Efficacy and safety of non-immersive virtual reality exercising in stroke rehabilitation (EVREST): a randomised, multicentre, single-blind, controlled trial

Background: Non-immersive virtual reality is an emerging strategy to enhance motor performance for stroke rehabilitation. There has been rapid adoption of non-immersive virtual reality as a rehabilitation strategy despite the limited evidence about its safety and effectiveness. Our aim was to compare the safety and efficacy of virtual reality with recreational therapy on motor recovery in patients after an acute ischaemic stroke.

Methods: In this randomised, controlled, single-blind, parallel-group trial we enrolled adults (aged 18-85 years) who had a first-ever ischaemic stroke and a motor deficit of the upper extremity score of 3 or more (measured with the Chedoke-McMaster scale) within 3 months of randomisation from 14 in-patient stroke rehabilitation units from four countries (Canada [11], Argentina [1], Peru [1], and Thailand [1]). Participants were randomly allocated (1:1) by a computer-generated assignment at enrolment to receive a programme of structured, task-oriented, upper extremity sessions (ten sessions, 60 min each) of either non-immersive virtual reality using the Nintendo Wii gaming system (VRWii) or simple recreational activities (playing cards, bingo, Jenga, or ball game) as add-on therapies to conventional rehabilitation over a 2 week period. All investigators assessing outcomes were masked to treatment assignment. The primary outcome was upper extremity motor performance measured by total time to complete the Wolf Motor Function Test (WMFT) at the end of the 2 week intervention period, analysed in the intention-to-treat population. This trial is registered with ClinicalTrials.gov, number NTC01406912.

Findings: The study was done between May 12, 2012, and Oct 1, 2015. We randomly assigned 141 patients: 71 received VRWii therapy and 70 received recreational activity. 121 (86%) patients (59 in the VRWii group and 62 in the recreational activity group) completed the final assessment and were included in the primary analysis. Each group improved WMFT performance time relative to baseline (decrease in median time from 43·7 s [IQR 26·1-68·0] to 29·7 s [21·4-45·2], 32·0% reduction for VRWii vs 38·0 s [IQR 28·0-64·1] to 27·1 s [21·2-45·5], 28·7% reduction for recreational activity). Mean time of conventional rehabilitation during the trial was similar between groups (VRWii, 373 min [SD 322] vs recreational activity, 397 min [345]; p=0·70) as was the total duration of study intervention (VRWii, 528 min [SD 155] vs recreational activity, 541 min [142]; p=0·60). Multivariable analysis adjusted for baseline WMFT score, age, sex, baseline Chedoke-McMaster, and stroke severity revealed no significant difference between groups in the primary outcome (adjusted mean estimate of difference in WMFT: 4·1 s, 95% CI -14·4 to 22·6). There were three serious adverse events during the trial, all deemed to be unrelated to the interventions (seizure after discharge and intracerebral haemorrhage in the recreational activity group and heart attack in the VRWii group). Overall incidences of adverse events and serious adverse events were similar between treatment groups.

Interpretation: In patients who had a stroke within the 3 months before enrolment and had mild-to-moderate upper extremity motor impairment, non-immersive virtual reality as an add-on therapy to conventional rehabilitation was not superior to a recreational activity intervention in improving motor function, as measured by WMFT. Our study suggests that the type of task used in motor rehabilitation post-stroke might be less relevant, as long as it is intensive enough and task-specific. Simple, low-cost, and widely available recreational activities might be as effective as innovative non-immersive virtual reality technologies.

Funding: Heart and Stroke Foundation of Canada and Ontario Ministry of Health.