The impact of non-invasive brain-computer interface technology on the therapeutic effect of patients with spinal cord injury: a summary of evidence based on meta-analysis.

Only 109 participants across nine studies, so every pooled estimate is built on a handful of people. Five of nine studies were self-controlled rather than randomised, mixing designs weakens the conclusion. Only one study reported allocation concealment and blinding, so subjective outcome exaggeration is possible. No long-term follow-up: eight studies measured immediately after the intervention and only one followed up at 8 weeks, so durability is unknown. Sensory and daily-living effects were fragile, disappearing when the largest study or the random-effects model was used. Chinese and English literature only, so relevant work in other languages was excluded. Seven different motor scales were combined into one score, which makes the single effect size hard to translate into points on any one test.

The paper does not report a funding source or a conflict-of-interest declaration.

Do not read the significant pooled scores as proof that BCI works for spinal cord injury. The authors themselves say this is a research hypothesis, not a basis for clinical recommendations, because the sample is tiny, most trials were self-controlled, blinding was almost absent, and the sensory and daily-living results became non-significant under sensitivity analysis.