Rhythmic auditory stimulation combined with split-belt treadmill training improves spatiotemporal gait parameters in hemiparetic stroke patients: a randomized controlled trial — a speed signal, not proof of restored symmetry
Clinical Question (PICO)
In medically stable adults aged 25–75 years, more than one month after stroke, with hemiparesis and residual assisted walking capacity, does rhythmic auditory stimulation plus split-belt walking training (RAS+SWT) improve three-week overground gait parameters compared with the same split-belt training without sound cues or conventional rehabilitation alone?
The outcomes were walking speed, cadence, affected- and unaffected-side step length, and a step-length-based limping index. No treatment-effect subgroup analysis was reported; stroke type, side, and baseline functional status describe applicability, not proven effect modifiers. This is an October 2026 paper selected through the PMC look-back search, not a new November 4 trial report.
Bottom Line
The authors conclude that combined training improves several gait parameters relative to conventional rehabilitation, while its demonstrated advantage over split-belt training alone is confined to walking speed: the between-group difference in change was 0.110 m/s (95% CI 0.022–0.197). This small, single-center, three-week study supports a preliminary speed signal—not established benefits for long-term independence, falls, or gait symmetry—and its exploratory comparison intervals were not multiplicity-adjusted.
Design
- Trial type: Prospective, single-center, assessor-blinded, three-arm randomized controlled trial. Participants and treating therapists were not blinded.
- N: 30 randomized; all completed treatment and assessments.
- RAS+SWT: 11.
- SWT alone: 8.
- Conventional rehabilitation: 11.
- Randomization: Independent statistician generated an Excel random-number sequence; predefined number intervals determined assignment. Allocation was kept in opaque envelopes labeled by enrollment number. Realized groups were unequal; no stratification was described.
- Setting: Shaanxi Provincial Rehabilitation Hospital, China.
- Enrollment: January–April 2026.
- Follow-up: Baseline to completion of three weeks of treatment; no later follow-up.
- Analysis: All 30 completers contributed data, with no reported missing outcomes. Linear mixed-effects models included time, group, their interaction, and a participant random intercept. The unadjusted base model was selected over a covariate-extended model. The report does not explicitly label its analysis intention-to-treat.
- Primary outcomes: Overground spatiotemporal gait parameters measured with Vicon three-dimensional motion capture. Change-score contrasts and Hedges’ g were also reported. Simple-effects testing used Tukey correction, but Table 3's exploratory 95% CIs were not adjusted for multiple comparisons.
Population
Inclusion Criteria
- CT- or MRI-confirmed stroke, age 25–75, and onset more than one month earlier.
- Stable vital signs and capacity to participate in exercise; low-to-moderate cardiac rehabilitation risk.
- Modified Rankin Scale (mRS) 3–5 and Functional Ambulation Category (FAC) below 2.
- Sufficient comprehension and cooperation, with S5Q score at least 3; the eligibility text also states no cognitive impairment.
Exclusion Criteria
- Unstable cardiac disease, including severe arrhythmia or valvular dysfunction.
- Resting blood pressure above 180/100 mmHg.
- Subarachnoid hemorrhage.
- Severe speech, visual, auditory, or psychiatric disorders interfering with participation.
- Pregnancy, breastfeeding, or concurrent participation in another intervention study.
Baseline Characteristics (overall or representative arm)
- Group mean ages: 47.55 years in controls, 45.25 with SWT, and 40.41 with RAS+SWT.
- 26 men and 4 women; 22 hemorrhagic and 8 ischemic strokes, calculated from the published arm counts.
- Overall mean time since stroke: 83.3 ± 31.01 days, range 36–155 days.
- All participants were rated FAC 1; actual mRS scores were 3–4. The authors acknowledge a conservative local FAC interpretation, with better walking capacity than that label might suggest elsewhere.
- Group mean modified Barthel Index scores ranged from 75.82 to 81.13; MMSE means ranged from 24.73 to 27.91. The latter should be considered alongside the stated cognitive eligibility criterion rather than assuming a uniformly cognitively intact cohort.
- Baseline walking speeds were 0.40, 0.36, and 0.37 m/s in control, SWT, and RAS+SWT groups, respectively. The combined group already had a lower limping index: 0.22 versus 0.42 in controls.
Interventions
- All arms: Conventional rehabilitation for 50 min/day—30 minutes of manual therapy and 20 minutes of active limb exercises—five sessions per week for three weeks.
