Is upper limb virtual reality training more intensive than conventional training for patients in the subacute phase after stroke? An analysis of treatment intensity and content

Intensity and repetition have been identified as key factors for promoting neural plasticity [1]. It has been stated that Virtual Reality (VR) training using either specially developed systems or off-the-shelf gaming consoles provides the opportunity to achieve many repetitions, salient stimuli and engages the patients in a motivating and intense way [2]. Furthermore, VR is supposed to deliver task-specific training and multi-sensory stimulation [3]. Some evidence has been found that supports the use of VR for improving arm function after stroke [4], although a recent multicenter trial could not corroborate the superiority of a commercial VR gaming system [5]. As VR is becoming progressively more used in neurorehabilitation a more detailed analysis of VR intensity and treatment components is indicated.

Intensity can be expressed as dosage. There is consent that a higher dosage of movement practice can contribute to better outcomes [6, 7]. However, how the term “dosage” should be defined or which factors of dosage are relevant for improved outcome, is unclear. When examining dose–response relationships, Lohse et al. [8] found a positive and significant relationship between amounts of therapy provided and motor function improvement after stroke. However, the authors pointed out the need of a more precise measure of active time and repetitions. In a recent review Lang et al. [9] emphasized the need for a deeper understanding of dose–response relationships. Measuring amount of practice in terms of therapy sessions - scheduled or actually conducted - has been widely used, but does not reflect intensity or active time. It has been demonstrated that patients spend less than two-thirds of their treatment sessions actively and that physiotherapists tend to overestimate the amount of active practice [10, 11]. Many VR systems provide a substantial advantage by the integrated registration of time spent actively practicing and other information on training performance [3, 12].

Like duration of therapy time, also many repetitions of meaningful and challenging exercises are regarded as beneficial for regaining motor skills after stroke [13]. Timmermans et al. [14] identified 15 components to characterize task-oriented training and examined their relation to effect sizes. They found random and distributed practice, clear functional goals and feedback to be associated with larger effect sizes. These components, however, can be found in both VR and conventional training.

Rand et al. [15] used accelerometers to compare the amounts of purposeful movements elicited in a group of patients with stroke using video games and a control group receiving traditional therapy. They found that playing video games resulted in more purposeful repetitions (median 271) than traditional training (median 48). Also cognitive and emotional involvement, considered as key factors for regaining motor skills, may be facilitated by many VR applications due to their playful character [16‐18].

The objective of this study was to compare the intensity, here defined as time spent actively using the affected upper limb, and the content of a VR training intervention and a conventional task-oriented intervention. We hypothesized that the intensity of training was higher in the VR group and that patients in VR would achieve more repetitions.

Video recordings of 50 patients (n = 25 in VR, n = 25 in CT) were analyzed with regard to total training time, active training time, activity rate and content of training. Patient characteristics are presented in Table 1.

The groupwise mean time measures are presented in Fig. 1 while the results of the comparison of the treatment groups are presented in Table 2. We observed a significant difference in activity rate of 10.3 % and active time of 6.1 min, while there was no significant difference in total time. This is visible in the univariate analysis, which only shows the differences between VR and CT in the blue graphs in Fig. 1. The multivariate analysis, taking into account the interaction of severity of paresis and kind of intervention received, showed that patients with severe paresis (red) achieved a significantly higher activity rate, while patients with mild paresis (green) did not. The ARAT at study admission had been 7.5 (6.6) for severe and 39.5 (9.6) for patients with mild to moderate paresis. Also active training time and total training time was different for the severely impaired patients only. This effect was significant as shown by the significant interaction in the multivariate analysis.

The content of the training differed in several respects. While the tasks in the VR comprised solely task-related functional movements, tasks in the CT covered a broader range including strength exercises and mobilization. There was no difference between VR and CT with regard to repetitions (p = 0.80). However, only discrete functional movements could be counted and not continuous movements which were frequently performed in both groups. Table 3 provides an overview of the content of training in VR and CT. Tasks in VR were shorter, 2–3 min, but frequently repeated during one session and then counted as one task.

VR training resulted in a higher activity rate per training session. This was especially pronounced for patients with severe paresis who also had longer training sessions and more active time in VR than in CT. For patients with mild to moderate paresis, the difference was present, but less pronounced and not statistically significant. Our results suggest that a higher activity rate was easier to reach in the VR group. This is in accordance with what generally is regarded as a main benefit of VR training [2]. The entertaining and persistent character of the training may facilitate a higher training intensity making it easier to achieve a higher activity rate. VR training normally provides a wide range of stimulating tasks. Interestingly, this seems to be particularly relevant for patients with little motor function. Purposeful, task-related exercises are regarded as most beneficial for arm motor function after stroke [6]. However, it can be very challenging for therapists to find appropriate active exercises for this patient group and to continue training when the patient shows signs of strain. In suitable VR environments patients experience that even small movements can be translated into purposeful actions. A sense of achievement, to be able to do something with an otherwise useless hand can be very motivating for patients with severe paresis. Furthermore, the multimodal feedback provided by the VR system may facilitate cerebral reorganization in a phase where highest plasticity can be expected [21].

