Effects of virtual reality-based cognitive training in older adults living without and with mild dementia: a pretest–posttest design pilot study

Objective Modern technologies are increasingly used in the development of cognitive interventions for older adults. Research into possible applications of virtual reality in such interventions has begun only recently. The aim of present study was to evaluate the effects of 8 sessions of VR-based cognitive training using the GRADYS game in healthy older adults (n = 72; aged 60–88) and older adults living with mild dementia (n = 27; aged 60–89). Results Older adults with mild dementia demonstrated worse baseline cognitive performance than participants without dementia. Both groups showed progress in training, which was greater in healthy older adults. There were also significant differences in cognitive functioning before and after the training. However, positive changes were revealed almost exclusively in the group of older adults without dementia. Based on the findings, we can recommend the GRADYS game for cognitive enhancement and as a possible counter-measure for cognitive decline experienced in normal cognitive ageing. Our results provide also support for the usefulness of VR technology in cognitive interventions in older adults. The use of the GRADYS game in persons living with dementia, however, would require several of the hardware and software modifications. Trial registration ISRCTN17613444, date of registration: 10.09.2019. Retrospectively registered


Introduction
The dynamic development of modern technologies raises questions as to the possibilities of their application in older adults in order to support their cognitive functioning. The efficacy of cognitive training in healthy older adults has been confirmed by meta-analyzes [1][2][3][4]. Several types of cognitive interventions [5] have proven beneficial also in older adults diagnosed with mild cognitive impairment (MCI) [6][7][8][9][10] which assumes cognitive functioning at the level between a healthy cognitive ageing and dementia [11]. Studies conducted on participants living with dementia are less clear. Some reviews and meta-analyzes have indicated positive effects of CI in dementia [12,13], but others have led to opposite conclusions [14,15]. Furthermore, positive effect on cognitive functioning in dementia is caused by cognitive stimulation, but not by cognitive training [16][17][18]. Currently, more and more studies on cognitive interventions employ modern technologies such as computerized cognitive training and video games. Meta-analyzes and systematic reviews provide evidence for the efficacy of this type of interventions in healthy older adults [19][20][21][22][23] and persons with MCI [24][25][26][27]. In persons with dementia, however,

BMC Research Notes
*Correspondence: lzajac@ukw.edu.pl 1 Faculty of Psychology, Kazimierz Wielki University, Staffa 1, 85-867 Bydgoszcz, Poland Full list of author information is available at the end of the article the evidence remains unclear and questionable [24,28,29]. A relatively new research area of growing interest in cognitive interventions concerns the use of virtual reality (VR). One of major benefits of VR is that it provides an immersive and naturalistic environment, which might increase the ecological validity of the intervention [30,31] and allows for training of cognitive skills that are relevant for real-world contexts [32][33][34]. This can make it easier for persons with dementia to benefit from cognitive training. Nevertheless, there is little research on the efficacy of VR-based cognitive interventions in relation to cognitive aging. In a systematic review of VR applications in healthcare, none of the included studies concerned cognitive interventions in older adults [35]. In a newest systematic review, concerning the efficacy of technologybased cognitive training in persons with MCI, VR was used in two out of 26 studies [26]. In the most recent meta-analysis on the efficacy of VR-based interventions in persons with MCI or dementia, VR-based cognitive training was used in 6 out of 11 studies and only one of them used a fully immersive technology [36]. This metaanalysis indicated a medium effect of VR-based interventions for cognition, larger in participants with MCI than in persons with dementia.
Recently, several studies on the use of VR-based cognitive training in healthy older adults or older adults with MCI have been published [34,[37][38][39][40]. However, the study samples were usually small, including case studies, and some articles have described only partial results from initial phases of the study or even only the design of the intervention.

Method
A pretest-posttest study design was applied to evaluate the effects of the VR-based cognitive training using the GRADYS game in older adults without and with mild dementia (Additional file 1: Figure S1, Additional file 2).
The exclusion criteria from the sample included the presence of mental disorders and serious somatic illnesses, as well as the presence of visual, auditory and motor impairments that could prevent the use of the game. All these criteria have been checked in a structured interview with the participant and verified by the caretaker in the case of participants with dementia. The initial sample comprised 150 participants, 75 in each group. The complete data were obtained from 99 participants . Participants who withdrew from the study, ceased their participation before the 3rd training session. A significant dropout of participants with mild dementia were mainly due to difficulties in understanding and coordinating the game control interface.
The administered intervention was based on the GRADYS game, which is a VR-based cognitive training containing four modules: attention, memory, language, and visuospatial processing. The storyline of each module scenario consists of tasks inspired by daily life. Each module has three difficulty levels. The game software included also a tutorial module to help participants to learn how to operate on the game interface. The game was controlled with the Oculus Rift DK2 and the Xbox 6DOF control pad.
Both study groups underwent eight individual training sessions, two per week. Each session consisted of two game modules: memory and attention or language and visuospatial processing. A single session lasted from 45 min to an hour. In each module the participants started at the lowest difficulty level and moved to a higher level in the next session having reached 75% accuracy in the previous one. Similarly, in the case of accuracy falling below 50%, the participants returned to a lower difficulty level. The participants were accompanied by a training assistant throughout the whole session.

