This Week in Physical Medicine & Rehabilitation — Sep 24, 2026

This AudioScholar briefing is generated by artificial intelligence for healthcare professionals and trainees. It is not medical advice.

Welcome to This Week in Physical Medicine and Rehabilitation. This week we're covering 10 notable papers spanning neuromodulation in neurorehabilitation, prognosis and risk prediction after spinal cord and neurological injury, and the evidence base behind our rehabilitation service models. Let's dive in.

We'll start with electrical stimulation, where two randomised trials tested whether adding a neuromodulation device to standard therapy buys anything extra. In Archives of Physical Medicine and Rehabilitation, Külekçioğlu and colleagues randomised patients with acute and subacute stroke and hemiplegic shoulder pain to transcutaneous electrical nerve stimulation, neuromuscular electrical stimulation, or both, on top of an identical hour-long rehabilitation programme five days a week for three weeks. Sixty of the seventy-five randomised participants completed the protocol. Every group improved on pain, Fugl-Meyer upper extremity scores, shoulder disability, Barthel Index and range of motion. The combination arm achieved significantly greater reductions in pain and shoulder disability than transcutaneous stimulation alone, but on shoulder disability it was no better than neuromuscular stimulation by itself, and the combination did not improve motor function, spasticity, or activities of daily living. So the signal here is confined to pain and shoulder-specific limitation in a small single-study sample, and the authors themselves call for larger trials before treating this as settled [1]. Alongside that, Spinal Cord published a randomised sham-controlled trial from a rehabilitation centre in Seoul, led by Cho, testing repetitive transcranial magnetic stimulation added to occupational therapy in cervical spinal cord injury, three sessions a week for five weeks. With just fourteen patients in the active arm and fifteen in the sham arm, the active group showed greater gains in strength, sensation and prehension on the GRASSP, and greater improvement in spinal cord independence. Worth noting the less comfortable detail: pain, spasticity and depression scores rose in both groups over the study period, with shoulder pain increasing significantly in the control arm. This is a very small trial, and the authors frame it as a contribution rather than a definitive result [2].

The second theme is prognosis, and three papers this week try to sharpen how we forecast recovery and risk. In Archives of Physical Medicine and Rehabilitation, Sugyo and colleagues applied latent class mixed modelling to more than four thousand longitudinal observations from 516 patients at a single specialised spinal cord injury centre in Japan, tracking lower extremity motor scores over the first year after cervical injury. Rather than one average recovery curve, the data separated cleanly into five trajectories: early high recovery, steep nonlinear recovery, moderate recovery with an early plateau, low recovery, and slow sustained recovery. Class assignment was highly confident statistically, and admission motor score was the one variable associated with membership across every comparison, with age, ASIA Impairment Scale grade and bony injury mapping onto specific classes. It's retrospective and single-centre, so this is a framework for understanding heterogeneity rather than a validated prediction tool, but it does argue that a plateau at three months in one patient and continued slow gain in another are distinct biological patterns, not noise [3]. Complementing that, Spinal Cord reports a multicentre prospective study from the Keio Spine Research Group, led by Isogai, following 363 patients with degenerative cervical myelopathy who had preoperative deltoid or biceps weakness out to two years after surgery. Over sixty percent of those patients regained full strength, which is genuinely reassuring for counselling. Persistent weakness was more likely in older patients, and on multivariable analysis a lower preoperative upper extremity motor score roughly tripled the odds of persistent weakness, as did prior cervical surgery, with additional musculoskeletal disease and hypertension carrying smaller increases in risk [4].

Staying with risk prediction, Archives of Physical Medicine and Rehabilitation also published work from Holtzer and colleagues on 109 ambulatory older adults with multiple sclerosis, average age about 65, followed prospectively with monthly telephone fall reporting over a mean of just over two years. They compared a simple patient-reported measure, the Patient Determined Disease Steps scale, against the short physical performance battery. A one standard deviation worsening on the patient-reported scale was associated with roughly a 28 percent increase in the odds of falling, while a one standard deviation better physical performance battery score was associated with about a quarter lower odds. Formal equivalence testing found the two measures statistically comparable in their association with prospective falls. In a cohort of this size that is a single-sample finding rather than a validated substitution, but it supports the case that a questionnaire taking seconds may carry similar falls-risk information to a supervised performance battery in older adults with multiple sclerosis [5].

