Comparative efficacy of virtual reality, robotics, and brain–computer interface interventions for upper limb rehabilitation after stroke: a systematic review and network meta-analysis
TL;DR
For improving upper-limb motor function after stroke, robotics-based interventions were supported by stronger evidence than other modalities, and no reliable conclusions can be drawn.
Abstract
Background Stroke often leads to persistent upper-limb motor impairment, which significantly impairs quality of life. Conventional physical therapy (CPT) has limitations, including insufficient intensity, limited patient engagement, and inadequate feedback. Emerging technologies such as virtual reality (VR), robotics (ROT), and brain–computer interfaces (BCI) have shown promise; however, direct comparisons among these approaches are lacking, and their relative effectiveness remains unclear. Objective This study aimed to systematically evaluate and compare the relative effectiveness of VR, robotics, and BCI on upper limb motor function, motor performance, and activities of daily living in stroke survivors using network meta-analysis. Methods PRISMA-NMA guidelines were followed. PubMed, Web of Science, Cochrane Library, and Embase were searched from inception to October 2025 for RCTs. Two reviewers independently screened studies, extracted data, and assessed risk of bias using RoB 2.0. A Bayesian random-effects network meta-analysis (R package gemtc) was performed to estimate relative treatment effects and calculate SUCRA values, along with sensitivity and subgroup analyses. Results 25 RCTs (1,145 stroke survivors) were included. The network evidence geometry was star-shaped, with conventional physical therapy (CPT) as the common comparator. For FMA-UE, ROT-RFE achieved the highest SUCRA ranking, although this estimate was based on a single study. ROT-CPT and ROT, supported by two and three studies respectively, provided more consistent evidence. For secondary outcomes, ROT-CPT ranked highest for MBI (SUCRA = 0.89), whereas VR-CPT ranked highest for WMFT (SUCRA = 0.59). Sensitivity analyses generally supported robustness, and subgroup analyses suggested patient characteristics may influence treatment effects. Conclusion For improving upper-limb motor function after stroke, robotics-based interventions were supported by stronger evidence than other modalities. Specifically, ROT (SUCRA = 0.71, 3 studies) and ROT-CPT (SUCRA = 0.70, 2 studies) demonstrated consistent and clinically meaningful improvements, representing more reliable options for clinical practice. While a single robotics variant (ROT-RFE, SUCRA = 0.91) achieved a numerically higher ranking, this estimate was based on one trial and should not be interpreted as definitive evidence of superiority. VR-based interventions showed modest benefits, whereas BCI-based interventions were supported by only one eligible study with extractable data and no reliable conclusions can be drawn. Systematic review registration https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD420251180631, identifier: CRD420251180631.