2.1 Study setting
This trial was a parallel randomized, single-blinded, controlled trial. Data collection took place in the electrically shielded Brain Laboratory of the First Affiliated Hospital, Sun Yat-sen University, Guangzhou, China. No change to the study design was made after commencement.
This study was approved by the Ethical Committee of The First Affiliated Hospital of Sun Yat-sen University. All participants were provided with a comprehensive explanation of the experimental procedure and a participant information sheet. Participants were asked to provide written consent prior to study enrolment. The study was on the Chinese Clinical Trial Registry (http://www.chictr.org.cn/searchprojen.aspx; Registration No.: ChiCTR-IOC-15006064; Retrospective registration: May 2015)
2.3 Recruitment and sample size
Recruitment took place between July 2016 and June 2018 in the inpatient ward of the Department of Rehabilitation Medicine, The First Affiliated Hospital of Sun Yat-sen University. The inclusion criteria were: (1) first ever occurrence of stroke within 1 to 6 months; (2) stroke occurrence confirmed by magnetic resonance imaging (MRI) or computed tomography (CT); (3) age between 40 and 80 years old; (4) have at least 20° of wrist flexion/extension and at least 10° of finger flexion and extension of the paretic limb; (5) be able to sit at least 30 minutes without assistance; and (6) no severe cognitive impairment (Mini Mental State Examination >21)37. The exclusion criteria were: (1) large area of cerebral infarction; (2) brainstem injury; (3) hand deformity; (4) visual field deficits; and (5) have already received elements of the intervention program over 1 week. Suitable participants were identified by the clinical team during the routine medical admission procedure. Patients were then approached by a member of the research team who was not involved in providing medical care to explain about the study. All approached patients were asked to contact a member of the research team to express their interest.
Sample size was based on a pilot trial where 16 subacute stroke participants were allocated into VR and control group. CNV latency during paretic hand movement post intervention was used as primary outcome measure for sample size calculation. The calculation of sample size was conducted in the software GPower ver 3.1.2, using the “A priori: Compute required sample size –given α, power, and effect size” as type of power analysis. Preliminary results indicated a mean CNV latency of 1740.22ms (SD 120.78) for the control group and 1614ms (SD 122.61) for VR group post intervention，which gave an effect size of 0.50. With α error probability of 0.05 and a power of 0.95，a sample size of 36 was sufficient for the present study. Figure 1 shows the number of participants at each stage of the study.
2.4 Randomization, Concealment, and Blinding
Participants were randomly allocated to either VR group or control group in a 1:1 ratio by simple randomization. The randomization schedule was calculated in statistical software SPSS (IBM SPSS statistics version 20, USA) by a statistical expert from the Faculty of Medical Statistics and Epidemiology, Sun Yat-sen University. The sequence of allocation were kept in sealed envelopes and revealed by a member of the research team after the participant was enrolled. Participants were blinded from their group allocation but were informed that they had an equal chance of allocation to the VR or control group before study participation. The outcome assessor was also blinded to group allocation. The treating therapists were not blinded to group allocation. There was no planned interim analyses or stopping rules as power calculation was conducted to enable analysis
Interventions were delivered for 2 weeks with 5 training sessions per week. Each session lasted for 60 minutes. Designated therapists provided either VR training or occupational therapy for upper limb function in addition to routine medical care and other rehabilitation deemed necessary by treating physicians.
2.6 Outcome measures
Primary outcome measure was CNV latency. Secondary outcome measures were EMG onset time, CNV peak amplitude and functional scales. CNV peak amplitude was adopted to assess the cortical excitability during the motor planning phase. Clinical assessment scales of Action Research Arm Test (ARAT)38 (range: 0-57); Upper Limb Fugl-Meyer Assessment (FMA-UL)39 (range: 0–66); and National Institutes of Health Stroke Scale (NIHSS) (range: 0-40) were adopted to assess functions. Outcome measures were not altered after commencement of the study. All outcome measures were assessed before and after intervention.
The VR intervention was delivered through an interactive training system (A2, YiKang Ltd., China) which comprised of a weight support exoskeleton arm with high precision displacement velocity sensors. Graded upper limb VR training exercises were provided to train the motion of reaching and reach to grasp. Parameters such as grasp power, object moving speed, interval between objects, lateral left or right dispersion of object start positions were adjustable to suit participants functional capability and progressive difficulty level. During VR intervention, participants were seated in front of a table facing the monitor, with the arms in the weight support exoskeleton arm. Figure 2A illustrates the setting of the VR system.
Participants in the control group received conventional occupational therapy that consisted of task-related practice of gross movements and dexterity, including grip and selective finger movement, and activities of daily living. Training frequency was matched to those in the VR group.
