At-Home BCI Rehabilitation Therapy for Chronic Upper Extremity Deficit After Stroke (BCI-REHAB): A Randomized Trial

Abstract

BACKGROUND:

At-home stroke rehabilitation methods have strong potential to supplement limited clinical resources, but realizing that potential for survivors in the chronic phase of recovery requires developing and validating more effective options than the home exercise programs typically used today. BCI-REHAB (At-Home BCI Rehabilitation Therapy for Chronic Upper Extremity Deficit After Stroke) compared an at-home brain-computer interface (BCI) therapy system (IpsiHand System) versus an at-home exercise program for improving upper extremity function in patients with chronic hemiparetic stroke.

METHODS:

Participants aged 18 to 85 years with a history of stroke ≥6 months before enrollment and right or left upper extremity paresis or plegia were screened and randomly assigned to either an at-home exercise program or to a BCI electroencephalogram system coupled to a range-of-motion assist handpiece. Participants completed 12 weeks of at-home therapy (5 sessions per week). The primary outcome was the change in the Upper Extremity Fugl-Meyer Assessment from baseline to 12 weeks.

RESULTS:

Overall, 109 participants were assessed for eligibility; 85 met the inclusion criteria and were randomized (43 intervention, 42 control). Of the 42 control participants, n=17 declined to participate further due to dissatisfaction with study arm assignment. A total of 62 participants were analyzed for the primary outcome (37 intervention, 25 control). The primary outcome (mean change) was significantly higher in the BCI group (6.0 [95% CI, 3.9–8.1]; P<0.0001) versus the control group (1.5 [95% CI, −0.1 to 3.1]; P=0.07), with an estimated treatment difference of 4.5 ([95% CI, 1.9–7.1]; P=0.0007). Participants who received the BCI intervention showed an estimated response rate of 55.5% (95% CI, 33.7%–77.2%), compared with 9.6% (95% CI, −2.9% to 22.1%) in the control group. This corresponds to an absolute increase in response of 45.8% ([95% CI, 20.9%–70.8%]; P=0.0003), yielding a number needed to treat of 2.2.

CONCLUSIONS:

At-home BCI therapy provides clinically meaningful improvement in upper extremity function for chronic stroke survivors compared with standard at-home exercise programs.

REGISTRATION:

URL: https://www.clinicaltrials.gov; Unique identifier: NCT05965713.

CLINICAL PERSPECTIVE.

What Is New?

• In the first randomized controlled trial of at-home, noninvasive brain-computer interface therapy, arm function improved significantly more with brain-computer interface than with a standard home exercise program (Fugl-Meyer treatment difference, 4.5 points), with a clinically meaningful response in 55.5% of patients versus 9.6% of controls (number needed to treat, 2.2), and no serious adverse events.

What Are the Clinical Implications?

• Noninvasive brain-computer interface therapy can produce clinically meaningful upper extremity recovery in chronic stroke survivors, beyond the window for spontaneous recovery.

• Because it is delivered at home without surgical risk, brain-computer interface–based therapy is a scalable option that may extend rehabilitation to survivors who cannot access in-clinic care.

Stroke is a leading cause of death and disability in the United States, with >795 000 strokes in the United States^1–3 and an estimated 12 million globally each year.⁴ Chronic upper-limb impairment of the arm and hand persists in ≈40% of survivors^5,6 and significantly impacts their daily living, quality of life, and the overall hope of recovery.⁷ Functional rehabilitation options are limited once survivors enter the chronic phase, and most conventional techniques are generally ineffective or plateau beyond 3 to 6 months poststroke.^5,8–11 Historically, stroke survivors have faced significant barriers to accessing advanced forms of therapy to improve their outcomes.^8,9,12–14 This is especially true in vulnerable minority and rural populations.¹⁵ There remains a critical unmet need for effective therapeutic options, particularly those that are accessible at-home and in rural populations.^16,17

Initial brain-computer interface (BCI) chronic stroke rehabilitation research hypothesized that BCI systems could measure movement-related signals from the central nervous system and provide meaningful feedback to direct plasticity.^18–20 BCI-enhanced cortico-sensory coupling is thought to induce Hebbian learning by involving simultaneous activation of presynaptic and postsynaptic neurons.^18,19 In the study by Humphries et al,²¹ magnetic resonance imaging findings suggested that contralesional BCI-induced motor function improvements in these populations may be partially driven by widespread decreases in motor network functional connectivity, potentially affecting inhibitory circuit activity through experience-dependent plasticity. In addition, an increase in theta–gamma coupling, mostly located in the hand regions of the primary motor cortex on the left and right cerebral hemispheres, positively correlated with motor recovery over the course of rehabilitation after BCI therapy.²² However, the precise synaptic and circuit mechanisms resulting in durable recovery, specifically after stroke, still require future studies.^23,24