Results
Better baseline cognitive performance was observed in the group of healthy older adults in general (Hotelling T 2 = 130.868; p < 0.001) and in the majority of cognitive measures (Table 1). In a few of the cognitive measures, no significant differences between groups were observed. Two performance indices in the sustained attention task (d2) were significantly better in the group of older adults with dementia. Both groups demonstrated progress throughout the training (the Friedman's test: p < 0.001 for each module in both groups). Yet the group of older adults without dementia showed larger gain in the majority of cognitive modules, except the attention module. Differences between the groups in terms of the maximum difficulty level reached until the last training session, analysed by the Mann-Whitney U test, were significant with regard to memory (p = 0.002), visuospatial processing (p < 0.001), and language (p = 0.006), but not attention (p = 0.053).
The results of repeated measures multivariate analysis of variance (RM MANOVA), supplemented by oneway tests, indicated improvement after the training, but mainly in the group of healthy participants (Tables 2 and  3).
The RM MANOVA with group as a between-subject variable ( Table 2) revealed a significant pre-post training difference in cognitive functioning. The interaction effect was not significant. However, the probability value was only slightly above the assumed value of statistical significance (α = 0.05), while partial eta squared indicated a large effect size [48], which suggested that the group may

Table 1 A comparison of baseline cognitive performance in two research groups
The means and standard deviations for all the above mentioned cognitive measures in both research groups are presented in Table 3  be a factor moderating the effect of training with regard at least to some cognitive measures. According to RM MANOVA computed for each group separately (Table 2), significant changes occurred only in participants without dementia. In the group of older adults with mild dementia the changes were not significant, however the observed power was low, in turn the effect size was high. It suggests that the statistical power of the test was insufficient to estimate the changes in this group with possible reasons being, at least partially, too small a sample size and a high sampling error.
In turn, one-way tests for the pretest-posttest differences in particular cognitive measures indicated several significant changes in the group of older adults with mild dementia. However, much fewer than in the group of healthy participants (Table 3).

Discussion
The results lead to a general conclusion that the GRADYS game may be effective cognitive intervention in older adults without dementia. However, the usefulness of the current version of the GRADYS game in persons with mild dementia is questionable.
As expected, before training, participants with mild dementia demonstrated lower cognitive performance than healthy older adults. In a few relatively easy cognitive measures no significant group differences were observed. Interestingly, two performance indices of d2 were significantly better in the group of older adults with mild dementia (WZ, WZ-B). The WZ indicator, however, is vulnerable to overestimation due to the skipping a part of the characters in the row, which was frequently observed in participants with dementia. The value of WZ-B, is a derivative of the WZ value.
Both groups showed progress in the course of training. Nevertheless, among older adults with mild dementia not only less progress, but also a there was a large withdrawal of participants from the sample. Therefore, the results obtained in the group of persons with mild dementia are less reliable and should be interpreted with caution.
Regarding the effects of the training on cognitive functions beyond the game environment, a significant difference in cognitive tests performance before and after training yielded in RM MANOVA suggests a positive impact of the training. At the same time, oneway tests revealed that positive changes were observed almost exclusively in older adults without dementia. Healthy older adults also improved in visuospatial processing, visual aspects of memory and working memory, but not in verbal learning and language. In contrast, in the group of older adults with mild dementia positive changes were observed only in the percentage of errors in the d2 test and in the copying task of ROCF, which belong to the easiest tasks used in the assessment. In this group also a negative cognitive change was observed (in BVRT).  Poor training effects in participants with mild dementia, despite its efficacy in healthy older adults are consistent with the results of previous studies, which have reported that in persons with dementia cognitive training is less effective [16][17][18]. We hoped that a near-natural VR environment would enhance training efficacy in this group, but we did not observe it.

Limitations
• The lack of a control group makes it difficult to isolate the training-related effects from the repeated measurement and practice effects. However, in order to reduce the likelihood of the practice effect occurrence in the majority of cognitive measures applied in cognitive assessment, alternative versions of the test were used in pretest and posttest (except for those, where no such versions exist). In addition, because the improvement occurred for some cognitive measures using different versions in the pretest and posttest, and at the same time the improvement did not occur for some measures using the same versions in the pretest and posttest, we conclude that the observed improvement cannot be attributed to the simple practice effect. Further development of the GRADYS game requires evidence of its efficacy obtained in randomised controlled trial. • Experiencing significant difficulties in game control due to an excessive cognitive burden in using new technology, older adults with mild dementia, were often unable to concentrate on cognitive tasks. To operate the game interface participants had to memorize and coordinate the functions of several buttons on the pad under the right and left thumbs as well as several ways of controlling the game with the Oculus. Additionally, the Oculus made it impossible for participants to visually verify their choice of buttons on the pad, so they had to maintain not only the functional but also spatial mapping in their working memory at all time. Interface control was difficult to achieve in participants with mild dementia even when they were provided with a practice tutorials. Thus, the cognitive burden of game control might have led to low learnability and caused a decreased efficacy of the training in older adults with mild dementia and their withdrawal from training. A significant dropout of participants with mild dementia due to the mentioned game control difficulties limits the conclusions to individuals with mild dementia who maintain a relatively high overall cognitive performance that allows them to understand and coordinate the game control interface. Further development of the game would requires making the game control more natural and easier.
Additional file 1: Figure S1. The structure of the intervention.