Virtual Reality Training Dose for Arm Recovery After Stroke: Clinical and Blood Biomarker Responses

Background Upper-limb motor impairment persists in many people after stroke. Virtual reality (VR)-based training added to conventional rehabilitation can improve upper-limb function, but the amount (dose) of VR training needed and the biological processes that accompany recovery are unclear.

Design Prospective, single-center, randomized (1:1:1), assessor-blinded, three-arm, parallel-group dose-response trial in 72 adults 6 to 24 months after ischemic or hemorrhagic stroke with moderate-to-severe upper-limb motor impairment (Fugl-Meyer Assessment for Upper Extremity [FMA-UE] motor score 15 to 50).

Randomization and blinding A computer-generated block randomization list will be prepared by an independent researcher who is not involved in recruitment, treatment or assessment; allocation will be concealed in sealed opaque envelopes. Participants and treating clinicians cannot be blinded. The outcome assessor, the laboratory staff performing the biomarker assays and the statistician will be unaware of group allocation.

Interventions (4 weeks) All participants receive conventional rehabilitation for 45 minutes per session, 5 days per week (neurodevelopmental approaches, functional upper-limb exercises, task-oriented training, stretching and strengthening). The low-dose VR group additionally receives 30-minute Neuroforma VR sessions 2 days per week (8 sessions), and the high-dose VR group 5 days per week (20 sessions). Scheduled therapy time is therefore 225, 285 and 375 minutes per week in the control, low-dose and high-dose groups. From week 4 to week 12, all participants follow the same standardized home exercise program. Adherence and adverse events, including cybersickness, are recorded daily during the intervention.

Assessments At baseline, week 4 (end of the intervention) and week 12 (8 weeks after the end of the intervention), the blinded assessor administers the FMA-UE, the Wolf Motor Function Test (WMFT) and the Action Research Arm Test (ARAT). Fasting venous blood is drawn between 08:00 and 10:00 at the same visits; serum is separated by centrifugation, stored at -80 °C and analyzed in batches with commercial ELISA kits for BDNF, irisin (FNDC5) and S100B.

Statistical analysis Analyses follow the intention-to-treat principle. The primary outcome, change in FMA-UE from baseline to week 12, is compared between groups with a linear mixed-effects model for repeated measures (weeks 4 and 12) adjusted for baseline FMA-UE; pairwise group comparisons are Bonferroni-adjusted. Secondary clinical outcomes are analyzed in the same way. The sample size of 36 required for the primary outcome (repeated-measures ANOVA, group-by-time interaction, f = 0.25, alpha = 0.05, power = 0.80, correlation among repeated measures = 0.50) was increased to 72 to allow pairwise dose comparisons and exploratory biomarker analyses and to cover an expected attrition of 20%.

Exploratory analyses Biomarker outcomes are exploratory; no adjustment for multiple comparisons will be made, and results will be interpreted as hypothesis-generating. Associations between changes in clinical scores and biomarker levels will be examined with correlation analysis. ROC analysis will assess whether biomarker changes identify participants who reach a clinically meaningful FMA-UE improvement (at least 5.25 points). A classification and regression tree (CART) model using baseline and week-4 clinical and biomarker data will be developed to predict treatment response and internally validated with 10-fold cross-validation; external validation is not part of this study. A model-based cost analysis will estimate the potential savings of response-guided treatment duration compared with a fixed 4-week program. These analyses are hypothesis-generating.

Funding Health Institutes of Türkiye (TÜSEB), project number 51841.