The effect of virtual reality on executive function in older adults with mild cognitive impairment: a systematic review and meta-analysis

Abstract Objectives This review aimed to summarize the currently available premium evidence to determine the effect of virtual reality (VR) on executive function (EF) in older adults with mild cognitive impairment (MCI), and to detect what level of immersive VR would be the most beneficial. Method Five electronic databases, namely, PubMed, Embase, PsycINFO, CINAHL, and Cochrane Library were searched. Our research team screened the studies and extracted data according to our inclusion criteria. The methodological quality of each study was rated using the PEDro scale. When three or more studies reported the same outcome, a meta-analysis was conducted using Review Manager 5.4.1. Results Finally, 14 randomized controlled trials with a total of 518 participants were included. VR training had an overall positive effect on cognitive flexibility, global cognitive function, attention, and short-term memory compared to the control groups. Additionally, semi-immersive VR was more effective in improving cognitive flexibility compared to the other two types of VR. The application of non-immersive level of VR had a significant effect on global cognitive function, attention, short-term memory, and cognitive flexibility. Conclusion VR may be effective in improving EF in older adults with MCI. However, the level of immersive VR that would be the most beneficial on EF still needs to be investigated with a greater number of well-designed studies.


Introduction
Dementia is considered to be a syndrome rather than a particular disease. it is mainly characterized by cognitive decline, and has a significant adverse impact on independent daily functional activities (Gale et al., 2018). it is estimated that by 2030, the number of people with dementia worldwide will reach 78 million (world Health Organization, 2021), which will place a heavy burden on global public health. The risk of acquiring dementia rises as an individual becomes older, especially once beyond the age of 53 (Tisher & Salardini, 2019). Mild cognitive impairment (MCi), a transitional stage of cognitive impairment lying between normally aging individuals and people with dementia (Tangalos & Petersen, 2018), may evolve into dementia within three years, with an incidence of up to 46% (Pal et al., 2018). As MCi is a progressive disease, early detection and treatment are necessary to slow down the progression of dementia (Knopman & Petersen, 2014). executive function (eF), a complex process involving multiple skills like working memory (wM), inhibition control, and cognitive flexibility (Blair, 2017;Diamond, 2013), plays an important role in our functional independence. The inhibition control, which is commonly evaluated by Stroop task test (Meier et al., 2020), allows subjects to set priorities and resist impulsive actions or responses; wM refers to information reprocessing, and is commonly measured by a digit span test (Diamond, 2013); cognitive flexibility typically involves set-shifting or task switching measured by the wisconsin Card Sorting inspired Task (wCST) (Dajani & Uddin, 2015). The diverse skills mentioned above are highly interrelated. People with executive dysfunction will have problems with planning, problem-solving, organization, and their information processing speed. One study has reported that executive dysfunction becomes more pronounced with normal aging as it progresses to MCi (Kirova et al., 2015). A cross-sectional trial concluded that individuals at an early stage of MCi showed poorer performance regarding eF than their healthy counterparts (Seo et al., 2016). Moreover, early dementia can be characterized by the poor performance of wM (Kirova et al., 2015). in that case, the decline in eF may lead to a decline in cognitive function in older people with MCi (Kirova et al., 2015).
virtual reality (vR) -a computer simulation of a real or imagined three-dimensional (3-D) environment, which allows users to have the same experiences they would get in a similar real situation (D'Cunha et al., 2019), has stimulated the interest of researchers and clinicians since its first use in 1994 (Diaz-Perez & Florez-Lozano, 2018). Compared to traditional pen-and-paper training, it is symbolized as a systematic and controllable intervention that makes use of data visualization and provides immediate feedback based on the participants' performance (Charles et al., 2020). vR is usually categorized into three types according to the form of connection with the physical world, including non-immersive, semi-immersive, and full-immersive vR Thapa et al., 2020). The vR system consisting of a modern 3-D head-mounted display and wireless hand controllers is considered to be full-immersive vR (Strong, 2020); semi-immersive vR comprises a high performance graphics computing system which includes a large screen monitor, a large screen projector system and multiple television projection systems, such as the BTS Nirvana interaction system (Maggio et al., 2018); non-immersive vR refers to a virtual environment delivered via a standard computer monitor or television and controlled by operating the mouse or keyboard (Strong, 2020).
Over the past several years, trials have investigated the effect of vR on eF, but inconsistent findings have been reported. Positive statistically significant differences on eF between a vR group and a control group have been reported in some trials (Liao et al., 2019;Thapa et al., 2020), whereas no significant differences were reported in other studies (Maggio et al., 2018;Mrakic-Sposta et al., 2018). Additionally, as all the articles mentioned above only focus on the overview of eF rather than detailed descriptions of the effect of vR on each domain of eF, like wM, inhibition control, and cognitive flexibility. Consequently, a precise description of the effect of vR on eF in older adults with MCi has not been established. Furthermore, a deeper immersion leads to a greater presence of a virtual environment (Strong, 2020), which may attract the participants' attention and result in more interaction with the vR training programs contributing to a better training result. However, so far, there has been no evidence that proves the effect different immersive levels have on improving eF. Therefore, this review aimed to systematically determine the effect of vR on eF in older adults with MCi, and to detect what level of immersive vR would have the greatest effect on eF in older adults with MCi.

