The effect of interactive cognitive-motor training in reducing fall risk in older people: a systematic review

Background It is well-known physical exercise programs can reduce falls in older people. Recently, several studies have evaluated interactive cognitive-motor training that combines cognitive and gross motor physical exercise components. The aim of this systematic review was to determine the effects of these interactive cognitive-motor interventions on fall risk in older people. Methods Studies were identified with searches of the PubMed, EMBASE, and Cochrane CENTRAL databases from their inception up to 31 December 2013. Criteria for inclusion were a) at least one treatment arm that contained an interactive cognitive-motor intervention component; b) a minimum age of 60 or a mean age of 65 years; c) reported falls or at least one physical, psychological or cognitive fall risk factor as an outcome measure; d) published in Dutch, English or German. Single case studies and robot-assisted training interventions were excluded. Due to the diversity of populations included, outcome measures and heterogeneity in study designs, no meta-analyses were conducted. Results Thirty-seven studies fulfilled the inclusion criteria. Reporting and methodological quality were often poor and sample sizes were mostly small. One pilot study found balance board training reduced falls and most studies reported training improved physical (e.g. balance and strength) and cognitive (e.g. attention, executive function) measures. Inconsistent results were found for psychological measures related to falls-efficacy. Very few between-group differences were evident when interactive cognitive-motor interventions were compared to traditional training programs. Conclusions The review findings provide preliminary evidence that interactive cognitive-motor interventions can improve physical and cognitive fall risk factors in older people, but that the effect of such interventions on falls has not been definitively demonstrated. Interactive cognitive-motor interventions appear to be of equivalent efficacy in ameliorating fall risk as traditional training programs. However, as most studies have methodological limitations, larger, high-quality trials are needed. Electronic supplementary material The online version of this article (doi:10.1186/1471-2318-14-107) contains supplementary material, which is available to authorized users.


Background
Falls are a major public health problem with one in three older people falling at least once a year [1]. Falling is associated with increased mortality [2], injuries [3], loss of independence [4] and adverse psychosocial consequences [5].
Exercise interventions that aim to improve physical risk factors, such as strength and balance training have been shown to reduce fall rates and fall risk [6,7], fallrelated injuries [8] and fear of falling [9,10] in older people. Systematic review evidence of 44 relevant exercise trials indicates a high exercise dose and challenging balance exercises are important components of successful programs [7]. Presently there is no evidence that cognitive training can lessen fall risk, but there is some evidence suggesting cognitive interventions have a positive impact on cognitive functioning in older populations [11]. The beneficial effects of physical activity decline after exercise cessation [12] and unfortunately low compliance and high drop-out rates in fall prevention studies are often reported [13,14]. Hence, exercise interventions that facilitate adoption and long term adherence may maximize the efficacy of fall prevention strategies.
Interactive cognitive-motor training (ICMT) requires participants to interact with a computer interface via gross motor movements, such as stepping, receiving immediate visual feedback from the projection screen and include high cost Virtual Reality training as well as less complex and inexpensive exergames [15]. It has been reported that ICMT participation is sufficiently intense to induce exercise-related physiological adaptations in older people [16]. In addition, ICMT requires parallel information processing, selective attention to task-relevant stimuli, inhibition of task-irrelevant stimuli and planning/ decision making with respect to the motor execution of the response. These cognitive functions (executive functioning (EF), attention and processing speed) decline with age [17,18] and if impaired increase fall risk [19]. Importantly, ICMT applications require both cognitive and motor involvement and there is evidence that combined training of cognitive and physical functioning leads to better results than isolated cognitive or physical exercises in older people [20][21][22][23].
Because of the potential of ICMT to improve adherence (through the provision of music, direct feedback on performance, positive reinforcement, realistic goal-setting, etc.) and subsequent higher doses of exercise, treatment efficacy may be larger than that achieved with traditionally delivered exercise programs and may lead to sustained improvements. Further, in areas where people have limited access to health care services or where transport is a major barrier for participation, ICMT may provide an effective alternative to enable exercise to be performed at home.
Targeting fall risk factors using ICMT may be effective in reducing falls and improving fall risk factors in older people. Two recent review articles found that exergames are feasible and can improve balance as well as balance confidence in the majority of included studies [24,25]. However, these reviews were either restricted to commercially available off-the-shelf games, included studies with age groups other than 65 years and over and/or were limited to few risk factors for falls. Therefore, the current systematic review aimed to 1) synthesize the currently available evidence on the efficacy of ICMT on falls and intrinsic risk factors for falls in older people and 2) determine how such interventions compare to traditionally delivered interventions in reducing the risk of falling in this group.