Our third theme is where the evidence base is thinner than we would like, and two systematic reviews in Clinical Rehabilitation make that uncomfortably clear. Park and colleagues conducted a network meta-analysis of dysphagia rehabilitation after head and neck cancer treatment, pulling in 21 trials with 18 of them, covering 1,139 participants, contributing to connected networks comparing swallowing exercise, neuromuscular electrical stimulation plus exercise, speech therapy plus exercise, and telerehabilitation. Exercise improved clinician-rated swallowing compared with usual care and improved instrumental measures compared with telerehabilitation. Beyond those two comparisons, nothing reached statistical significance. Different interventions ranked top for different outcome domains, and confidence in the evidence ranged from moderate all the way down to very low, with every patient-reported comparison rated very low. The authors' conclusion is that no intervention is consistently superior and that selection has to be individualised while adequately powered trials are done [6]. The same journal published a systematic review from Cointo and colleagues on fall prevention after lower-limb amputation, covering 23 studies and 832 participants across virtual reality, resistance training, proprioceptive neuromuscular facilitation, electrical stimulation, dual-task training and preoperative education. Resistance training and proprioceptive neuromuscular facilitation were associated with improved fall risk, comprehensive multicomponent protocols with reduced fall incidence, and preoperative education with better balance. But a meta-analysis of ten studies found that virtual reality improved balance confidence without significantly improving fall risk or balance itself — a useful reminder that confidence and capability can move independently. The authors are explicit that with low to very low certainty and high risk of bias throughout, these findings are exploratory and not a basis for clinical mandates [7].

Finally, three papers ask how rehabilitation gets delivered and by whom. In Clinical Rehabilitation, Sheerin and colleagues ran a cluster non-randomised controlled trial across four intervention hospitals and one control hospital, comparing therapy-led conservative management of acute hand injuries — fractures, mallet, volar plate and central slip injuries — against usual medical-led care in 133 adults. There was no statistically significant between-group difference in QuickDASH function at eight weeks or six months, or in any secondary outcome, while both groups improved substantially over time. The therapy-led pathway generated significantly fewer onward referrals and fewer healthcare contacts. The authors read comparable recovery at lower resource cost as support for greater use of therapy-led pathways, though this was not a randomised design and was not formally powered as a non-inferiority trial [8]. On the paediatric side, Disability and Rehabilitation reports a randomised controlled trial from Kim and colleagues of a real-time computer vision and artificial intelligence hand rehabilitation programme in 45 children with cerebral palsy aged eighteen months to seven years, delivered as three fifteen-minute sessions a week for eight weeks against a no-intervention control. There were significant time-by-group effects on the Box and Block Test and palmar pinch strength bilaterally, with compliance around 92 percent and caregivers reporting satisfaction and reduced burden. The absence of an active comparator limits what can be attributed to the technology rather than to added therapy time, but feasibility for home-based delivery looks real [9]. And in the American Journal of Physical Medicine and Rehabilitation, Cho and colleagues describe a 63-year-old woman with dissociative seizures and multi-domain disability after traumatic brain injury, managed on an inpatient acquired brain injury unit with physiatrist-led diagnostic clarification, team training, structured family education and function-oriented, functional-neurological-disorder-informed rehabilitation. Her seizures decreased in frequency and severity and she was discharged home with meaningful functional gains. It is a single case report, so it carries no comparative weight, but it illustrates a rehabilitation pathway for a condition where structured pathways remain scarce [10].

If you only have time for one paper this week, make it the multiple sclerosis falls study in Archives of Physical Medicine and Rehabilitation [5]. It asks whether a one-item patient-reported mobility scale carries the same prospective falls information as a supervised performance battery, and the answer it returns — statistical equivalence in this cohort — reopens a practical question about how falls risk screening is structured in busy clinics.

Here is what this week's evidence adds up to in Physical Medicine and Rehabilitation. First, added neuromodulation shows narrow, outcome-specific benefit: combined electrical stimulation helped shoulder pain and shoulder-specific disability after stroke but not motor function or independence, and repetitive transcranial magnetic stimulation in cervical spinal cord injury looked promising in a trial of under thirty patients — both are hypothesis-supporting, not practice-settling. Second, recovery after cervical cord pathology is heterogeneous but partly predictable: five distinct lower limb trajectories emerged from a large single-centre cohort, and in degenerative cervical myelopathy most patients with preoperative proximal weakness regained full strength, with severity at baseline and prior cervical surgery marking those least likely to. Third, where we lean hardest on rehabilitation — dysphagia after head and neck cancer, falls after amputation — the pooled evidence is sparse, imprecise and mostly low certainty, with no consistently superior intervention identified. Fourth, therapy-led hand injury care produced comparable functional recovery with fewer healthcare contacts, though from a non-randomised cluster design that stops short of proving equivalence. And finally, a simple patient-reported mobility scale performed comparably to objective performance testing for prospective falls in older adults with multiple sclerosis — a finding that still needs external validation before it displaces anything.

That's your roundup for This Week in Physical Medicine and Rehabilitation. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And think of one colleague — in any specialty — who never has time to keep up with the literature. Tell them about AudioScholar: a free ten-minute weekly for every specialty, to listen to in any podcast app, or to read at audioscholar dot C C.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.