Prior studies in chronic stroke have demonstrated motor recovery benefits from BCI-based interventions that couple brain activity to neuromuscular electrical stimulation or motion-assist orthoses.^25–27 Importantly, Bundy et al²⁰ demonstrated reliable BCI control in patients with hemiparesis based on motor signals from the contralesional hemisphere, ipsilateral to the paretic hand, establishing a feasible and consistent control signal source for therapeutic BCI in hemiparetic populations. However, the clinical relevance of this approach for addressing unmet rehabilitation needs depends on demonstrating effectiveness and feasibility in at-home therapy settings.

Subsequent research has supported the utility of BCIs in patient-centric home environments. At-home BCI therapy with a motion-assist handpiece demonstrated a statistically significant increase of 6.2 points in the Action Research Arm Test in 10 chronic hemiparetic stroke survivors with moderate-to-severe upper-limb motor impairment.²⁸ Rustamov et al²⁹ conducted a pooled analysis of 2 prospective clinical trials (https://www.clinicaltrials.gov; Unique identifiers: NCT04338971 and NCT03611855) evaluating safety and efficacy in 26 adults after 12 weeks of at-home BCI therapy; all 26 participants demonstrated motor improvement. The mean Upper Extremity Fugl-Meyer (UEFM) score increased by 8.1 points (P<0.001), exceeding the minimal clinically important difference (MCID) of 5.25 and indicating a functionally meaningful change in motor performance.³⁰ This BCI system, which uses intentional brain activity to drive impaired upper extremity movement for at-home stroke rehabilitation, received Food and Drug Administration breakthrough device designation and de novo market authorization.³¹

BCI-REHAB (At-Home BCI Rehabilitation Therapy for Chronic Upper Extremity Deficit After Stroke) was a randomized clinical trial comparing an at-home BCI therapy system versus an at-home exercise program for upper extremity rehabilitation in the chronic phases of stroke. The purpose of the trial was to evaluate the efficacy of BCI-based therapy tested against standard exercise therapy and assess the potential for meaningful motor improvement in the home environment.

Methods

Data Availability Statement

Supplemental Material is available with this article and upon written request per National Institutes of Health guidelines.

Study Design

This randomized controlled trial was conducted at Washington University in St. Louis and affiliates. The study was approved by the WIRB-Copernicus Group, Inc (WCG IRB) and subject to appropriate regulatory approvals (US Food and Drug Administration Investigational Device Classification: DEN200046, Class II). The study was conducted according to the Declaration of Helsinki and written informed consent was obtained from all participants. CONSORT 2025 guidelines (Consolidated Standards of Reporting Trials) for randomized controlled trials were followed. There was no patient or public involvement in the design, conduct, and reporting of the trial.

Inclusion criteria were adults between 18 and 85 years, with a history of stroke ≥6 months before enrollment and upper extremity paresis/plegia. Exclusion criteria were participants too cognitively impaired to understand tasks or provide informed consent, receiving any formal upper extremity therapy, who had contractures in the affected wrist and digits, who had receptive aphasia, and unilateral visual inattention. In addition to inclusion and exclusion criteria, participants underwent an electroencephalogram (EEG) test before randomization as described below.