- RAS+SWT: An additional 25-minute treadmill session: 5 minutes with equal belt speeds, 10 minutes at a 2:1 speed ratio with the shorter-step leg on the fast belt, and 10 minutes of equal-speed recovery. Sound cues began during the final 2 minutes of baseline and continued during adaptation; they stopped for recovery. Metronome tempo was individually set 5–10% above baseline cadence and reduced when stability or synchronization was difficult. The detailed protocol specifies ceiling-mounted speakers, despite the randomization narrative's reference to wearing auditory devices.
- SWT alone: The same additional treadmill duration, phases, belt ratio, and leg placement, but without auditory stimulation.
- Conventional rehabilitation alone: No added treadmill or sound-cue period. Total therapy was 50 rather than 75 min/day; therapist contact was reported as 55 versus 80 min/day.
- Safety and adherence: Therapist supervision, body-weight support adjusted to balance/fatigue, handrails, usual orthoses where applicable, and vital-sign monitoring. All participants completed all prescribed sessions. Each treadmill arm received 375 additional training minutes over the study, calculated as 25 × 5 × 3.
Outcomes
Primary Outcomes (change from baseline to three weeks):
The following are the paper's Table 3 mean differences in change, with unadjusted 95% CIs. These are continuous outcomes, so HRs, ORs, and RRs do not apply. No subgroup-specific estimates or pairwise P values accompany this table; none are inferred here. Positive speed, cadence, and step-length differences favor the first treatment; a negative limping-index difference indicates less asymmetry.
| Outcome | RAS+SWT versus control | RAS+SWT versus SWT | SWT versus control |
|---|---|---|---|
| Walking speed, m/s | 0.167 (0.090 to 0.245) | 0.110 (0.022 to 0.197) | 0.058 (0.010 to 0.105) |
| Cadence, reported steps/min | 9.340 (4.456 to 14.223) | 4.407 (−3.636 to 12.451) | 4.932 (−1.615 to 11.479) |
| Unaffected-side step length, m | 0.077 (0.039 to 0.116) | −0.079 (−0.344 to 0.185) | 0.156 (−0.107 to 0.420) |
| Affected-side step length, m | 0.081 (0.035 to 0.127) | 0.048 (−0.009 to 0.105) | 0.033 (−0.005 to 0.070) |
| Limping index, unitless | −0.065 (−0.258 to 0.129) | −0.089 (−0.290 to 0.112) | 0.024 (−0.149 to 0.198) |
The mixed-model time-by-group interactions were significant for walking speed (F = 10.68, P < 0.001), cadence (F = 5.01, P = 0.014), and affected-side step length (F = 6.21, P = 0.006). They were not significant for unaffected-side step length (F = 1.47, P = 0.247) or limping index (F = 0.430, P = 0.655).
RAS+SWT walking speed increased descriptively from 0.37 ± 0.14 to 0.54 ± 0.23 m/s. The speed change contrast versus SWT had Hedges’ g = 0.988. Neither the cadence nor either step-length contrast between the two treadmill arms excluded zero. The lower post-treatment limping index in the combined arm does not establish superior improvement: baseline values already differed and the interaction was null.
Secondary Outcomes:
The report does not specify a separate patient-centered secondary efficacy hierarchy. Walking independence, falls, community mobility, quality of life, and persistence of benefit were not established. Gait speed and motion-capture parameters should not be relabeled as disability or real-world activity outcomes.
Adverse Events / Safety:
The authors report no serious adverse events during treatment, no withdrawals, and complete adherence. They do not provide a detailed arm-level inventory of minor adverse events or evidence sufficient to exclude uncommon harms. Bleeding, recurrent stroke, and mortality were not reported as comparative efficacy/safety endpoints.
Figures
Source: Du et al., Frontiers in Neurology (2026), PMC13713294, CC BY 4.0 — fneur-17-1904812-g001.webp. Click image to expand.
Source: Du et al., Frontiers in Neurology (2026), PMC13713294, CC BY 4.0 — fneur-17-1904812-g002.webp. Click image to expand.
Source: Du et al., Frontiers in Neurology (2026), PMC13713294, CC BY 4.0 — fneur-17-1904812-g003.webp. Click image to expand.
