The research team, led by Cristian D. Guerrero-Mendez and colleagues at the University of Campinas in Brazil, searched seven scientific databases from their inception through November 2025, following PRISMA reporting guidelines and registering the review protocol in advance on PROSPERO. They sorted the eligible trials into four distinct categories of electromyography-based intervention: EMG-triggered electrical stimulation, in which a patient’s own muscle activity fires electrical pulses to the target muscle once an activation threshold is crossed; that same stimulation combined with other therapies such as mirror therapy, task-oriented training, or motor imagery; EMG-controlled robotic platforms that estimate motor intention to drive assistive devices; and EMG biofeedback, which converts muscle activity into real-time visual or auditory signals. Each category was then compared against conventional, non-EMG rehabilitation across outcomes organized by the World Health Organization’s International Classification of Functioning, Disability and Health framework.

The headline finding is a striking asymmetry. EMG-triggered electrical stimulation used in isolation showed no advantage whatsoever over conventional therapy. On the Fugl-Meyer Assessment for the upper extremity, one of the gold-standard measures of post-stroke motor recovery, control groups actually fared slightly better immediately after treatment, with a small standardized mean difference of −0.53 favoring conventional care. On the Action Research Arm Test, which measures real-world arm function, the two approaches were statistically indistinguishable at every time point examined, from the end of treatment out to follow-ups of three months or longer. For a technology that has been promoted for decades as a way to reawaken paralyzed muscles by linking voluntary effort to immediate stimulation, the message is blunt: the trigger alone is not the therapy.

But when EMG-triggered stimulation was woven into a broader rehabilitation program, the picture transformed dramatically. Combined interventions produced large improvements in upper-limb motor function, with a standardized mean difference of 1.53 on the Fugl-Meyer scale, and moderate-to-large gains in grip strength, wrist flexion strength, wrist range of motion in both flexion and extension, manual dexterity on the Box and Blocks Test, and the Manual Function Test. Crucially, the benefits extended beyond raw impairment measures into the domains that matter most to patients: combined therapy improved scores on the Barthel Index of daily living independence and on both subscales of the Motor Activity Log, a patient-reported measure of how much and how well the affected arm is actually used in everyday life. In the language of the ICF framework, the combined approach produced improvements across body functions and structures, activity, participation, and patient-reported outcomes simultaneously.

The robotic arm of the evidence told a different and equally instructive story. EMG-based robotic platforms, which read a patient’s motor intention from surface muscle signals and translate it into smooth, assisted movement of the wrist or fingers, did not outperform conventional therapy on broad functional measures such as the Fugl-Meyer Assessment or the Action Research Arm Test. Yet they excelled at one specific and clinically stubborn problem: spasticity, the pathological muscle stiffness that afflicts many stroke survivors. Robotic interventions produced large reductions in finger and wrist spasticity on the Modified Ashworth Scale immediately after treatment, and remarkably, the wrist benefits persisted at long-term follow-up of three months or more, with an effect size of −2.32. The authors suggest this is because EMG-driven robotic assistance encourages active patient engagement during movement intention and execution, reinforcing the sensorimotor coupling thought to underlie neural plasticity.

EMG biofeedback, the fourth category, remains the most uncertain. Only a handful of eligible trials existed, permitting just two meta-analyses, both conducted immediately after intervention. The available evidence pointed to large improvements in wrist extension range of motion and moderate gains in Brunnstrom staging, a measure of motor recovery stage, but the certainty of this evidence was rated as very low to low under the GRADE framework. Notably, none of the included studies exploited modern biofeedback applications such as serious games, exoskeleton control, or neurofeedback-based motor learning, relying instead on classic auditory or visual feedback. The authors argue this represents an untapped frontier rather than a dead end.

Why does combining EMG-triggered stimulation with other therapies work so much better than stimulation alone? The review offers a mechanistic explanation grounded in motor learning theory. Isolated muscle activation, the authors note, does not satisfy current stroke rehabilitation guidelines, which emphasize intensive, repetitive, task-oriented practice. EMG-triggered stimulation can amplify patient involvement by rewarding voluntary effort with immediate sensory feedback, but that activation must then be channeled into functional movement patterns. When paired with mirror therapy, motor imagery, or task-specific training, the stimulation provides the neuromuscular spark while the complementary therapy directs it toward meaningful motor goals, addressing voluntary control, coordination, sensorimotor integration, and daily-living function in a single integrated program.