EEG Signal Test

After meeting the inclusion criteria, all participants underwent an EEG evaluation to ensure that EEG features sufficient for device control could be detected. To conduct the test, trained occupational therapists (OTs) placed a commercial EEG system (Wearable Sensing, Inc, DSI-7, 10–20 standard electrode locations P3, P4, C3, C4, F3, F4, Pz, and Fz) on the participant’s head. Each participant completed a minimum of 1 and up to 3 signal intent qualifications where EEG signals were collected while patients performed a visually cued motor qualification task. Regardless of existing motor function level, participants were instructed to imagine movements of the affected hand. Spectral power changes were measured by the EEG and validated by a trained specialist to ensure that the subject had neural activity present in the unaffected hemisphere, allowing for detection of intended movement of the affected hand using an r² analysis. Further details regarding the EEG test process have been published previously.^22,29 If no control signal was identified within 3 signal intent qualification attempts, the subject was excluded due to EEG test failure. As part of the EEG test, participants also completed cognitive and motion-related assessments. Participants without aphasia who scored an ≥8 on the Short Blessed Test were excluded due to indication of cognitive impairment. For users with aphasia who could not complete the Short Blessed Test, the Mississippi Aphasia Screening Test was administered. Participants who scored a 6 or lower were excluded, as this score indicates inability to understand verbal and written instructions. In addition, all participants were screened for inattention/neglect using the Mesulum Cancellation Test (unstructured). Those who omitted ≥6 targets were excluded. Active and passive range of motion measurements and a Modified Ashworth Scale score of ≤3 were completed for upper extremities to confirm the patient had range of motion to comfortably fit and wear the IpsiHand handpiece.

Randomization and Masking

Participants who met the inclusion/exclusion criteria and passed the EEG test were randomly assigned (1:1) to either home-based BCI rehabilitation or a home exercise rehabilitation regimen program (HEP). Randomization used the Web-based National Institutes of Health Clinical Trial Randomization Tool, which used maximally tolerated imbalance randomization with permuted blocks. The randomization sequence was generated within the platform, and study personnel had no access to upcoming assignments. No clinical or demographic characteristics were used in determining allocation. The randomization allocation was communicated to a local OT by the unblinded study coordinator.

Procedures

After the EEG test, participants randomized into the intervention arm completed an in-person baseline assessment of function status and a training session with the BCI system (IpsiHand System, Neurolutions, Los Angeles, CA) shown in Figure 1. All training and assessments were completed by licensed OTs who were not blinded to arm assignment. Participants were instructed to use the device for home-based therapy sessions 1 hour per day, 5 days per week, for a continuous 12 weeks. During the therapy sessions, the system would guide the patient through a series of routines in which they would imagine opening and closing their affected hand. At the instant the BCI detected sensorimotor cortex EEG spectral features associated with imagined ipsilateral hand movements (continuously recorded from C3 and C4 and based on the control signal determined in the EEG signal intent qualification), the handpiece opened and closed the participant’s hand in a 3-finger pinch grip as described in prior publications.^20,22,28,29 Patients did not receive occupational or physical therapy services while participating in the self-directed BCI Therapy. At week 6, participants were contacted to set up their 12-week completion visit.

Figure 1.

Flow diagram of brain-computer interface (BCI) therapy system starting with motor intention and signal decoding so that user intent can be translated to the handpiece. The handpiece end effector opens and closes the fingers in a 3-finger pinch grip based on user intention. EEG indicates electroencephalogram.

Participants allocated to the HEP control group received a structured home exercise program designed to control for nonspecific motor and sensory engagement associated with the intervention. They completed an in-person baseline assessment of function and an HEP training session with an OT. The program consisted of a customized set of range of motion and functional tasks targeting the hand, wrist, elbow, and shoulder. Participants were instructed to complete 1 hour of home exercise per day, 5 days per week, over the course of 12 weeks. Exercises included passive range of motion, stretching, and a series of standardized functional activities of daily living such as wiping a table, washing hands, pulling a chair away from a table, opening and closing drawers, holding a utensil and bringing it to the mouth, lifting a cup, bringing a phone-like object to the ear, and turning on or off light switches. Each participant received a daily log to record adherence to the prescribed regimen and to document the dates on which exercises were completed. At week 6, participants were contacted to set up their 12-week completion visit. For both arms, when applicable, botox injection date was monitored so that it coincided within a week of the participants’ baseline start date. Mid-trial, all participants randomized into the HEP were offered access to the commercial IpsiHand System after study completion as an incentive.

Outcome assessments were completed on day 1 (baseline) and after 12 weeks of participation. There were no interim clinical appointments in either arm of the study. All OTs conducting assessments had a minimum of 5 years of direct clinical experience working with adults after neurological stroke and completed the formal Fugl-Meyer Assessment certification course developed at The University of California, Los Angeles.³² Assessments were completed in-person by a licensed OT and included: UEFM (primary), Manual Long Form Change from baseline to 12 weeks in UEFM total score, Modified Ashworth Scale score, Timed Up and Go Test, Trail Making Test Part A and Part B, modified Rankin Scale, Gross Grasp Force (Dynamometer), and Western Aphasia Battery (Bedside Record Form). In-person UEFM assessments were filmed to allow for a blinded third-party review and assessment of bias.

