Journal Archive

As of May 1, 2018, research literature related to virtual reality of stroke patients was collected. Overseas data were searched for ‘stroke’, virtual reality, training’ through Pubmed, the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, and EMBASE. The search was limited to English data from 1991 to October 1, 2018. A total of 2813 articles were searched.

Approval of request for exemption from deliberation by the institutional life review committee for data retrieval and collection; It took place from January 2018 to May 2018 after receiving (IRB No. SUIRB20). In addition, the data search was supplemented through the references of the searched papers. The search was limited to papers in English and papers published from 1991 to January 2018 that provided data from the DB. The data collection procedure of the study to be analyzed was based on a systematic literature review process. Literature search was conducted by a researcher and a research assistant with the advice of a meta-analysis expert. First, in MEDLINE, 1) mesh words, which are control words of simulation education intervention, were checked, and then all four mesh words were included. After including the related central keywords identified in the abstract, EMBASE checked the control words in EMTREE and performed the same method as above. In the Cochrane Library CENTRAL and CINAHL, the search was performed through the mesh identified in MEDLINE. Related articles were searched through a core electronic database. Duplicate data were removed from the literature managementDB (reference management database, [Refworks]). The original text of the selected thesis was reviewed after first checking the related thesis through the title and abstract of the thesis. Two researchers independently searched and selected the data, and inconsistent literature was matched through discussion. The authors and publication year, country, study design, age of subjects, number of subjects, characteristics of intervention type, and research results were extracted from the finally selected study and recorded in the electronic coding book. If the data did not match, the original text was reviewed and a final decision was made.

A total of 2813 articles were searched through thesis titles and abstracts in EMBASE, MEDLINE, Cochrane Library CENTRAL and CINAHL.Duplicate data were excluded through the reference management database ([Refworks]), leaving 499 articles.With the exception of 352 studies that did not meet the selection criteria, 47 studies were selected first. The original texts of the 47 first-selected studies were reviewed, and among them, the final 21 studies were selected, excluding studies that did not meet the selection criteria (Figure 1).

Figure 1. 
Flow diagram of study screening

For RCT research, the 7-item RoB (The Cochrane's Risk of Bias) tool developed by The Cochrane Bias Method Group was used. Methodological quality evaluation of the thesis was conducted after conducting an RCT research pre-test with a professor with experienced in meta-analysis. Two researchers evaluated each of the separate questions, and those that did not match were re-evaluated after reviewing the original text.

The general characteristics of the research paper and the intervention method are analyzed by 12 questions (research conducting country, research design, subject, sample size, subject, intervention type, intervention period, number of interventions, intervention time, control group type, measurement tool, outcome variable) made by the frame. The average duration of one intervention, the total number of applications and the total number of weeks of application for the intervention method was investigated. For the effect size of upper extremity function, the Box and Block test, the modified Ashworth scale, and the joint ROM were selected. For the balance variables, the berg balance scale measurement tool was selected, and the gait variables were the stand-up walking test, stride length, and gait function. The walking rate was selected. The effect size of the intervention was meta-analyzed using the RevMan 5.3.3 program of the Cochrane Library. Analysis was conducted when there were 3 or more papers that provided the same outcome variable, and subgroup analysis was performed when there were 2 or more papers. When analyzing data, if there was a mean difference value and standard deviation before and after the experiment in the target paper, the values ​​were used. If there are only the mean and standard deviation before and after the experiment, the mean difference value is the value obtained by subtracting the pre-experiment from the post-experiment, and the standard deviation is calculated. The effect size was calculated using a random effects model in which weights were reset in consideration of variation among subjects of individual studies and heterogeneity between studies when homogeneity was not secured. The homogeneity of the subject studies was confirmed through Cochrane's chi-square test and I² test. I² value becomes 0% when there is no heterogeneity, 30-60% means moderate heterogeneity, and more than 75% means high heterogeneity. When data input was performed two or more post-mortem measurement of outcome variables, the average value was included, and the standard deviation was calculated using the pooled standard deviation formula. The statistical significance of the effect size (ES) was determined by the overall effect test and the 95% confidence interval (CI), and the significance level of 5% was the standard. As for the interpretation of the effect size, an effect size of ES=0.20-0.50 means a 'small effect', an effect size of ES=0.50-0.80 means a 'moderate effect', and an ES=0.80 or more means a 'large effect' (Cohen, 1988).