Source: Du et al., Frontiers in Neurology (2026), PMC13713294, CC BY 4.0 — fneur-17-1904812-g004.webp. Click image to expand.
Criticisms
- Dose and attention confounding: The combined-versus-control comparison adds 25 minutes of treatment every day. It tests an expanded rehabilitation package, not the isolated effect of sound cues. The RAS+SWT-versus-SWT comparison is the cleaner, approximately time-matched test of cueing.
- Small and selective sample: Only 30 patients from one center, with 8 in the SWT arm; predominantly younger men and hemorrhagic stroke survivors. The locally interpreted FAC 1 label should not be used to generalize to all highly dependent walkers.
- Short follow-up and surrogate outcomes: Three-week overground performance does not prove durable everyday mobility, fewer falls, or independence. Early post-stroke recovery and additional practice complicate interpretation of within-arm gains.
- Multiplicity and model selection: Several outcomes and contrasts, exploratory unadjusted CIs, and data-dependent choice between models warrant caution. Table 3's SWT-versus-control speed CI excludes zero, while the corrected within-SWT pre/post test has P = 0.054; these are different comparisons and testing procedures, not interchangeable significance claims.
- Symmetry was not convincingly changed: The time-by-group limping-index test was null, and baseline asymmetry favored RAS+SWT. A significant post-treatment group difference is not a significant treatment effect on change.
- Measurement/reporting questions: Cadence is described as an average of ipsilateral step rates yet reported in steps/min; conventional total-cadence comparability needs clarification. The population's cognitive eligibility, local FAC grading, and speaker-versus-wearable description also warrant clarification. These issues are not repaired by guessing missing definitions.
- Registration and missing data: Registration is stated, but the registry history and prospective prespecification were not independently verified. There were no observed missing outcomes; the withdrawal criteria would require scrutiny in a larger trial with actual attrition. No treatment-effect subgroup results were reported.
Funding
The authors state that no financial support was received for the work or its publication and declare no commercial or financial conflicts. No sponsor role is described. Registration: ChiCTR2600121963; ethics approval: Shaanxi Provincial Rehabilitation Hospital, 2025134. This is a Chinese registry identifier, not an NCT number.
The paper
- Authors. Minghui Du et al.
- Title. Rhythmic auditory stimulation combined with split-belt treadmill training improves spatiotemporal gait parameters in hemiparetic stroke patients: a randomized controlled trial.
- Journal. Frontiers in Neurology.
- Year. 2026; volume 17, article 1904812; published October 14, 2026.
- DOI. 10.3389/fneur.2026.1904812
- PMCID. PMC13713294
- PMID. 42952509
- License. The source article and attributed figures are distributed under CC BY 4.0. This review paraphrases and critically interprets the report; the figures are unaltered linked reproductions.
Deep Dive — click to expand
What this is
A plausible combination—spatial adaptation from a split-belt treadmill and temporal cueing from a metronome—produced a short-term gait-speed signal in 30 stroke survivors. The clinically interesting question is narrower than whether the treatment package beats usual care: does adding sound to an already staffed treadmill session help enough to matter? The directly relevant estimate is 0.110 m/s versus SWT alone, with a wide 95% CI of 0.022–0.197. The following frameworks are editorial interpretation, not trial-tested protocol recommendations or a statement of current guidelines.
1. Shadow Audit
The attractive multidimensional conclusion hides two distinct comparisons. Relative to conventional care, the combined arm received 75 rather than 50 therapy minutes per day: a calculated 50% increase in scheduled treatment and 375 additional treadmill minutes. Those extra minutes may explain part of the package's advantage. Only the comparison between treadmill arms meaningfully isolates adding auditory cues, and that comparison showed an incremental signal for speed—not demonstrated superiority for cadence, step length, or symmetry.
The other shadow is the limping index. RAS+SWT started at 0.22 and controls at 0.42. With a null time-by-group interaction (P = 0.655), the lower final value cannot be promoted into a causal symmetry-restoration result. Faster is supported more clearly than more symmetric.
2. Inversion Engine
For the opposite practical conclusion—sound adds too little to justify routine use—to remain plausible, the true incremental benefit need not be zero. It need only be below a patient-important threshold. The 0.022–0.197 m/s interval versus SWT spans small through potentially meaningful benefits.