Compliance in the BCI arm was assessed remotely using device-recorded data including date, time, duration, and number of repetitions. HEP compliance was recorded by the participant and attested to at completion. All UEFM assessments were video recorded. All adverse events were collected and monitored systematically throughout the study by the sponsor.

Outcomes

The primary outcome was change in UEFM score from baseline to the 12-week completion assessment. A clinically meaningful response was prospectively defined as a ≥5.25 gain from baseline on the UEFM assessment.³⁰ Secondary end points included Modified Ashworth Scale (Elbow Flexion), Timed Up and Go Test, Trail Making Test Part A and Part B, modified Rankin Scale, Gross Grasp Force (Dynamometer), and Western Aphasia Battery (Bedside Record Form).

Statistical Analysis

The trial was designed based on a primary estimand defined with the end point of change from baseline to 12 weeks in UEFM total score and a population-level summary being the difference in least square means of the change associated with the model described below. The sole predefined intercurrent event was device use <80% of the duration described by the protocol, which applies only to the BCI intervention group. Of primary clinical interest was comparison of device used as intended versus no device, and thus the intercurrent event was handled using the hypothetical strategy³³ to reflect the benefit in a future population under ideal device conditions, when initiating and continuing treatment as planned. An additional estimand, reflecting the expected benefit of the study intervention in a target population irrespective of observed intercurrent events, handled intercurrent events using a treatment policy strategy. Estimands for secondary end points were defined similarly.

A group sequential approach, with a single interim analysis, was preplanned in the statistical analysis plan. The interim analysis was prespecified to occur after two-thirds of the enrolled population completed primary outcome assessment at week 12. The Lan-DeMets implementation of the O’Brien-Fleming stopping boundary was used in the statistical assessment of interim efficacy results, conducted with a 2.5% (1-sided) nominal significance level.³⁴ The study sample size was based on meeting minimum power requirements, defined as 80% power, for analysis of the primary estimand. For the required pooled SD, a value of 5 was assumed, and a minimal clinically meaningful difference of 4.25 was used. Minimal clinically meaningful difference was defined as the difference between MCID and historical control arms.^30,35 Under these assumptions, calculations showed that a planned sample size of up to 23 participants per group (46 total with nonmissing 12-week UEFM total score) was required. To account for a potential 25% dropout rate, a total of up to 62 participants were targeted for enrollment.

Measured variables were summarized using descriptive statistics. To describe the observed variability in the primary outcome and perform inferences on differences between randomized groups, statistical analyses were based on an ANCOVA model, where change in UEFM total score was modeled as a linear function of independent variables representing randomized group assignment and baseline UEFM total score. Standard diagnostic plots were used to assess model fit, which was found to be adequate. Once the model was fit, group differences based on least-squares means were computed. As the primary estimand used the hypothetical strategy to handle intercurrent events, only postbaseline data for those who did not experience the predefined intercurrent event were used, resulting in a missing outcome value at 12 weeks for those who did experience the intercurrent event. Multiple imputation methods were then used under missing-at-random assumptions, using group-specific regression models with baseline UEFM total score as the sole predictor, with the imputed value used in the calculation of within-participant change. A total of 500 complete data sets were generated, and Rubin’s rules³⁶ were applied to combine the results and construct the group estimates using the parameter estimates and associated standard errors. For the estimand that used the treatment policy strategy, a similar approach was used in which missing values were imputed using the baseline-observation-carry-forward–like procedure discussed in O’Kelly and Ratitch.³⁷ Participant-level covariates known to be predictive of outcome—age and cerebrovascular accident lesion location—were added to the above-described model, with the group effect reassessed as a series of supportive analyses. Responder analyses based on a predefined clinically significant improvement in UEFM total score of 5.25 points were conducted based on a similar imputation strategy. Analyses of secondary outcomes proceeded similarly as the primary analysis.

As the primary outcome, UEFM total score, was assessed by an unblinded evaluator, interrater agreement was evaluated in a subset of participants for whom blinded evaluators independently scored the UEFM total score based on video-recorded assessments. Agreement between unblinded and blinded UEFM total scores was assessed using Bland-Altman analyses, conducted separately at baseline, at study completion, and for the change score. For each analysis, the mean difference (bias), defined as the unblinded score minus the blinded score, and the corresponding 95% limits of agreement were estimated. Bland-Altman analyses were also stratified by randomized treatment group at study completion to evaluate potential differential bias by group. All analyses were implemented using the SAS statistical software package (SAS Institute, Inc, Cary, NC), version 9.4.