Database searches and keywords search strategy
This systematic review was conducted in accordance with the Preferred Reporting items for Systematic Reviews and Meta-Analyses (PRiSMA) statement guidelines used extensively in health care interventions (Liberati et al., 2009) in November 2020. Five electronic databases, namely, PubMed, Cochrane Library, PsyciNFO, CiNAHL and embase, were searched using the terms 'virtual reality' , 'executive function' , and 'mild cognitive impairment' , combined with Boolean characters 'AND' and 'OR' . The details of the keywords search strategy are presented in Appendix 1. Additionally, a backward search was conducted by looking up the reference lists of eligible studies.  (Diamond, 2013) or an N-back test (Owen et al., 2005); (3) inhibition control: assessed by a Stroop task (MacLeod, 1991) or a go/no-go task (Cragg & Nation, 2008); (4) cognitive flexibility: assessed by a wisconsin Card Sorting Test (wCST) or a Trail Making Test-B(TMT-B) (Arnett & Labovitz, 1995

Exclusion criteria
• Participants: healthy older adults, individuals with schizophrenia or depression; • intervention: vR used for assessment; vR used for the control group; • Outcome measurement: no specific assessment of eF; Study design: studies other than RCT, such as reviews, casecontrol studies, and case reports.

Data extraction and quality assessment
The author (YD) screened the title, abstract, and full-text of each paper according to the inclusion criteria, and extracted the following information: the first author's last name, publication year, the contents of the treatment, and the eF assessment tools used. All of the information was confirmed by the second author (LX). if any controversy arose, the third author was consulted. All of the extracted data are presented in Table 1. The methodological quality of each study was assessed using the Physiotherapy evidence Database (PeDro) scale (Cashin & McAuley, 2020), resulting in a score ranging from 1 to 10. The research team then reviewed each item of the PeDro scale with reference to the PeDro official website to perform a scrutinized rating of each paper, with a higher rating indicating a better methodological quality (low quality: 1-3; fair quality: 4-5; good quality: 6-8; excellent quality: 9-10).

Data synthesis
The meta-analysis was performed using Review Manager 5.4.1. The post-training data with the mean and standard deviation (SD) were extracted to conduct the meta-analysis when more than three studies reported the same outcome. For continuous data, the effect size was reported as the mean deviation (MD) or standard mean difference (SMD), with 95% confidence intervals (Ci). A P value ≤ 0.05 was considered statistically significant. Furthermore, i 2 was used to measure the statistical heterogeneity. if i 2 was above 50%, the random-effects model was selected; otherwise, the fixed-effects model was selected. in addition, a subgroup analysis was performed to evaluate the effect of the different levels of immersive vR on eF.

Study selection
The initial search yielded a total of 180 records. After removing the duplicates, 143 articles were left for screening according to titles and abstracts. Of them, 25 articles were selected. After a further screening of the 25 articles, two of the studies were excluded because the full-text was inaccessible (Jprn, 2017;Park et al., 2018), while remaining nine studies (Appel et al., 2019;Hsieh et al., 2018;Jacoby et al., 2013;Liao et al., 2020;Maggio et al., 2020;Man et al., 2013;Mirelman et al., 2013;Mirza & Yaqoob, 2018;Zając-Lamparska et al., 2019) were excluded for not meeting the inclusion criteria. The detailed screening information is shown in Figure 1. Finally, 14 RCTs were found to be eligible for inclusion in this review.