Table 1 shows the general characteristics of 21 virtual reality intervention studies analyzed in this study. There were 3 titles before 2010 (14.3%), 6 films from 2011 to 2013 (28.6%), and 12 films from 2014 to 2016 (57.1%). The source of the study was in the order of 10 Korea (47.6%) 2 USA (9.5%) 2 Brazil (9.5)% Canada 2 (9.5%) Other countries (UK, Israel, Spain) 5 (23.8%) appear. As for the research design, all 21 papers were presented as RCT experimental papers. The mean age of the participants in the total study was 60 years, 8 (38.1%) 50-60 years old, and 10 61-70 years old (47.6%). The total number of participants in the study was 642. The types of interventions were virtual reality training (VRT) in 16 patients (76.2%) and VRT + general physical therapy (EST, NDT, occupational) in 5 patients (23.8%). As for the intervention period, the average intervention time for each virtual reality training (VRT) was 38.4 minutes/time, and the weekly average was 4.12 times/week to 4.5 weeks. In the intervention group, in which virtual reality training (VRT) and general physical therapy were added, 42.0 minutes/time for virtual reality training (VRT), and 4.2 times/week to 4.4 weeks for general physical therapy 19.0 minutes/time, and 4.0 times/week in 4.5 weeks. Finally, the outcome variables were physical activity, balance, muscle strength and endurance, and quality of life. Finally, the outcome variables were physical activity, balance, strength and endurance, and quality of life. Among them, the measurement variables of gait affecting daily living activities were the Timed Up and Go test (TUG) stride length, gait function, and gait rate. The Berg balance scale was selected as the balance variable, and the box and block test, the modified Ashworth scale, and the joint range of motion were selected as the measurement variables of upper extremity function. The effects of virtual reality training on gait, balance, and upper extremity function of stroke patients were reviewed.

Measured variables on gait as an intervention effect of virtual reality training: Timed Up and Go test (TUG), stride length, and gait function A total of 6 studies were presented to measure the effect size of all four variables in walking rate on walking, and the number of subjects was 166, 84 in the experimental group and 82 in the control group. The effect size for gait was -2.19 (95% CI: -3.99, 0.38), and the effect size between the experimental group and the control group was statistically significant (p=.02). And the heterogeneity test did not show a significant result (p=.94), indicating homogeneity (Figure 2).

Figure 2. 
Forest plot of effect on gait in virtual reality rehabilitation intervention

First, four studies were presented to measure the effect size of the standing walking test, and the number of subjects was 120, 61 in the experimental group and 59 in the control group. The effect size for TUG was -1.74 (95% CI: -3.67, 0.19), and the effect size between the experimental group and the control group was not statistically significant (p=.08) (Figure 4).

Figure 4. 
Forest plot of effect on gait in virtual reality rehabilitation (TUG)

A total of 3 studies were presented to measure the stride effect size, and the number of study subjects was 60, 30 in the experimental group and 30 in the control group. The effect size between the experimental group and the control group was -0.45 (95% CI: -0.97, 0.06), and the effect size between the experimental group and the control group was not statistically significant (p=.08) (Figure 6).

Figure 6. 
Forest plot of effect on virtual reality rehabilitation gait (Step length)

A total of 3 studies were presented to measure the effect size of the gait function, and the number of study subjects was 60, 30 in the experimental group and 30 in the control group. The effect size on the gait function was -0.30 (95% CI: -0.81, 0.21), and the effect size between the experimental group and the control group was not statistically significant (p=.25) (Figure 8).

Figure 8. 
Forest plot of effect on virtual reality rehabilitation intervention on gait (Stride)

A total of 3 studies were presented to measure the effect size of the walking rate, and the number of subjects was 60, 30 in the experimental group and 30 in the control group. The effect size on the gait function was -0.22 (95% CI: -0.42, 0.02), and the effect size between the experimental group and the control group showed a statistically significant result (p=.03). However, the size of the effect was -0.22, indicating that the effect was small (Figure 10).

Figure 10. 
Forest plot of effect on gait in virtual reality rehabilitation intervention (Cadence)

As an intervention effect of virtual reality training, the effect size of the berg balance scale was tested as a measurement variable on balance. A total of 5 studies were presented to measure the BBS effect size, and the number of study subjects was 100, 50 in the experimental group and 50 in the control group. The effect size for BBS was -2.05 (95% CI: -3.55, -0.61), and the effect size between the experimental group and the control group was statistically significant (p=.005). The size of the effect was -2.05, indicating that the effect was very large (Figure 12).