The authors compare their combined arm's roughly 0.17 m/s within-group gain with previously published 0.13–0.16 m/s meaningful-change estimates. But the increment attributable to adding sound is 0.110 m/s, not that within-group gain. Moreover, those benchmarks come from different cohorts and are individual-change anchors, not universal thresholds for between-group mean differences. A larger, equal-dose trial whose incremental-effect interval lies above a justified patient-centered threshold would strengthen adoption; a precise near-zero estimate or loss of benefit at follow-up would undermine it. The current trial settles neither inversion.
3. Second-Order Catalyst
The first plausible adopters are specialist inpatient rehabilitation teams that already have a split-belt treadmill, a safety harness, and therapists experienced with assisted walking. For them, adding a metronome can be evaluated without buying the capital platform or claiming a new universal standard. This is a candidate for a monitored service evaluation in patients resembling this cohort, after clinical review—not an unsupervised home treadmill instruction.
The key institutional change is measurement: document the treadmill dose, cue tempo, stability, and an overground speed measure before and after treatment, then test retention after discharge. Merely recording a better end-of-session treadmill performance would fail to reproduce the study's overground assessment and would miss its largest unanswered question.
4. Asymmetric Leverage
The low-cost component is sound; the expensive component is the treadmill session and its supervision. If a session already exists, a useful speed improvement could offer favorable marginal value. If a center has no equipment or trained staff, this trial is not an economic justification to create a new treadmill program: there is no cost-effectiveness comparison, and the conventional-care arm received less total treatment.
The asymmetric next experiment is therefore not a larger advertisement for the combined package. It is an adequately powered, dose-matched comparison of cueing versus no cueing, retaining overground outcomes and adding real-world walking, falls, and follow-up. That isolates the inexpensive decision from the resource-intensive background platform.
5. Paradigm Destroyer
The reflex this paper should displace is equating gains in walking speed with correction of gait asymmetry, not the broader habit of providing conventional rehabilitation. Its own data separate these endpoints.
A defensible two-sentence local evaluation protocol would read: “When considering auditory cueing during an existing supervised split-belt program, individualize tempo to stability and record overground speed and symmetry separately. Do not promise independent walking or durable symmetry gains on the basis of this three-week trial; reassess retention and patient-important mobility before treating an early speed gain as success.”
MVP — Minimum Viable Proof
The minimum verified statement is that, among 30 selected participants, adding RAS to an otherwise similar SWT program was associated with a 0.110 m/s greater three-week speed change (95% CI 0.022–0.197; exploratory, unadjusted interval). That is enough to motivate a supervised, measured evaluation where the platform already exists. Routine practice-changing proof would additionally require a reproducible, clinically important incremental benefit, durable overground transfer, and safety in a broader population; this paper does not provide those elements.
Best Combination
Read this alongside the primary mechanistic literature without treating distinct forms of auditory feedback as the same intervention. Lewek et al.'s randomized trial found that error augmentation was not superior to error minimization or conventional treadmill training with verbal feedback for step-length asymmetry. Cherry-Allen et al.'s 2024 study examined 15 adults and used binary auditory reinforcement after adaptation, not concurrent metronome pacing; immediate retention of improved step-length asymmetry supports a motor-learning rationale, not direct replication of this clinical regimen. These prior-study descriptions were checked against Europe PMC's primary-paper abstracts; a formal systematic review was not performed.
For clinical interpretation, Tilson et al. estimated a 0.16 m/s meaningful individual gait-speed change in a cohort assessed 20–60 days after stroke, while Bohannon et al. identified 0.13 m/s in a small inpatient rehabilitation cohort. Together these papers help frame the scale of a speed change; they do not validate transferring a threshold mechanically to the incremental treatment contrast or to this study's population. Combine the motor-learning hypothesis with equal-dose comparative testing and retention measurement—not with an unverified claim of guideline endorsement.
Overvalue Warning
- Do not turn statistical separation into a universal recovery claim. A speed contrast in a small, predominantly male and hemorrhagic-stroke sample does not establish fewer falls, independent community ambulation, efficacy in every stroke phenotype, or superiority for symmetry. A nonsignificant cadence or step-length contrast is also not proof of equivalence.
- Do not equate no serious events in 30 people with established safety. Three weeks, close supervision, selected cardiac risk, body-weight support, and no detailed minor-event table leave substantial uncertainty about broader implementation. Neither durable efficacy nor safety without the study's staffing and safeguards has been shown.
Educational use notice. The Time Is Brain Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.