Characteristics of the included studies
The included 14 studies contain a total of 518 participants. Among the studies, there was a considerable difference in sample size, ranging from 10 to 114, with 37 being the average sample size. The total duration of the training was from four to 60 h. except for one study (Amjad et al., 2019) that did not report information on age, the mean age of all the participants in the other 13 studies was 74.7 years old. More detailed information about the characteristics of the included studies is shown in Table 3.
Only one study using semi-immersive vR reported a statistically significant improvement (P < 0.01) compared to its control group (Maggio et al., 2018). However, the effect size could not be determined because of insufficient data.

Inhibition control
Three studies (Liao et al., 2019;Tarnanas et al., 2014) have assessed the inhibition control using the Stroop Color and word Test (SCwT), but the results did not reveal a significant difference between the intervention and control groups in those studies.

Cognitive flexibility
The

Global cognitive function
All 14 studies except one (Liao et al., 2019) tested the effect of vR-based training on global cognitive function in individuals with MCi. However, we extracted data from only 10 of the studies with a total of 320 participants due to the lack of well-presented data for the mean and SD for the remaining three studies (Maier et al., 2020;Monteiro-Junior et al., 2017;Optale et al., 2010). According to the meta-analysis, vR training resulted in an overall significant improvement on global cognitive function compared with the control group (SMD = 0.63, 95% Ci [0.06, 1.20], P = 0.03, i 2 = 81%; Figure 4).

Attention
Seven of the studies with a total of 229 participants ( 2020) were included in the meta-analysis which assessed attention using the TMT-A scale. However, one study (Mrakic-Sposta et al., 2018) was excluded from the meta-analysis due to the lack of post-training data. From the forest plot, it could be seen that there was an overall significant positive difference between the experimental and control groups (MD = −12.31, 95% Ci [-24.59, −0.04], P = 0.05, i 2 = 94%, Figure 5).

Short-term memory
Short-term memory was evaluated by the DST-F scale in five of the studies. However, a meta-analysis was conducted with four out of the five studies ( Figure 6) between the intervention and control groups.

Effect of the type of VR
Subgroup analysis based on the level of immersive vR were performed on cognitive flexibility, wM, global cognitive function, attention, and short-term memory. No significant differences among the subgroups were found for most of these outcomes, except for cognitive flexibility (P < 0.01). However, a greater number of significant positive results for cognitive flexibility, global cognitive function, and shortterm memory were achieved in the subgroup using non-immersive vR compared to the subgroups using the other two types of vR.

Compliance and attrition factors
Dropouts were reported in eight studies (Amjad et al., 2019;Hughes et al., 2014;Liao et al., 2019;Maier et al., 2020;Mrakic-Sposta et al., 2018;Optale et al., 2010;Tarnanas et al., 2014;Thapa et al., 2020), and all participants completed all the assessments and interventions in the remaining six studies. The dropout rate ranged from 3% to 20% in the eight studies; in three (   15%, mainly because of loss contact, hospitalization, dissatisfaction with the vR training or technical problems.

