Brain-computer interfaces in poststroke rehabilitation: a meta-analysis of randomized clinical trials

OBJECTIVE Stroke is a leading cause of long-term disability, with conventional rehabilitation often failing to achieve substantial motor recovery, particularly in patients with severe paresis or in chronic stages. Brain-computer interfaces (BCIs) offer a novel rehabilitation approach by translating neural signals into real-time external feedback. The authors performed a systematic review and meta-analysis of randomized controlled trials (RCTs) to evaluate the efficacy and safety of noninvasive BCIs for poststroke motor rehabilitation. METHODS A systematic literature review was performed based on the PRISMA guidelines using 3 databases. Eligible RCTs enrolled stroke patients receiving noninvasive BCI-assisted motor rehabilitation compared with conventional therapies. The primary outcome was the Fugl-Meyer Assessment for Upper Extremity (FMA-UE) improvement. Secondary outcomes included the Action Research Arm Test (ARAT), Motor Activity Log (MAL), Modified Barthel Index (MBI), and Modified Ashworth Scale (MAS). Effect sizes were pooled using random-effects models and expressed as mean differences (MDs), standardized MDs (SMDs), or odds ratios, each with corresponding 95% confidence intervals (CIs). RESULTS Thirty-two RCTs comprising 1187 patients were included with no heterogeneity or significant imbalances in baseline characteristics across groups. A BCI was significantly superior in FMA-UE score improvement compared with controls (MD 3.85, 95% CI 2.84–4.86; p < 0.01), with benefits sustained at follow-up. Within-group analyses revealed greater improvement in the BCI arm from follow-up to baseline (MD 8.18, 95% CI 5.77–10.60; p < 0.01). A BCI was also associated with higher ARAT (MD 7.18, 95% CI 2.4–12.0; p < 0.01) and MAL (SMD 0.59, 95% CI 0.34–0.85; p < 0.01) scores, although between-group differences for these endpoints were not statistically significant. For the MBI, a subgroup analysis did not demonstrate significant differences, but a sensitivity analysis revealed a significant improvement in the BCI group (p = 0.042). There were no significant differences in the within- and between-group analyses of the MAS. A subgroup analysis suggested a synergistic benefit with the BCI combined with neuromuscular electrical stimulation. Adverse events were infrequent and generally mild; 2 withdrawals in the BCI group were reported due to seizure and electrode allergy. Notably, all heterogeneity was successfully resolved through sensitivity analyses, supporting the robustness of the findings. CONCLUSIONS Noninvasive BCI-assisted rehabilitation is a safe and effective adjunct to conventional therapy, enhancing motor recovery after stroke. While all included RCTs evaluated noninvasive systems, the potential value and efficacy of invasive and minimally invasive BCIs may require further consideration.