Literature search strategy
A two-stage process for the identification of potentially relevant studies was used. First electronic databases (Medline (Pubmed), EMBASE (Ovid), Cochrane CENTRAL) were searched from their inception to 31/12/2013. We combined free-text and MeSH terms using a broad range of synonyms, related terms and variant spelling. Second we scanned all reference lists of review articles and included appropriate trials. The Games for Health Journal and the authors own database were hand-searched for relevant articles. No language restrictions were applied to this initial search. Three semantic search loops were used. The first contained terms related to the study design, the second related to ICMT, the third included key words relating to risk factors for falls and fall outcomes. Finally we limited our search to older populations. The search strategy used for PubMed can be found in Additional file 1.

Inclusion/exclusion criteria
Studies were included if a) at least one treatment arm contained an ICMT component; b) the sample included had a minimum age of 60 years or a mean age of 65; c) at least one physical, psychological or cognitive factor associated with falls or/and fall count data were included as an outcome measure; d) the article was published in Dutch, English or German. In case of multiple publications for one study, all articles were used to obtain maximum information.
Studies were excluded if they were published in abstract form only or designed as a single case study. We also excluded applications in which participants sat while exercising and all robot-based systems, as it was unclear what movements were passive, active or partially supported and therefore different underlying mechanisms may have applied. Finally, studies were excluded if they attempted to change disease-specific outcomes but included if they contained older populations with diseases to investigate fall-related outcomes for which no different underlying mechanisms could be assumed.
Ethical approval was noted for all published papers included in the review. No further ethics approval was sought.

Data extraction and analysis
Two independent reviewers (DS, EdB) scanned titles and abstracts and full texts if necessary to determine eligibility for each article. Any disagreement was solved by discussion. Extracted data were entered into Microsoft Excel/ Word templates specifically developed for this review and piloted using the five first included articles.
The following data were extracted: sample size, population characteristics (age, ethnicity, country, physical function and performance, co-morbidity, falls in previous year), setting (community, hospital, long-term care), ICMT system used, dosage, program of the control group, trial duration, relevant outcomes and assessment instruments, baseline and retest values (between and within group comparisons) and adverse events. Outcome measures of interest included falls as defined by the Prevention of Falls Network Europe [26] and physical, psychological and cognitive measures that have been associated with falls in older people.
Authors were contacted by Email in cases where eligibility could not be established and to clarify any uncertainty regarding intervention content.

Quality assessment of included studies
Risk of bias was assessed by two independent reviewers (EdB, DS) using the Downs and Black scale for randomized and non-randomized trials [27]. This scale contains 27 items assessing reporting (10 items), internal (13 items) and external (3 items) validity and power (1 item). We modified two items: item 23 (randomisation) and item 27 (study power). For item 23, the method used to generate the randomized sequence (as opposed to a simple statement indicating the trial was randomized) was required to meet this criterion for this item as this is standard in the CONSORT statement. For item 27, authors needed to report if and how they determined their sample size a priori (item 27). Disagreements were solved by discussion or by a third person (TV, SL). For studies where one or more of the authors for this review were involved, the bias risk assessment was undertaken by a third person (TV).
Due to the heterogeneity in study designs, outcome measures and populations used we considered conducting a meta-analysis was not appropriate. A descriptive summary of the results was therefore carried out in lieu. The PRISMA-statement was followed for reporting items of this systematic review [28].