The review is equally candid about the field’s weaknesses. Using the Cochrane Risk of Bias 2 tool, the team judged only eight of the 38 studies to be at low risk of bias, while 13 raised some concerns and 17, nearly half, were rated high risk. Intervention protocols varied wildly, from as few as 4 sessions to as many as 126, and from 2 to 21 sessions per week, making it difficult to isolate the effect of the EMG component itself, since in some trials only the experimental group received the additional therapy. The authors also highlight practical and technical constraints: patients with severe paresis may generate no detectable muscle signal at all, rendering EMG triggering impossible, while chronic patients may produce compensatory rather than task-specific activation. Electrode displacement, crosstalk from neighboring muscles, and low signal-to-noise ratios further degrade reliability, and the skin preparation and calibration these systems demand add to therapist workload.

The authors’ evidence gap map, a visual grid of intervention types against clinical outcomes, reveals where the field stands and where it stumbles. Evidence clusters heavily around immediate post-intervention assessments, while medium- and long-term follow-up data remain scarce, and large swaths of the map are simply blank. Looking forward, the team calls for adequately powered, multicenter randomized trials with rigorous randomization and standardized outcome measures, and points to high-density EMG, which offers far greater spatial resolution and was used in none of the included studies, as a promising next-generation interface. For now, the practical prescription is clear: EMG-based interfaces should be treated as complementary tools within comprehensive rehabilitation programs, not stand-alone replacements. The muscle signal is a messenger, the review suggests, not a medicine, and it heals best when embedded in the rich context of conventional, task-oriented therapy.

Subject of Research: Effectiveness of electromyography-based interfaces for hand and wrist motor rehabilitation after stroke

Article Title: Impact of EMG-based interfaces for hand motor rehabilitation in stroke: A systematic review with evidence gap map and meta-analysis

Article References: Guerrero-Mendez, C. D., Batista, N. P., Alves Filho, J. O., Germer, C. M., & Elias, L. A. (2026). Impact of EMG-based interfaces for hand motor rehabilitation in stroke: A systematic review with evidence gap map and meta-analysis. Medical & Biological Engineering & Computing. https://doi.org/10.1007/s11517-026-03660-7

Image Credits: AI Generated

DOI: 10.1007/s11517-026-03660-7

Keywords: stroke rehabilitation, electromyography, EMG-triggered electrical stimulation, biofeedback, rehabilitation robotics, spasticity, meta-analysis, systematic review, hand function, motor recovery, ICF framework, neurorehabilitation

Researchers at the Belgian National MS Center in Melsbroek, working with collaborators in the Netherlands, the United States, and Canada, examined whether hand function, assessed with the Nine Hole Peg Test, could serve as a reliable primary outcome measure in non-ambulatory people with progressive multiple sclerosis. Their findings, published in the Journal of Neurology, show that hand function deteriorates measurably and consistently in this population over two years—and that a particular way of scoring the test may be sensitive enough to power future clinical trials in a group of patients who have long been left on the sidelines of therapeutic research.

The Nine Hole Peg Test is among the simplest instruments in neurology. It asks participants to pick up nine small pegs, one at a time, place them into a series of holes, and then remove them again, with the entire sequence timed in seconds. What the test lacks in sophistication it makes up for in practicality: it is quick, inexpensive, reproducible, and executable by people who use wheelchairs. It has long been a component of the Multiple Sclerosis Functional Composite, but it has rarely been trusted to carry the weight of a primary endpoint on its own. The new research set out to test whether it could.

The team drew on a well-characterized database of patients followed at the Melsbroek center, applying strict inclusion criteria. In total, 131 people were included, all with progressive multiple sclerosis, all with a baseline EDSS score of at least 7.0—indicating severe disability and, in most cases, inability to walk more than a few meters even with assistance—and all with recorded NHPT and EDSS measurements at baseline, 12 months, and 24 months. This three-point longitudinal design allowed the researchers to track not just whether hand function changed, but how quickly and by how much, across a crucial two-year window that matches the duration of many phase 2 trials.

One of the study’s most important methodological contributions lies in how it defined worsening. Rather than committing to a single threshold, the investigators applied four distinct definitions of NHPT worsening: a classical definition based on overall performance, one based on the best-performing hand, one based on the worst-performing hand, and a composite definition capturing worsening in either hand, abbreviated as EH-NHPT. Each definition captures a different aspect of what clinicians and patients might consider a meaningful loss of manual dexterity, and the differences between them turned out to be substantial for trial design.

The results were striking. Across all four definitions, hand function consistently worsened over time, confirming that manual dexterity decline is a real and pervasive feature of advanced progressive multiple sclerosis rather than a statistical artifact. Under the classical, best-hand, and worst-hand definitions, approximately 25 percent of patients met the criteria for worsening at 12 months, rising to roughly 40 percent at 24 months. But the either-hand definition proved the most sensitive of all: 35.9 percent of patients showed worsening at one year, and 52.7 percent at two years. In other words, by the end of two years, more than half of this cohort had experienced a meaningful decline in the function of at least one hand—a signal strength that could dramatically reduce the sample sizes and costs of future trials.