Figure 12. 
Forest plot of effect on the balance of virtual reality rehabilitation interventions (BBS)

As an intervention effect of virtual reality training, the effect size of each of the three variables was tested with the box and block test, the modified Ashworth scale, and the joint range of motion as the measurement variables on the upper extremity function. First, four studies were presented to measure the effect size of the box and block test, and the number of study subjects was 222, 113 in the experimental group and 109 in the control group. The effect size for the box and block test was 0.46 (95% CI: -1.03, 0.12), and the effect size between the experimental group and the control group was not statistically significant (p=.12) (Figure 14).

Figure 14. 
Forest plot of effect on upper extremity function of virtual reality rehabilitation intervention (BBT)

A total of 3 studies were presented to measure the effect size of the modified Ashworth scale, and the number of subjects was 83, 42 in the experimental group and 41 in the control group. The effect size for the modified Ashworth scale was 0.04 (95% CI: -0.26, 0.33), and the effect size between the experimental group and the control group was not statistically significant (p=.80) (Figure 16).

Figure 16. 
Forest plot of effect on upper extremity function of virtual reality rehabilitation intervention (MAS)

A total of three studies were presented to measure the effect size of the joint ROM, and the number of subjects was 60, 30 in the experimental group and 30 in the control group. The effect size for the joint ROM was 0.05 (95% CI: -0.56, 0.46), which was not statistically significant (p=.84). The effect of virtual reality training on upper extremity function was found to have no effect size in the measurement variable box and block test, the modified Ashworth scale, and the joint ROM (Figure 18).

Figure 18. 
Forest plot of effect on upper extremity function of virtual reality rehabilitation intervention (ROM)

The walking style of stroke patients is usually characterized by slow, excessive effort, urgency, and poor coordination. In addition, the method of movement shows a synergistic method of mass movement flexion and extension without selective control of the trunk and extremities (Caillet et al, 2003). Although the pattern of gait exhibited by each stroke patient varies considerably, stroke patients generally exhibit a significantly slower gait speed compared to normal people (Hsu et al, 2003). In this regard, Wagenaar and Beek (1992) reported that stroke patients were associated with decreased speed and shortened girth. von Schroeder et al (1995) showed that stroke patients showed a decrease in speed, an increase in the gait cycle, and an increase in the biceps support phase, compared to a normal person, and the paralyzed side showed a decrease in stance phase, an increase in swing phase, etc. was said to represent. The asymmetrical gait patterns in stroke patients include a slow gait cycle and gait speed, a difference in stride length between the paralyzed side and the non-paraplegic side, and a short stance phase and long swing phase on the paretic side (Mauritz, 2002). The cause of the gait pattern is slow movement due to a decrease in selective motor control ability, compensation for this on the uncompensated side, loss of proprioceptor sensory loss, and decreased movement control of the ankle joint such as stiffness, joint flexion due to abnormal coordination between flexor and extensor This is because a characteristic gait pattern called 'hemiplegic gait' appears by inducing a pattern or joint extension pattern (Sakuma et al, 2014). Due to this unique gait pattern of 'hemiplegic gait', stroke patients show an inefficient gait pattern that consumes a lot of energy when walking. After stroke, weakness in lower extremity muscle strength is a major factor in performing functional activities such as walking and maintaining posture. Foot dropping occurs due to pathological reasons such as weakness of the dorsiflexor, increased activity of the plantar flexor, and contracture, resulting in a decrease in the angle of dorsiflexion in more than 45% of patients (Bethoux et al, 2014). In this study, gait variables were selected as standing walking test, stride length, gait, and gait rate.

The results of this study showed that the gait rate function [standardized mean difference (SMD) - 0.22, 95% confidence intervals (CI) - 0.42 to –0.02] was statistically significant. It was found that there was no statistically significant effect on TUG, guaranteed distance and stride function. These results are similar to the results of Mirelman (2008) and Yang et al (2008) showing that the effect of virtual reality training on walking speed was not statistically significant. However, in the study result of Jaffe et al (2004), the VR intervention group had a significantly greater effect than the comparison group at the fast walking speed in the study on the effect of virtual reality training on slow walking speed and fast walking speed. In the study of Jung et al (2012), the VR intervention group showed improved results than the control group in the evaluation results of the standing walking test measurement tool (Cohen's d = 0.78), but it was not statistically significant between the two groups. These results show that contradictory results can be derived depending on the intervention method or measurement tool.

