls, or osteoarthritic patients with balance impairment or older adults who were classified as frail or adults with various motor and cognitive deficits or residents of long-term-care facilities or patients with severe neurocognitive disorders. Because we found a high degree of heterogeneity within the groups diagnosed with a particular deficit, we decided to exclude from further analysis all studies that included diseased individuals. However, we have left a summary of all studies included in the original analysis in Supplementary Tables 1–4.

Three independent authors (KT, UM, and LŠ) conducted the search for available studies on a selected topic. The screening was performed in four steps. First, the titles were screened by the reviewers to determine whether they were suitable for our meta-analysis. Then, abstracts were assessed to determine whether the study topic met the selected inclusion and exclusion criteria. The inclusion criteria were selected as follows (as mentioned above) and are described in Table 1: the mean age of the participants was 60 years or more, the type of intervention was a motor-cognitive intervention, the comparison group was either passive or with included motor intervention, the outcomes of the studies were gait speed under a single- or a dual-task condition, TUG and BBS test and the study design was a RCT. The exclusion criteria were: no control group, irrelevant outcomes, unsuitable measurement of gait (e.g., measures were performed on a treadmill), inadequate results and unsuitable measurement of balance (e.g., measures were performed on a force plate). Third, the full text articles were read, the required information was selected and included (if appropriate) in the meta-analysis. Finally, the references of the included studies were reviewed for possible inclusion. If the full text of any paper was unavailable or the data of the study were incomprehensible (certain data on results were missing, e.g., standard deviation, or we were unable to deduce the value of the results from the reported data, e.g., data were reported in graphs), the corresponding author was contacted by email. Disagreements about the inclusion/exclusion of certain RCT were resolved by discussion or by a third person when no consensus could be reached (EdB, AP).

The Physiotherapy Evidence Database (PEDro) scale was used to assess the risk of bias and quality of included studies (Maher et al., 2003). This scale helps the reader to quickly assess whether a clinical trial presents reliable and meaningful results for use in clinical practice. Points are awarded only when a criterion is clearly met. In addition, points are awarded according to the specifics of the article and if the article meets those specifics (eligibility criteria were specified, subjects were randomly assigned to groups, assignment was concealed, the groups were similar at baseline regarding the most important prognostic indicators, there was blinding of all subjects, there was blinding of all therapist who administered the therapy, there was blinding of all assessors who measured at least one key outcome, measures of at least one key outcome were obtained from more than 85% of the subjects initially allocated to groups, all subjects from whom outcome measures were available received the treatment or control condition as allocated or, where this was not the case, data for at least one key outcome was analyzed by “intention to treat”, the results of between group statistical comparisons are reported for at least one key outcome). The quality assessment score was interpreted as follows: studies scoring 6–10 points on PEDro quality assessment were of “high quality”, studies scoring 4–5 of “fair quality” and studies scoring 0–3 of “poor quality”. The evaluation of the studies is available in Table 2.

All included studies were divided into three groups (Tables 3–5) according to the type of dual-task intervention they performed. According to Herold's definition (Herold et al., 2018), we divided the studies into those that can be performed simultaneously or sequentially. Furthermore, studies that included training with a dual-task performed simultaneously were further divided into studies that performed simultaneous motor-cognitive training with additional cognitive task or those who performed simultaneous motor-cognitive training with incorporated cognitive task. In the category of simultaneous motor-cognitive training with an incorporated cognitive task, there were very few studies that did not use exergaming. We excluded the studies that did not meet the definition of exergaming and formed a group with exergaming studies only. The third group is therefore called exergaming.