That sensitivity came with an important caveat. Under the either-hand definition, a notable minority of patients—15.3 percent at 12 months and 9.2 percent at 24 months—showed paradoxical improvement, meaning that one hand’s performance crossed back above the threshold despite the overall trajectory of decline. This pattern reflects the inherent volatility of measuring function in a disease with fluctuating symptoms, where fatigue, temperature, spasticity, and day-to-day variation can all influence performance on a timed task. Trial designers will need to weigh the increased statistical sensitivity of the either-hand definition against the noise introduced by these apparent reversals, potentially through confirmatory re-testing protocols already common in walking-based endpoints.

The study also searched for predictors of decline, with sobering results. Logistic regression analyses of baseline characteristics found only modest associations: older age and poorer baseline hand function slightly reduced the likelihood of meeting the either-hand worsening definition at 24 months, plausibly because patients whose hands were already severely impaired had less room left to deteriorate. At the 12-month mark, no significant predictors emerged at all. The practical implication is that clinicians designing trials cannot easily pre-select patients likely to worsen rapidly; instead, the widespread and largely unpredictable nature of hand function decline in this population becomes an argument for the measure’s utility, since a high proportion of any enrolled cohort will be at risk of reaching the endpoint.

The broader context makes these findings all the more consequential. Progressive multiple sclerosis remains one of neurology’s toughest therapeutic challenges. The landmark approvals of ocrelizumab for primary progressive disease and siponimod for active secondary progressive disease marked genuine breakthroughs, but both trials enrolled predominantly ambulatory patients with lower disability scores. Meanwhile, the underlying biology of progression—inflammation within the central nervous system compartment, smouldering neurodegeneration, mitochondrial injury, and progressive gray matter atrophy—continues unchecked in people with advanced disability. Recent phase 3b work has begun to push into more advanced disease stages, and the field increasingly recognizes that the population excluded by walking-based endpoints is precisely the population with the greatest unmet need.

There is also a compelling efficiency argument. Clinical trials in progressive multiple sclerosis are notoriously expensive and slow, and the field has grappled for years with how to streamline them. A simple, cheap, chairside test of manual dexterity that can be administered in minutes without gait-dependent machinery could open trials to larger and more diverse patient populations while potentially shortening follow-up periods, given the substantial two-year event rates documented in this study. Futility-design strategies, which allow early stopping of unpromising therapies, could also benefit from the larger event rates that a sensitive hand-function endpoint provides. None of this diminishes the need for careful validation—the current study is observational, retrospective, and drawn from a single center—but it offers exactly the kind of feasibility and trajectory data that trial designers need before committing to an endpoint.

For the hundreds of thousands of people worldwide living with severe, non-ambulatory progressive multiple sclerosis, the symbolism of this research may matter as much as the statistics. A population that has effectively been told the trial system has nothing for them now has a concrete, evidence-based proposal for their inclusion, built on a test they could complete in a clinic chair in under a minute. Whether the Nine Hole Peg Test ultimately earns the status of a primary outcome measure will depend on validation in prospective, multi-center settings and on consensus among regulators and trialists. But the central question posed by the researchers—whether hand function can step up—now has a data-driven answer: hand function declines reliably in this population, the either-hand definition captures that decline more sensitively than any alternative, and the path toward more inclusive trials in progressive multiple sclerosis may well run through nine holes, nine pegs, and a stopwatch.

Subject of Research: The Nine Hole Peg Test as a potential primary outcome measure for clinical trials in non-ambulatory people with progressive multiple sclerosis

Article Title: Clinical trials in non-ambulatory people with progressive multiple sclerosis: can the nine hole peg test step up as primary outcome measure?

Article References: D’Haeseleer, M., Strijbis, E. M. M., Cutter, G., della Faille, L., D’hooghe, M. B., Kos, D., Lison, A.-L., Mostert, J., Nagels, G., Repovic, P., Uitdehaag, B. M. J., Van Remoortel, A., Van Schependom, J., & Koch, M. W. (2026). Clinical trials in non-ambulatory people with progressive multiple sclerosis: can the nine hole peg test step up as primary outcome measure?. Journal of Neurology, 273(10), Article 608. https://doi.org/10.1007/s00415-026-14140-1

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14140-1

Keywords: multiple sclerosis, progressive multiple sclerosis, Nine Hole Peg Test, hand function, non-ambulatory, clinical trials, outcome measures, EDSS, disability worsening, manual dexterity, neurology, trial endpoints