In this study, the berg balance scale was selected as a tool to measure the effect of virtual reality training on balance, and the results were statistically significant (SMD –2.05, 95% CI -3.50 to −0.61), which was found to have an effect. Balance disorders are thought to be common after stroke, slowing recovery of daily activities and mobility, and increasing the risk of falls (Loewen et al, 1990; Lamb et al, 2001). In the study of Kim Jung-hwi (2005), after virtual reality training, dynamic average balance, anteroposterior stability limit, and left-right stability limit, which are dynamic balance variables, increased significantly in the training group (p<.01). There was a significant difference in both comparisons with the control group (p<.01). After virtual reality training, the berg balance scale score significantly increased in both the training group and the control group (p<.01), and there was a significant difference in the comparison of the mean difference between groups before and after training (p<.01). According to the research results (Ski æ ret-Maroni et al, 2016) and Kim et al (2011) that virtual reality-applied stepping exercise is effective in improving balance in patients with chronic stroke, balance ability of stroke patients through virtual reality games Nintendo Wii using reported to have an effect on improvement. Llorens et al (2014) reported a significant increase in BBS score and 10-m gait test over time in a study on the effect of virtual reality-based stepping exercise on balance improvement in chronic stroke patients. It is considered to have a significant impact on ability. However, in the study result of Kim (2005), it was found that the method using virtual reality did not show a significant effect on the evaluation of static balance because it consisted of a dynamic exercise task that moves the whole body. Therefore, there are still insufficient areas to conclude that virtual reality training has a significant effect on the balance of stroke patients, indicating that further research is needed.

There was no significant effect on upper extremity function (box and block test, modified Ashworth scale, range of motion). Patients with hemiparesis after stroke show abnormal movement patterns due to abnormal muscle tone, hyperopia reflexes, joint movements, coordination, and association reactions, and the motor function of the damaged limbs is greatly reduced due to musculoskeletal disorders and sensory disturbances. As a result, it adversely affects physical activity related to daily life (Langhorne et al, 2009). In 9 studies with a total of 190 participants, the virtual reality training intervention showed a moderate to moderate effect compared to the non-intervention group in the results of the virtual reality training intervention on upper extremity function (Coupar 2012; Kim 2011a; Kwon 2012; Yavuzer 2008). In addition, virtual reality is starting to be used recently as an intervention to help functional recovery of patients with disabilities, and patients can provide various feedbacks through virtual reality similar to reality even after the onset of disease (Lucca, 2009). In addition, it was argued that the complex provision of upper extremity sensory and functional activity to stroke patients contributes a lot to upper extremity motor function rehabilitation (Scalha et al, 2011). Virtual reality training is repetitive and focused training for improving upper extremity function necessary for daily life, and the difficulty of training can be adjusted according to the patient's functional level and needs. In addition, it has the advantage of being able to experience various virtual environments through active exercises that can be done by oneself (Rizzo and Buckwalter, 1997a) suggested that virtual reality training would greatly affect the upper extremity function of stroke patients, but the results of this study showed contradictory results. However, a study that applied virtual reality as a rehabilitation method for upper extremity function of stroke patients (Merians et al, 2002) and a study that applied virtual reality to patients with cerebral palsy and Erb's palsy to increase the range of motion and movement speed of the hand (King, 1993; David et al, 2001) results. Broeren et al (2004) also reported that there was a meaningful recovery of upper extremity function as a result of intensive training using a virtual reality program that induces upper extremity use in chronic stroke patients. Considering the above results, it is implied that it is somewhat unreasonable to conclude that virtual reality training is not effective for upper extremity function of stroke patients. In this study, the upper extremity function was limited to the box and block test, the modified Ashworth scale, and the joint range of motion, indicating the need for further analysis through various measurement tools. In addition, the upper extremity function plays a greater role in daily life movements than walking and balance, so it is thought that it is not significantly affected by the virtual environment and stimuli compared to walking and balance. This study has the following limitations. Although studies were selected through literature review, only the results of studies published in English were considered. Therefore, all relevant publications could not be identified. In this study, as the gait, balance and upper extremity function measurement variables, the gait variable was the Timed Up and Go test (TUG), the stride length, gait, and gait rate were BBS, the upper extremity function was the box and block test, and the modified Ashworth scale. The measurement results were derived by limiting the joint range of motion. Therefore, it is somewhat difficult to generalize the results of this study as there is a possibility that different results may be obtained depending on the selection of the measurement variables depending on the intervention method. In future studies, it is suggested that more intervention methods and studies considering measurement variables are needed.