Identified studies
The initial search yielded 426 articles. Of these 98 were obtained as full text and 37 studies were identified as eligible for inclusion in this review - Figure 1 shows the flow chart of the selection process.
The included studies could be classified into five categories according to the physical exercise component of the intervention: Step training -dynamic balance programs involving step training using step pads (pressure sensors); this type of training involved rapid or well-timed steps with weight transfers in multiple directions. ii) Balance board training -static and dynamic balance programs using balance boards/platforms; this type of training was characterised by feet in place exercises for most movements and therefore only small movements of the centre of mass. iii) Balance board plus aerobic training -static and dynamic balance plus aerobic training using balance boards and inertial sensors; this type of training involved exercises described under ii) and additional aerobic training (i.e. step aerobics, walking in place). iv) Multi-component programs with low challenge of balance -full body fitness programs using inertial sensors and/or motion capture devices; this type of training usually simulated sports and involved aerobic, resistance, power and agility components with a low balance challenge. v) Aerobic programs -locomotive training using VR displays; this type of training included VR treadmill training and involved continuous rhythmic movements with a low balance challenge.
Exergame step training has also been reported to improve step velocity (reaction time, movement time) [58,61], step accuracy [50,62] and measures of static and dynamic balance [50,55,58,62]. Inconsistent results were found for mobility (timed up and go test) [50,58,59] and balance confidence and falls-efficacy [50,55,58,59,62]. Two studies reported step training did not lead to improvements in pen and paper tests of attention and EF [55,58]. However, several studies have shown step training improves measures of dual tasking [58,59,61,62] and performance in a test that combines stepping and EF [58].
Consistently, studies have shown balance board training can improve performance in balance batteries (e.g. BBS, POMA) between baseline and re-assessment [36,46,47,52,57,63,64]. Some studies have also reported significant between-group differences using passive [49,52] and active [47,63] control groups, whereas others have not -passive control: [57]; active control: [36,46,53,57]. Balance board training has been shown to improve postural sway in the majority of uncontrolled trials after training [40,41,56,66] and when compared to a sham control group [40]. However, balance board training with the IREX Juggler application was found to be ineffective in reducing sway in healthy older people [46], and two studies have reported increases in sway after ICMT [45,53]. Balance board training has been found to improve strength and power measures after training [47,48,52] and when compared to passive [52] and active [47] controls. However, in one study, no between-group difference was found in patients after knee replacement using the Wii balance board as an adjunct to standard rehabilitation [35].
There are inconsistent results for the efficacy of balance board training with respect to falls-efficacy and balance confidence [35,47,52,53,63]. Few balance board interventions have reported on changes in cognitive performance, including tasks under divided attention. Padala et al. found no improvements in global cognition (MMSE) scores after an eight week training program in people with mild Alzheimer's disease [36]. In relation to dual task performance, Yen and colleagues found improvements in sway under divided attention when relying more on vestibular feedback [34], but Kubicki et al. found that the use of a platform as an adjunct to standard strength and balance training did not improve dual task gait speed compared to strength and balance training only in frail older people [45].

iii) Balance board plus aerobic training
Eight studies with a combined sample of 202 participants investigated the effect of combined balance board and aerobic training interventions (ICMT only [37,42,43,51], and ICMT plus other intervention components [31,33,54,65]) ( Table 3). None of these studies reported results for falls and none were powered to do so.
Combined balance board and aerobic training improved static and dynamic balance [31,33,42,43,65] and mobility [51,65] in several studies. However, such training was not effective for these outcomes in a geriatric hospital setting [37] or as an adjunct to mobility training in PD patients [33]. Wii balance board and bicycle training improved depression scores after six weeks training [54], but inconsistent results have been reported for measures of balance confidence and falls-efficacy [37,42,65].
Two studies investigated the impact of combined balance board and aerobic training on cognitive measures. In the study by Maillot et al., 12 weeks of Wii training improved EF and processing speed but not visuo-spatial skills in sedentary older people [51]. In the second by Pompeu et al., PD patients improved their global cognitive function (MOCA) after seven weeks of Wii and traditional mobility exercises but no between-group difference was apparent when Wii training was compared to traditional training of a similar dose [33]. This intervention also did not lead to improvements in dual task performance.