The meta-analyses were performed using Comprehensive Meta-analysis software (version 3.0; Biostat Inc., Englewood, NJ, USA). The mean differences and 95% confidence intervals (Cls) were calculated for the included studies. We applied the random-effects model of the meta-analysis in all comparisons to determine the effect of the motor-cognitive intervention on gait and balance. Due to the high heterogeneity of the measured variables, the effect sizes were reported in Hedges-‘g. To calculate each effect size we used reported mean value of selected parameter, their standard deviation and sample size of the included study. The following established criteria were used to interpret the magnitude of motor-cognitive intervention for gait and balance improvements: trivial (<0.20), small (0.21–0.60), moderate (0.61–1.20), large (1.21–2.00), very large (2.01–4.00) and extremely large (>4.00) changes (Hopkins et al., 2009; Fraser et al., 2017; Wongcharoen et al., 2017; Laatar et al., 2018). Heterogeneity across studies was assessed using the I² statistics, which is a measure of inconsistency used to quantify between-study variability. A value of 25% is recommended to represent low statistical heterogeneity, 50% moderate and 75% high statistical heterogeneity (Higgins, 2003). In addition, the sensitivity analysis excluded studies with poor methodological quality, i.e., the study's PEDro score was 3 or less. The publication bias was assessed by examining the asymmetry of the funnel plots using Egger's test (Egger et al., 1997; Sterne et al., 2011). Significant publication bias was considered if the p - value was < 0.10.

The initial search yielded 6,314 results. After duplicates were removed, 2,546 articles remained to be considered. After screening titles and abstracts, 890 records were excluded. Full-text reading of 351 articles revealed that 262 articles did not meet our inclusion criteria. We excluded 262 articles with the following reasons: no matched control group in the study (n = 44), unsuitable protocol (e.g., only intervention group, without control group) (n = 20), irrelevant outcomes (n = 32), unsuitable motor-cognitive interventions (e.g., additionally added cognitive tasks during interventions, incorporated cognitive task that was not part of exergame intervention) (n = 81), unsuitable measurement of gait (e.g., measures were performed on a treadmill) (n = 11), inadequate outcomes (n = 20), and unsuitable measurement of balance (e.g., measures were performed on a force plate) (n = 6) and not randomized controlled trials (n = 48). After excluding studies that included diseased older adults, there were 58 studies that we included in the quantitative synthesis. Details of the study selection process are presented in Supplementary Figure 1.

Based on the quality assessment, 44 out of 58 studies were high quality, 11 of fair quality and only 3 of low quality (Table 2). The intervention characteristics, including the type of intervention, a description of the motor and cognitive components and the frequency and dose of training described in the included studies are summarized in Tables 3–5. To facilitate the review of included studies, we have subdivided all studies according to the type of training performed by the experimental group (sequential motor-cognitive training, simultaneous motor cognitive training with additional cognitive task and exergaming).

We included 7 studies with sequential motor-cognitive training (presented in Table 3). These studies included samples ranging from 14 to 147 participants (range of age 69.7–81.9 years). The most common training approach (N = 6) was a combination of aerobic/resistance/strength/balance training and performing cognitive tasks on the PC.

There were 19 studies with simultaneous motor-cognitive training with an additional cognitive task (presented in Table 4). Studies included samples ranging from 17 to 286 participants, with the age of participants in both groups (experimental and control group) varying from 64.7 to 85.3 years. The most common training approach used was a combination of balance training while simultaneously performing different cognitive tasks (the number of studies with that type of training is 9).

We included 33 studies that performed exergaming (Table 5). Studies included samples ranging from 9 to 153 participants. The age range of participants in both groups (experimental and control) was between 65.2 and 86.9 years. The training approach in the studies included different types of interventions using PC and consoles with game controls, e.g., Nintendo Wii and Xbox Kinect, as one form of exergaming training offered to the participants.

Our meta-analysis showed that motor-cognitive intervention can improve gait speed under single-task conditions. There was a small but significant effect that suggests that motor-cognitive interventions for this specific gait parameter may be beneficial for healthy older adults. However, further analysis showed that such interventions are no more effective than other conventional interventions in improving gait speed under a single-task conditions and that the type of motor-cognitive intervention is not a moderating factor for a positive effect.

Although walking was considered a fairly simple task until recently, it requires a large amount of higher-level cognitive input (Mirelman et al., 2018). For this reason, we hypothesized that motor-cognitive interventions would be more beneficial compared to motor training alone. The fact that motor-cognitive training does not contribute more to gait improvement in the active comparison groups than conventional forms of motor training alone is an important point to discuss. The finding is in line with one other systematic review (Gavelin et al., 2021) showing that the addition of cognitive training to physical exercise does not reduce physical efficacy of the training, and exergaming was only superior to passive control for both physical and cognitive outcomes. However, this review also found that motor-cognitive training is likely to be most effective for cognition.