Role of the Funding Source

The study was funded by the National Institutes of Health as part of SBIR award R44HD105579 and Neurolutions, Inc. The funder (Neurolutions, Inc) contributed to the study design and provided administrative support but had no role in data collection, analysis, interpretation, or the decision to submit the manuscript for publication.

Results

Of the 109 participants assessed for eligibility, 85 met the study inclusion criteria and were randomized between October 2, 2023 and August 31, 2025 (43 assigned to intervention, 42 assigned to control; Figure 2). The most common reasons for ineligibility were an inability to pass the EEG test signal qualification session (9.2%, n=10), Unilateral Visual Inattention (6.4%, n=7), and a Modified Ashworth Scale score of the affected upper extremity of ≥3 (3.7%, n=4). After randomization, 21 participants withdrew from the study before baseline, and 2 enrollments were excluded from analysis due to protocol deviations. Seventeen of the 21 participants declined to participate further due to dissatisfaction with assignment into the HEP control arm. The prespecified interim analysis was conducted with complete data sets at a sample size of 17 HEP and 28 BCI participants by an unblinded, third-party, statistician. At the time of the interim analysis, the unblinded statistician made the recommendation to end the trial early for efficacy. After the interim analysis, further enrollment was ended due to lack of clinical equipoise, and the remaining enrolled participants completed treatment. A total of 62 participants were analyzed for the primary outcome (25 HEP and 37 BCI). Ten of the participants included in the analysis completed a baseline assessment but withdrew before 12-week completion. Demographics are shown in Table 1, and baseline characteristics are shown in Table 2.

Figure 2.

Consort diagram/trial profile. BCI indicates brain-computer interface; EEG, electroencephalogram; and HEP, home exercise rehabilitation regimen program.

Table 1.

Demographics

Table 2.

Baseline Characteristics

The primary outcome, change in UEFM from baseline to 12 weeks, was significantly higher in the BCI group 6.0 ([95% CI, 3.9–8.1]; P<0.0001), than HEP control group 1.5 ([95% CI, −0.1 to 3.1]; P=0.07; Table 3; Figure 3). Treatment differences were estimated to be 4.5 ([95% CI, 1.9–7.1]; P=0.0007). Participants who received the BCI intervention showed an estimated clinically meaningful response rate of 55.5% (95% CI, 33.7%–77.2%), compared with 9.6% (95% CI, −2.9% to 22.1%) in the control group. This represents an absolute improvement in response of 45.8% (95% CI, 20.9%–70.8%), which was statistically significant (P=0.0003), resulting in a number needed to treat of 2.2. When statistically adjusting for age and cerebrovascular accident lesion location, results persisted. The treatment policy estimand analysis, assuming missing values were not missing at random and imputed using the baseline-observation-carried- procedure, still demonstrated a significant treatment difference (BCI-HEP) of 2.5 ([95% CI, 0.3–4.6]; P=0.024).

Table 3.

Primary Outcome Summary Statistics—All Recorded Data

Figure 3.

Response and change in Upper Extremity Fugl-Meyer (UEFM) score (primary end point). A, Measured UEFM scores at baseline and week 12 completion. B, Change in measured UEFM score at baseline and week 12 completion. The primary end point is denoted with an X; brain-computer interface (BCI) group 6.0 ([95% CI, 3.9–8.1]; P<0.0001) and home exercise rehabilitation regimen program (HEP) group 1.5 ([95% CI, −0.1 to 3.1]; P=0.07). C, Estimated mean and SE based on primary estimand and multiple imputation analysis.

Comparison of unblinded, in-person UEFM assessments to blinded UEFM done over video review showed a mean bias of −2.4 (95% CI, −3.2 to −1.5) and limits of agreement of (−8.3 to 3.6) for 51 comparable participants evaluated at baseline and −2.4 (95% CI, −3.6 to −1.3) and limits of agreement of (−10.4 to 5.5) for 49 comparable participants at study completion. The estimated bias in the assessment of UEFM change from baseline to completion was −0.04 (95% CI, −1.1 to 1.0) and limits of agreement of (−6.8 to 6.7).