iv) Multi-component programs with low challenge of balance
Four studies involving 134 participants investigated the effect of multi-component interventions (ICMT only [30,38,60], ICMT plus other intervention component [29]) ( Table 4). No intervention reported results for falls and none were powered to do so. A study using the Sony eyetoy in a higher functioning sample of participants with diabetes demonstrated improved functional measures of static and dynamic balance as well as strength [38]. In contrast, two studies in lower functioning residential aged care participants found no improvements in physical outcomes [29,60]. The aforementioned study in diabetic people also showed improvements after training in fallsefficacy [38], and Rosenberg et al. found 12 weeks Wii sports training program improved depression scores and global cognitive functioning [30].

v) Aerobic programs
Two studies with a combined sample of 34 participants investigated the effect of aerobic interventions involving VR treadmill training (ICMT only [32], and ICMT plus other intervention components [39]) ( Table 5). Neither study reported results for falls or fall-related psychological measures, but both showed improvements in balance and mobility [32,39]. In the study by Mirelman et al., VR treadmill training improved EF and showed larger improvements in dual task gait performance than regular treadmill training in people with Parkinson's disease [32].

Effect of interactive cognitive-motor training on falls
The review findings indicate the effect of ICMT on falls is uncertain. Only one of the 37 studies included falls as an outcome measure and due to its modest size (n = 60), this study could be considered to be of a pilot nature for a fall prevention RCT. Encouragingly, the study found a larger reduction of falls in the training group compared to the control group using standing balance training under different sensory conditions [40], as well as improvements in balance and fear of falling; parameters previously reported as mechanisms of effective fall prevention interventions [7,67].

The effect of interactive cognitive-motor training on fall risk parameters
The within-group and passive control group comparisons indicate ICMT can improve balance and strength. The majority of studies placed a strong emphasis on balancethe most important component in effective fall prevention exercise interventions [7]. Clinical test batteries (POMA, BBS) in particular, appeared to be sensitive to change and consistently improved. These test batteries provide combined scores for different functional balance tasks which adds power, reduces measurement error and increases the likelihood of finding valid differences [68]. No studies, however, have reported in which sub-tasks participants improved.
Interestingly, two studies found an increase in sway after feet-in-place training [45,53]. Higher COP velocity and amplitude predict falls [69] which would suggest that the interventions increased fall risk. However, other authors have suggested that an increase in sway after training may relate to improved compensatory strategies [70]. There have also been inconsistent findings regarding intervention effects on one leg stance, functional reach and timed up and go performance. This may be due to the use of off-the-shelves games in many studies. These were not developed to improve clinical outcomes in older people and therefore may lack the task-specificity and/or lack the training principle of progressive overload [71]. The null findings might also be explained by the small sample sizes in many studies and the related low power of detecting significant differences.
It is also possible that psychological consequences of falling can affect quality of life through reduced confidence and activity restriction [72]. Fear of falling and balance confidence improved after training in the review studies that had durations of more than four weeks. However, improvements in falls-efficacy as measured in most trials with versions of the Falls Efficacy Scale (FES, mFES, FES-I, icon-FES), appeared to be not related to the instrument used, the training content or exercise dose. These findings accord with the literature showing that traditional exercise leads to reduced fear of falling in some studies with no clear indication of superiority of one exercise modality [73]. The review findings also indicated ICMT improved depression scores in people both with and without sub-syndromal depression. Depressive symptoms have been consistently associated with falls in older people [74], and exercise is considered an effective strategy for reducing depressive symptoms [75]. However, whether this is due to physiologic, psychological or cognitive factors remains unclear [76].
Cognition, especially EF and attention, are associated with falls in older people [19], and the association between impaired EF and reduced gait speed is one suggested pathway for this association [77]. ICMT improved gait speed and EF in the majority of the review trials, and especially when tasks involving both cognitive and physical components (such as walking under conditions of divided attention) were included; findings were consistent with the literature indicating that cognitive functioning can be enhanced by physical and cognitive exercise [11,78]. It has been suggested that exercise overcomes agerelated overactivity of executive networks in the prefrontal cortex which facilitates motor actions involved in motor planning [79], and that regular physical activity improves efficiency of executive control during more complex tasks involving switching and conflict resolution [80,81]. Thus, improved coordinated motor performance, especially under real-life multitask conditions, could be a possible mechanism for ICMT reducing fall risk in older people.