2.8 ERP paradigm
The ERP experimental procedure and paradigm adopted in the present study was in accordance with a published study31. Participants were seated in front of a table in an electrically shielded laboratory with his/her shoulder flexed to 0 and 10°, elbows flexed to 130° and wrists orientated in a neutral position. The motion of palm opening and closing of the hand occurred in the horizontal plane. Figure 2B illustrates the staring position of the experiment. All of the participants were first fully explained on the procedure of the experiment. This was followed by 5 to 10 minutes of practice to familiarise with the protocol. Signal recording then began.
The EPR paradigm was as follow: A white fixation point “+” first appeared in the centre of the screen for 500 ms. Then, visual and auditory cues (S1) were given simultaneously for 2000 ms. A picture cueing grasp motion with either the left or right hand was displayed on the screen, accompanied by an auditory cue to indicate right or left side palmar grasp. During S1, the participants were required to judge the palmar grasp task. Then a grey reaction window (S2) of 3000 ms appeared, the participants performed left or right palmar grasp, and avoid making compensatory movements. Then a dark screen of 2000 ms appeared, the participants restored their fingers, and then entered the next trial. The experiment consisted of 40 trials for each hand, totalling 80 trials. The order of the trials were randomised. A diagram of the ERP paradigm can be found in previous published study 31.
EEG activities were recorded by a 32-channels QuickAmp amplifier and Ag/AgCl scalp electrodes (BrainProducts, Germany). The electrodes were positioned in accordance with the international 10-20 system. Electrodes were filled with conductive gel to maintain the impedance below 5 kΩ EEG. EMG signals were recorded in DC mode and sampled at 1000 Hz synchronize with event markers.
Regional of interest
Six electrodes of interest related to motor function were extracted for EEG analysis (left hemisphere: F3 and C3, right hemisphere: F4 and C4, midline region: Fz and Cz). The F3, F4 and Fz lies over the pre-motor cortex and the C3, C4 and Cz lies over the primary motor cortex. Figure 2C illustrates the locations of the electrodes. For the purpose of statistical comparison, the left and right side hemispheres were flipped right to left in participant with right hemispheric lesion so that the “left” hemisphere was always the lesioned hemisphere22.
The EMG activities were simultaneously recorded with EEG. Two surface electrodes were placed along the extensor digitorum using 2 surface electrodes with a 2-cm inter-electrode distance. The EMG reaction time was measured with the time from S1 to the EMG onset as shown in Figure 3A.
2.11 Signal processing
EEG signals were new referenced to the bilateral mastoid. Eye movement artefacts were removed through the Ocular Correction Independent Component Analysis (ICA) as part of standard operating procedure 14. EEG and EMG signals were filtered using a 50 Hz notch filter and a bandpass filter from 0.1 to 30 Hz. In order to acquire stimulus-locked ERPs, the EEG and EMG signals were segmented into epochs of 500 ms pre to 3000 ms post aligned to S1. The baseline was corrected according to the first 200 ms of the epochs, which was the 200 ms time window before S1 onset.
The EMG reaction time was measured with the time from S1 to the EMG onset. As the maximum CNV amplitude was detected between 1300 ms to 1800 ms after S1, peak detection was used to detect the maximum CNV amplitude in this time window. The maximum amplitude must be higher than the amplitude of 2-3 points on the front and back sides40. The CNV latency was calculated as the time from S1 to the maximum amplitude. The mean CNV amplitude was calculated from 1300 to 1800 ms to acquire the average amplitude of the CNV potential (Figure 3B).
2.11 Statistical analysis
Statistical analyses were performed using SPSS version 22 (IBM SPSS statistics version 20, USA). The Wilcoxon signed-rank test was applied to verify the statistical significance of the changes in the ARAT, FMA-UL, and NIHSS within groups, and Mann-Whitney U test was conducted to compare between the two groups. For EMG reaction time and CNV latency analyses, three-way mixed analysis of variance (ANOVA) were calculated, with task (paretic hand movement vs. non-paretic hand movement) and TIME (baseline vs. post-training) as within-subject factors, GROUP (VR group vs. control group) as between-subject factor. Four-way mixed ANOVA, with TASK (paretic hand movement vs. non-paretic hand movement), ELECTRODE (F3, C3, F4, C4, Fz, and Cz) and TIME (baseline vs. post-training) as within-subject factor, TREATMENT (VR group vs. control group) as between-subject factor were performed to assess the training effects. The Greenhouse-Geisser adjustment was applied to adjust the degrees of freedom if the assumption of the Mauchly’s test of sphericity was not significant. Separate mixed model ANOVAs were tested for each level of electrodes. If there was a significant interaction in the within-subject factor and between-subject factor, then subsequent independent sample t-tests were performed to further investigate the differences between the two groups. The significance level for all statistical analysis was set at 0.05. Bonferroni-adjusted significance tests were performed to correct the p-values of electrodes for multiple comparisons. Thus, the corrected significance level for electrode was α= 0.05 ÷ 6 = 0.00841.