Secondary end points including Modified Ashworth Scale (Elbow Flexion), Timed Up and Go Test, Trail Making Test Part A, modified Rankin Scale, Gross Grasp Force (Dynamometer) and Western Aphasia Battery (Bedside Record Form) did not differ significantly between intervention and control groups (all P>0.05). The Trails Making Test Part B time was higher in the BCI arm than HEP (BCI-HEP) 30.5 s ([95% CI, 3.5–57.4]; P=0.027). Full secondary outcomes are included in the Supplemental Material.

There were no serious or general adverse events reported in the study. There were no significant protocol deviations that affected the rights, safety, or well-being of participants. There were 3 inclusion criteria deviations noted postrandomization in which the participants were removed from the analysis data set. Additional deviations related to informed consent issues and missed visits were minor and did not affect the per protocol analysis related to the scientific integrity of the study.

Discussion

BCI-REHAB is the first randomized clinical trial to study at-home, noninvasive, BCI therapy for upper extremity deficit after chronic stroke. Participants assigned to the BCI intervention arm experienced clinically meaningful improvements in motor function compared with participants assigned to an at-home exercise rehabilitation program. More than half of the participants assigned to BCI therapy achieved a clinically meaningful response after 12 weeks of therapy, with no reported serious or general adverse events. Outcomes for both the BCI treatment and control groups were consistent with prior published outcomes.^29,35 Taken together, this study supports that BCI rehabilitation is a reasonable treatment option for patients with chronic stroke with upper extremity paresis or plegia.

Several aspects of these results make them impactful for patients with stroke. Although the trial inclusion criteria allowed enrollment as early as 6 months after stroke, the mean time from date of stroke for enrolled participants was 5.4 years, with participants in the BCI arm ranging from 6 months to 18 years poststroke. The results confirm that it is possible to achieve meaningful improvements in the chronic stage after a stroke, with improvements unlikely to be attributable to spontaneous recovery typically seen in the early postacute period. Notably, in long-term chronic stroke survivors, there is often a decrement in function over time.³⁸

Further, these gains achieved with noninvasive BCI therapy were accomplished in the patients’ homes. This is important for 2 reasons. First, these findings confirm that BCIs can operate in relatively uncontrolled, nonmonitored patient-centric environments to record meaningful brain signals and perform consistently to achieve clinically relevant results. Second, the system is used entirely in the home environment, which overcomes patients’ reticence to visit exercise and rehabilitation centers. Fatigue and distance from exercise facilities are among the main reported barriers to exercise in patients with chronic stroke, along with lack of a person to help and knowledge on how to practice exercise.⁷ The final notable aspect is that this intervention is noninvasive. Improvement can be accomplished with minimal risk and does not require invasive surgical treatment such as implantation of a vagus nerve stimulator.^39,40

Although the home-based setting is certainly an advantage, this study also revealed patient sentiment about standard stroke rehabilitation approaches in the home. Community perception and willingness to complete home-based exercise therapy for chronic stroke upper extremity rehabilitation is poor.⁷ This was evidenced here, as participants randomized into the HEP declined to participate further due to dissatisfaction with not being assigned to the active device intervention arm and experienced a high withdrawal rate. Mid-trial, all participants randomized into the HEP were offered access to the commercial system after study completion as an incentive. This subsequently improved control arm retention at the point of randomization. This outlines one of the major difficulties in randomized trials studying chronic stroke rehabilitation, namely that survivors have given up on traditional exercise and will refuse to participate if not offered another option for recovery. The chronic stroke community and their caregivers already perceive a lack of clinical equipoise regarding home-based exercise programs.

The findings from this study position BCI therapy as a valuable component within a comprehensive, tiered rehabilitation strategy for chronic stroke survivors with persistent upper extremity deficits. In the chronic phase, optimal care could include current standards, including physical therapy and constraint-induced movement therapies, when patients retain sufficient residual function. Even when these conventional approaches have been exhausted or are contraindicated due to severe impairment, BCI therapy emerges as a promising next-line intervention that can meaningfully improve motor function without surgical risk. Importantly, these therapeutic modalities should be viewed as synergistic rather than competing approaches. For instance, vagus nerve stimulation therapy for stroke requires patients to meet specific UEFM assessment thresholds for candidacy, yet many chronic stroke survivors fall below these functional requirements.⁴¹ BCI therapy’s demonstrated ability to produce clinically meaningful improvements in UEFM scores positions it as a potential bridge intervention that could elevate patients’ functional status to enable vagus nerve stimulation candidacy. This sequential, building-block approach allows each therapy to optimize the foundation for subsequent interventions, ultimately expanding treatment options and hope for recovery in a population that has historically been considered to have reached their rehabilitation plateau.