Comparison of interactive cognitive-motor training with traditional training regimens
In studies that compared ICMT to equivalent training programs (similar content, same dose) most comparisons did not show significant differences, suggesting equivalence of training programs. In a few studies, however, ICMT was found to be better than traditional balance and strength or aerobic training in improving physical and cognitive outcomes [32,34,37,39]. These studies were conducted in clinical settings; possibly indicating higher levels of motivation, higher exercise dose and closer supervision. Three of these four studies were also of high methodological quality, so it is possible that other included studies may have failed to demonstrate differences in physical and cognitive outcome measures due to methodological limitations.
The notion of combining cognitive and physical training is based on interrelationships between cognitive and motor functions [82]. Postural control does not simply consist of automated motor tasks but depends on input from higher cortical centres [83], especially from neural networks associated with attention and EF [84]. In addition to good evidence demonstrating cognitive functions improve following exercise interventions [78] there are also preliminary findings suggesting seated cognitive training has beneficial effects on motor functions [85][86][87]. For example, Verghese and colleagues found eight weeks of seated computer game play training improved gait speed under single and dual task conditions in lowfunctioning older people; an effect that could not be accounted for by increased levels of physical activity [87].
Using enriched environments in ICMT that require those central processes in addition to motor execution may improve outcomes more than traditional exercise training due to the ecological validity as well as the involvement and interaction of additional modifiable risk factors. In our review however, we were unable to establish consistent differences in functional domains in favour of ICMT. This heterogeneity may have been due, in part, to the low statistical power of many of the included studies. In a related study with a larger sample that did not include standing exercise, VR bike training significantly improved several measures of executive functioning compared with traditional stationary bike training [88]. This VR training effect also exceeded the sum of effects of separate training regimens as reported in the literature, suggesting a synergistic effect [88]. Other studies support this finding in that they report combined physical and cognitive training leads to larger improvements in cognitive, physical and emotional outcomes compared to physical or cognitive training only [20][21][22][23].
The feasibility of the lower-cost ICMT exergames and their equivalence with traditional training programs suggest several advantages. ICMT fulfil several criteria to increase adherence and adoption to effective exercise interventions, such as realistic goal-setting, positive reinforcement while exercising, feedback, and the ability to self-monitor one's performance [89][90][91]. In addition, due to their easy use and relative low costs they could be deployed in the homes of older people with possible significant cost savings [92]. However, further research is required in this area as only two studies have applied systems within older people's homes [43,58] and no studies have conducted cost-effectiveness, cost-utility or cost-benefit analyses of their interventions.

Limitations of this review
We acknowledge this review has certain limitations. First, it is possible we neglected some trials that were not published in the main databases or referred to by other articles. Second, studies published in languages other than English, German or Dutch were not included. Third, it was not always possible to accurately describe and characterise the included studies due to inadequate reporting. Additional information sought from study authors was obtained for 14 studies [30][31][32]34,35,37,40,42,45,54,56,58,62,63] which assisted in providing more detailed descriptions of the interventions trialled. Finally, due to the heterogeneity in study designs, outcome measures and populations we were unable to conduct a meta-analysis.

Conclusions
This review shows that the effect of interactive cognitivemotor training on falls remains unclear with only one study including falls as outcome measure. There is evidence from multiple small studies showing that ICMT improved physical and cognitive factors associated with falls in older people but inconsistent findings have been obtained for psychological measures associated with fear of falling. Limited evidence from few studies suggests that ICMT are equivalent to traditional exercise interventions in their effect on fall risk factors.
These review findings have to be regarded with caution due to methodological issues, small sample sizes and poor reporting of the included studies. There is a need for highquality trials sufficiently powered to show differences in fall rates between groups. In addition, larger trials are required to identify small but meaningful differences between ICMT groups and equivalent traditional training controls. Underlying mechanisms should be explored to determine the interplay between sensorimotor and cognitive functions. Although cost-saving in theory, no studies have investigated cost-effectiveness of their interventions and only a few studies have administered ICMT in the home setting. Future studies therefore should examine these aspects of trial provision.