Different feedback modes of brain–computer interface training for upper limb motor function after stroke: a protocol for a systematic review and network meta-analysis

Abstract

Background Upper limb motor impairment after stroke is a major contributor to long-term disability and can have lasting effects on patients’ activities of daily living, social participation, and quality of life. Brain-computer interface (BCI) training establishes a closed-loop training process that links central nervous system activity, external feedback, and sensory reafferent input, and is considered a promising approach to facilitating neuroplastic reorganization and motor recovery. Existing studies on upper-limb rehabilitation after stroke have investigated several BCI feedback modes. However, the relative efficacy of these feedback modes has not been systematically compared across available randomized controlled trials. Methods This study protocol has been registered in PROSPERO with registration number CRD420261370474 and will be conducted in accordance with the PRISMA-P statement and the PRISMA-NMA extension. PubMed, Embase, Web of Science, the Cochrane Library, and Scopus will be systematically searched from inception to March 2026. Only parallel-group randomized controlled trials will be eligible for inclusion. Participants will be patients with post-stroke upper-limb motor impairment. The experimental interventions will consist of BCI training with different feedback modes, and intervention nodes will be defined according to specific feedback categories, including functional electrical stimulation, robot- or exoskeleton-assisted feedback, visual feedback, virtual reality feedback, and combined multimodal feedback. Control groups will include conventional rehabilitation, sham BCI, no additional intervention, and other active interventions. The primary outcome will be the Fugl-Meyer Assessment for Upper Extremity (FMA-UE) score. The secondary outcomes will be the Action Research Arm Test (ARAT) score and the Wolf Motor Function Test (WMFT) score. Two reviewers will independently perform study selection, data extraction, and risk-of-bias assessment, and any disagreements will be resolved through discussion or adjudication by a third reviewer. If the evidence network is sufficiently connected, a network meta-analysis will be performed to compare the relative efficacy across different intervention nodes. Local inconsistency will be assessed using the node-splitting method, and the surface under the cumulative ranking curve (SUCRA) will be used to rank different feedback modes according to efficacy. Conclusion This protocol specifies the methods for comparing different BCI feedback modes in post-stroke upper-limb rehabilitation and provides a transparent framework for the planned systematic review and network meta-analysis. Systematic review registration https://www.crd.york.ac.uk/prospero/, identifier CRD420261370474.