Discussion
This review aimed to determine the effect of vR on eF in older adults with MCi. According to the results, vR had a significant positive overall effect on cognitive flexibility, global cognitive function, attention, and short-term memory, compared to the control groups, offering a general conclusion that vR training may have a positive effect on eF in older adults with MCi. Current work indicates that vR has a significant positive effect on cognitive flexibility, but no significant effect on wM and inhibition control. After looking at an analysis of the insignificant results, it seems that the heterogeneity between the eF outcome measurements and the training content of the vR program may have led to the negative results. For example, the content of the vR in five of the studies Monteiro-Junior et al., 2017;Tarnanas et al., 2014) that reported the outcome measures of wM, such as driving (J. S. , making fruit cocktails (J. H. , and playing soccer , tended to focus on more on short-term memory and attention than on wM, the ability to store and process information (Baddeley, 1992). A similar mismatch between test and training content occurred regarding inhibition control. Therefore, an optimized strategy would be to orient training programs toward training goals. The following two reasons may explain the significant effect of vR on cognitive flexibility. Firstly, cognitive flexibility refers to the ability to switch flexibly between different tasks. Accordingly, a vR program will commonly contain two or more tasks specifically requiring cognitive flexibility. in addition, cognitive flexibility is a high-level cognitive control that involves basic cognitive skills, thus the improvement in attention and global cognitive function may be what is promoting the improvement in cognitive flexibility.
A significant improvement over control groups was reported in the global cognitive function and attention in vR groups. Regarding the result for attention, explanations for this phenomenon can be elaborated as follows: firstly, all of the included studies emphasized that vR can motivate and fully engage the participants by creating an artificial interactive environment; secondly, attention was the target training domain in the eight studies (Faria et al., 2016;Maggio et al., 2018;Maier et al., 2020;Mrakic-Sposta et al., 2018;Tarnanas et al., 2014). As for the result of global cognitive function, given that various kinds of vR were provided, including juice making (Thapa et al., 2020), shopping (J. S. , playing games (Maier et al., 2020), and practicing Tai Chi (Liao et al., 2019), the participants were exposed to a rich virtual environment involving various cognitive abilities, which may have led to the enhancement of their global cognitive function.
According to the subgroup analysis of the different outcomes, there is weak evidence to suggest which level of immersive vR was the most beneficial. However, a subgroup difference (P < 0.01) was achieved for cognitive flexibility. it can be seen from the forest plot that a significant positive difference was achieved in the subgroup using semi-immersive vR, and there was no overlap of the 95% Ci in comparison with the other two subgroups. Thus, indirect evidence supports the hypothesis that semi-immersive vR was better than full-immersive vR and non-immersive vR for promoting cognitive flexibility. except for that, according to the meta-analysis positive significant differences can be found mostly when using non-immersive vR. However, a definitive conclusion that non-immersive vR is the most effective in clinical practice cannot be made due to the lack of experimental studies directly comparing the effect of non-immersive vR with the other two types of vR.
From the results of the included studies, it can be seen that the effect of vR on eF may be associated with the content of the training protocol. Three studies (Amjad et al., 2019;Liao et al., 2019; using vR-based physical and cognitive training together reported a larger number of positive effects on eF compared to the reports of other studies using vR-based cognitive training alone, which may indicate that physical training can also help improve eF. This finding is consistent with the previous review (Hötting & Röder, 2013), showing that physical training helps improve cognitive function, suggesting that combining the two types of training may be more effective for people with cognitive impairment. Furthermore, studies using magnetic resonance imaging (MRi) have shown that a human's gray matter in the frontal brain regions (Colcombe et al., 2006) and the hippocampus (erickson et al., 2011) increased after physical exercise interventions. Physical exercise prepares the brain to respond to cognitive training, which will then trigger changes in neurons in specific networks associated with training skills. Therefore, vR-based physical and cognitive training together may be better than just vR-based cognitive training alone.
The overall attrition rate in the 14 studies seems good (less than 15%), except for three studies (Liao et al., 2019;Maier et al., 2020;Mrakic-Sposta et al., 2018) in which it was from 16% to 20%. The main reasons for the high attrition rate in these three studies included dissatisfaction with randomization, hospitalization, and low motivation. Although a number of studies have shown that vR training can improve the participants' motivation, which is the main reason for the high compliance rate, we still need to optimize the program to get more subjects to participate and persist in training so as to achieve the best state. Strategies for improving the adherence rate in older people with MCi may include the following aspects: provide more support and feedback during the training period; choose a suitable training format for the participants, such as endurance/resistance training, which can significantly affect the adherence rate (Di Lorito et al., 2020); choose shorter (in weeks) or less frequent (in weekly sessions) interventions as that may make it easier for people to adhere to the training (Xu et al., 2020).

Conclusion
in conclusion, vR is a promising technology that can be used to enhance eF. More accurately, vR has a positive effect on cognitive flexibility, but a non-significant effect on wM and inhibition control. Additionally, it was shown to have a positive significant effect on global cognitive function, attention, and short-term memory in older adults with MCi. Semi-immersive vR was found to be more effective in improving cognitive flexibility compared with the other two types of vR. in addition, vR-based physical and cognitive training together may help improve eF more than vR-based cognitive training can by itself, but further studies with direct comparisons between these two training protocols are needed to verify this conclusion. Last but not least, reducing the attrition rate can increase the reliability of a study. Thus, it is necessary to optimize the training program to guarantee that more participants will engage in and complete the whole process so as to achieve the optimum training results.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
The author(s) reported there is no funding associated with the work featured in this article.