In general, our patients were active during 72 % their training sessions, which is more than identified in a previous review article where only 60 % active time was found [10]. Although the exact dose–response of rehabilitative treatment approaches is still unclear, the results of several reviews indicate a beneficial effect of augmented training time on functional outcomes and activities of daily living [7‐9]. It is recommended to find ways to increase therapy time after stroke and VR training seems to be an appropriate treatment alternative which also can increase adherence to exercises. Our intervention took place in a clinical setting under the supervision of a therapist, which facilitated adherence. However, in a home setting without supervision, it may be difficult to achieve recommend levels as demonstrated in a study by Standen et a. [22] Active time in therapy expressed as activity rate was the main focus of the current study. Measuring activity rate seems to be the most objective and tangible aspect of motor training. Comparing VR to dose-matched interventions in terms of scheduled therapy sessions only may not reflect the real extent of activity. [23] There was no difference for total time and active time between VR and CT when looking at all patients, which was expected since the training in both groups was supposed to be matched for these parameters per protocol. However, for the more impaired patients also longer total time was revealed, not only a higher activity rate and longer active time. Longer sessions in VR may indicate that it was easier to perpetuate training for a longer time.

While our main focus was on training aspects related to time, we also intended to explore other features of the training provided in both groups. Counting repetitions for discrete functional movements was of very limited value in this study. The number of repetitions did not reflect activity rate or intensity, since both the CT and the VR training comprised continuous tasks, such as steering a plane (VR) or cleaning a table or writing (CT). VR comprised probably a larger part of continuous tasks that could not be counted as repetitions. Severely impaired patients with no grasp function frequently had no repetitions or no grasps registered, although they had been active through large parts of the training session. Training in VR consisted of exclusively functional tasks, while training in CT also comprised other tasks, mainly mobilization. However, repetitions of discrete functional movements registered were substantially, roughly 3 times, higher in both VR and CT than those registered by Kimberley et al. [24], who found only a mean 40.64 repetitions per session. An also relatively low number of only 39 active and 12 purposeful movements per session were counted by Lang et al. [25]. Another study by Lang et al. revealed that functional UL movements occurred in 51 % of training sessions [26]. Comparison with these earlier results indicates that our patients both achieved a very high activity rate and relatively many repetitions. Still, the activity rate and number of repetitions does not come close to what is regarded as facilitating neural plasticity based on animal studies [21, 27]. Activity rate seems to be the closest we can get to gauge intensity. However, it has been demonstrated that motor learning and cortical reorganization is not based on stereotype repetitions and the simple use of the hand [28]. Plastic changes only occur when skilled tasks are practiced, and may not automatically generalize to untrained tasks [29, 30]. Motivation and engagement can facilitate general motor learning and retention [31]. It has been claimed that these requirements can be met by enriched virtual environments rather than by conventional therapy [2].

We also registered the kind of feedback delivered. The benefit of explicit feedback is still unclear. In some studies a beneficial effect of knowledge of performance on motor learning was found, while other studies suggest that explicit information interferes with implicit learning [32, 33]. In VR feedback was provided in a multi-modal way, since the patients could see scores reached, heard applause, felt vibrations on the hand. Also the training report that describes progress over a series of sessions can be regarded as feedback. Verbal feedback from the therapist comprising knowledge of performance and knowledge of results was provided in both groups. However, knowledge of performance with the intention to control for excessive compensatory trunk movement was provided in CT only. This may have allowed for more compensatory strategies in VR [32]. However, knowledge of performance can also be provided in VR by therapists or included as a part of the virtual environment as in the study by Subramanian et al. [34].

The fact that a session was video-recorded may have influenced both the patient and the therapist and could have resulted in altered behavior [35] and most likely in a way that effective training time was increased and small talk and extensive breaks discouraged. This may have affected CT more, since in VR the course of the sessions including breaks usually was pre-programmed. Since our study was cross-sectional we cannot tell if more was better in this case. Instead we have to wait for the results of the main study. The motor abilities of our patients were not assessed at the day of the video-recording. Therefore we cannot tell how much the initially severe motor impairment had improved that special day. We also have to acknowledge that the initial categorization based on active finger and wrist extension can be debated. However, we can tell from the video-recordings that patients with initially severe arm motor impairment still suffered from pronounced paresis at the day of recording. Although the video recordings were taken at a random day, we cannot rule out to have picked a session that was not quite representative for all sessions of this respective patient. Large deviations would have been reported by the therapist.

A markedly higher activity rate for patients with severe paresis in the VR group suggests that VR training facilitates more active training time. Also a longer active time and total time was registered for patients with severe impairments, indicating that it may be easier to adhere to the training. Only functional tasks were observed during training period in the VR group compared to only 65 % of the time in the CT group. Several modes of feedback were registered in the VR group. It remains to find out if these differences in quality of training between the programs will cause any difference in arm function.