Limitations

This study had several important limitations that may affect the generalizability of findings. A key limitation was the lack of a sham control and resultant lack of blinding for both participants and assessors, with all parties being aware of assignment to BCI or HEP arms of the trial. This could have resulted in bias in the assessment of the primary end point, although analysis of study outcomes compared with a blinded video reviewer showed agreement.

Baseline UEFM score was not used as an inclusion or exclusion criterion, allowing participants across the full range of scores (0–66). Although this allowed for broad applicability of the results, it meant that participants with baseline UEFM scores >60 could not achieve the MCID of 5.25 points, potentially underestimating treatment effects in higher-functioning participants. 1 participant in the BCI arm of the study had a baseline UEFM of 61, allowing for a maximum possible improvement of 5.

In addition, the 12-week intervention period may have been insufficient to capture the full therapeutic potential of the BCI system, as recent evidence demonstrates that stroke survivors continue to show meaningful improvements beyond 12 weeks of use, with some participants requiring longer durations to achieve MCID thresholds. Although the ability to achieve meaningful improvement in 12 weeks is highly encouraging, the study does not capture the full possibility of recovery that may come with extended use beyond 12 weeks, and future, longer duration, studies will be conducted. We note that recently published real-world evidence suggests continued improvement beyond 12 weeks.⁴²

The study experienced uneven participation between intervention and control groups due to community perceptions regarding home exercise programs and participant dissatisfaction with control group assignment, which may have introduced bias in the results. This trial demonstrated that offering the therapy after control home exercise may improve retention in home exercise control groups perceived to be ineffective. In addition, there were no interim clinical appointments between baseline assessment and 12-week study completion. Although this demonstrated what is possible for patients in an entirely unsupervised environment, it also limited the ability to improve patient adherence and engagement.

The study was powered to assess UEFM change. Among the prespecified secondary end points, only one (Trails Making Test Part B) reached statistical significance, favoring the HEP arm. The trial was powered for the primary end point; the secondary end points were exploratory and individually underpowered to detect between-group differences, with no adjustment for multiple comparisons. The wide 95% CI of Trails Making Test Part B (3.55–57.38 s) indicates substantial uncertainty about the true effect size. In addition, the closely related Trails Making Test Part A showed no significant between-group difference, which supports interpreting the Trails Making Test Part B result as hypothesis-generating rather than a contradiction of the significant primary end point. In future work, we plan to expand data collection to more comprehensively evaluate secondary end points that may capture clinically meaningful changes beyond the primary outcome. Additional real-world evidence studies will assess impact, including activities of daily living, quality of life, and caregiver burden.

Finally, although chronic stroke survivors often face significant barriers to accessing advanced rehabilitation therapies—particularly among minority and rural populations—this study did not specifically examine whether home-based BCI therapy improved access for these underserved groups, limiting our understanding of the technology’s potential to address existing healthcare disparities in stroke rehabilitation.

Conclusions

Chronic stroke survivors, averaging over 5 years poststroke, demonstrated statistically significant and clinically meaningful improvements in upper extremity function after 12 weeks of home-based BCI therapy, exceeding MCID more often than through a conventional home exercise program alone. These findings establish at-home BCI therapy as a promising, noninvasive rehabilitation intervention that can meaningfully restore hand and arm function in individuals with persistent motor deficits long after the initial stroke event.

ARTICLE INFORMATION

Acknowledgments

The authors thank participants for their contribution to this research project. The authors also acknowledge Washington University School of Medicine and Neurohub for institutional support, the clinicians (G. Kopp, L. Leonard, and K. Guetzloe) for their contributions, and the Veraxia team for their involvement.

Disclosures

Dr Wilk, L. Souders, K. Carr, Dr Coker, and Dr Leuthardt own stock and are employees of Neurolutions Inc. The other authors report no conflicts.

Supplemental Material

Figures S1–S4

Tables S1–S2

CONSORT Checklist

Funding Statement

This work was supported by National Institutes of Health (R44HD105579) and Neurolutions Inc.