The effect size of the overall improvement in gait speed under a single-task conditions was small, with high statistical heterogeneity between studies. Because of the great heterogeneity in the methods and measurements of the studies, more studies within each subgroup would be needed to draw definitive conclusions. When planning future motor-cognitive interventions, overall effectiveness is important and a deeper understanding of causation is needed (e.g., type and design of intervention, quality of research conducted). There were 27 studies in which motor-cognitive intervention had a positive effect and 14 studies in which motor-cognitive intervention had a negative effect. Among the studies with the positive effect, the highest effect of intervention had the study by Sadeghi et al. (2021) (quality score = 6/11, see Table 2). The experimental design of this study involved a motor-cognitive intervention with visual context displayed on a PC, a confirmed human-computer interaction with tasks performed dynamically. The effect of an intervention study conducted by Jardim et al. (2021) (quality score = 6/11), in which participants performed aerobic, resistance, and stretching training while simultaneously solving cognitive tasks, was similarly high. The third largest effect was in the study by Pothier et al. (2018) (quality score = 4/11), where participants in addition to the aerobic and resistance training, performed cognitive training on a PC. Taken together, analysis of studies with larger effect sizes did not identify any pure trends that could currently provide an answer to the most effective designs of motor-cognitive interventions.

Finally, an additional analysis was performed excluding the study by Eggenberger et al. (2016) due to poor quality (PEDro score ≤ 3). The exclusion of the study did not affect the final conclusion of the results, as the additional calculations only confirmed the results reported above.

We included twenty studies in the meta-analysis to evaluate the effect of a motor-cognitive intervention on gait speed under dual-task conditions, which yielded a non-significant effect. Regarding the quality (assessed by the PEDro scale) of the included studies, 6 studies were of “good quality” and 5 studies were of “fair quality”. Heterogeneity was moderate (82%), and the dual-task assessment methods varied considerably. When interpreting our non-significant results on gait speed under dual-task conditions, it should also be considered that dual-task walking used different cognitive tasks and were combined into one effect size (e.g., walking with n-back task, verbal fluency task, backward couniting, Go/No Go task). Therefore, future studies should investigate this effect considering different subcategories of cognition as a secondary task.

We found that motor-cognitive intervention has a small positive effect on TUG performance for healthy individuals. Moreover, our analysis showed that motor-cognitive intervention is more effective than other conventional interventions in improving the TUG test, but the type of motor-cognitive intervention is not a moderating factor for a positive effect.

The overall effect size was small, with the majority of included studies showing a positive effect of motor-cognitive intervention on TUG test. We included 41 studies, of which 33 had a positive effect size and 8 had a negative effect size. The three interventions related to TUG performance with high effect sizes differed in the type of motor-cognitive intervention as well as in the study quality ratings; one study conducted motor-cognitive training with an additional cognitive task (Jardim et al., 2021; quality score = 6/11). The experimental design included the simultaneous performance of aerobic, resistance, and stretching training, in addition to the performance of various cognitive tasks. Two studies conducted motor-cognitive training with an incorporated cognitive task (Sápi et al., 2019, quality score = 3/11, Kinect balance training; Sadeghi et al., 2021, quality score = 6/11, balance training in virtual reality). On the other hand, the highest negative effect was found by Medeiros et al. (2018) (quality score = 7/11), but it was still only a small negative effect. In the latter study, participants in the experimental group performed a combination of aerobic, flexibility, strength, and balance training while completing a cognitive task. The authors explained the negative effect by the type of the sample (participants had exercised before the intervention) and the relatively short duration of the intervention (12 weeks). Similar to gait speed under a single-task condition, no clear trend for the most effective design of motor-cognitive intervention can be derived for the TUG test.

In addition, the type of control group (passive vs. active) moderated the effects of motor-cognitive interventions, suggesting that motor-cognitive interventions are better able to improve multicomponent tasks of dynamic balance and mobility function as measured by the TUG. Indeed, both the passive and active control groups had a significant effect on the results. The effect of the passive control group was moderate, while the effect of the active control group was smaller, as expected (ES = small), but still statistically significant. Since TUG is a multicomponent test that examines balance, gait speed, and functional ability (Beauchet et al., 2011), it achieves higher ecological validity compared with less complex straight-line walking without an additional task. Motor-cognitive interventions could therefore be a promising strategy to improve dynamic balance and mobility in older adults. When performing a sub-analysis of different types of motor-cognitive interventions (although there was only a non-significant trend with P = 0.064), both the additional and incorporated interventions had a positive effect on TUG, but the sequential intervention did not. One possible explanation is that performing motor and cognitive tasks at different times (sequential motor-cognitive training) may not be as stimulating for improving complex movement tasks as performing these tasks simultaneously. In addition, Prosperini et al. (2021) conducted a meta-analysis in which they found that exergaming interventions can have a positive impact on the balance of people with neurological disorders. This study (Prosperini et al., 2021) was performed on a group of diseased older adults, and we cannot directly confirm our findings with the above-mentioned study, but related results may help us draw meaningful conclusions on this topic. That is, a motor-cognitive intervention with incorporated cognitive task or exergaming has already been shown to have a positive effect on the selected population.

In addition, our results were also confirmed by the exclusion of two studies that were considered to be of poor quality by the PEDro assessment (Bieryla and Dold, 2013; Sápi et al., 2019). Reanalysis confirmed a small but positive effect of motor-cognitive intervention on TUG scores and the effect of the intervention was still moderated by the control group.

The 11 studies included in our meta-analysis showed a positive but small effect on Berg Balance Scale scores (BBS). Thus, motor-cognitive intervention may be beneficial for healthy older adults while improving BBS score, but no differences were found between control groups or type of intervention. All included studies had a positive effect on BBS. The highest effect of the intervention was found in the studies by Karahan et al. (2015) (quality score = 7/11) and Norouzi et al. (2019) (quality score = 8/11). Participants in the experimental group in the study by Karahan et al. (2015) performed exergaming training, where they exercised on the Xbox 360 Kinect. The control group participated in balance training at home. On the other hand, participants in the experimental group in the study by Norouzi et al. (2019) performed resistance training using an isokinetic training device while performing cognitive tasks. Considering that BBS evaluates static and dynamic balance, the possible explanation for the results in favor of the control group could be the implementation of conventional physical therapy, that is motor training. The conventional motor training may have a better effect on balance parameters (BBS) than walking or playing video games. In addition, the existing literature summarizing the effect of simultaneous motor-cognitive training with incorporated cognitive task is inconsistent; Howes et al. (2017) and Pacheco et al. (2020) concluded that exergaming can improve static balance measured with BBS, whereas Chen et al. (2021) did not reach this conclusion.

There are also some limitations as well as future directions that should be mentioned. First, despite a high heterogeneity among motor-cognitive approaches in terms of the selected cognitive or motor task, duration, and frequency, we pooled and summarized the data for the meta-analysis. Second, our meta-analysis included participants with a mean age of 60 years, allowing for the possibility that some individuals in the studies were younger than this age. Third, the results of publication bias indicated the presence of bias in TUG and BBS outcomes. Future studies should focus on other aspects of functional mobility and examine the effectiveness of such motor-cognitive interventions on activities of daily living, such as navigating parks and grocery stores, ability to drive, and others. Finally, the current scarcity of literature on motor-cognitive interventions in specific disease populations may open new avenues of discussion for the implementation of such training. For example, technology-driven exergames with forms of extended reality (XR) that combine real and virtual environments and relate to human-machine interactions generated by computers and wearable technologies will provide services for remote monitoring, training, and telerehabilitation (Meulenberg et al., 2022). The nature of engagement in XR allows training and/or rehabilitation exercises to feel similar to physically performed actions, while seemingly being more engaging, motivating, and stimulating than conventional practice.