Clinical and neurophysiological effects of manual acupuncture for upper-limb motor dysfunction after ischemic stroke: a three-arm randomized controlled trial

Clinical Question (PICO)

In right-handed adults aged 30–85 with unilateral upper-limb motor impairment within three months of ischemic stroke, does a four-week course of standardized manual acupuncture plus conventional rehabilitation improve upper-limb motor function and daily activity more than superficial nonacupoint sham needling plus the same rehabilitation, or rehabilitation alone? The primary outcome was baseline-to-endpoint change in the Fugl-Meyer Assessment of the Upper Extremity (FMA-UE); the Modified Barthel Index (MBI) and exploratory surface-EMG and fNIRS measures were secondary outcomes.

Bottom Line

Among 84 randomized participants (28 per group), manual acupuncture plus rehabilitation produced an 18.04-point mean FMA-UE gain over the intervention, versus 9.46 with sham acupuncture and 8.39 with rehabilitation alone. The adjusted between-group advantages were 8.57 points over sham (95% CI, 4.55–12.59; Bonferroni-adjusted p < 0.001) and 9.64 over rehabilitation alone (95% CI, 5.62–13.66; adjusted p < 0.001). MBI improvement also favored manual acupuncture. These are promising four-week results from one small center—not proof of durable benefit, a specific neural mechanism, or effectiveness across the full range of stroke survivors.

Design

Trial type: Single-center, three-arm, assessor-blinded randomized controlled trial; acupuncturists were unblinded. Participants in the two needling groups were not told whether they received manual or sham needling; the rehabilitation-only group could not be blinded.
N: 84 randomized; 28 per group. All 84 had baseline and endpoint data and were included in the full analysis set. Three endpoint assessments occurred before the scheduled week-four visit because of early discharge; no outcome imputation was performed.
- Manual acupuncture + conventional rehabilitation: 28.
- Sham acupuncture + conventional rehabilitation: 28.
- Conventional rehabilitation alone: 28.
Randomization: Computer-generated, 1:1:1 permuted blocks of six, without stratification. Sequentially numbered, opaque, sealed envelopes were used after eligibility confirmation and baseline assessment.
Setting: Departments of Rehabilitation Medicine and Geriatric Rehabilitation, Affiliated Panyu Central Hospital, Guangzhou Medical University, Guangzhou, China.
Enrollment: December 2025–April 2026. The trial was registered before first enrollment.
Follow-up: Four-week intervention and endpoint assessment; three participants discharged early had endpoint testing before discharge (one in the sham group after three weeks; two in the manual-acupuncture group on days 22 and 25).
Analysis: Full analysis set, defined as all randomized participants with observed baseline and endpoint data, analyzed in randomized groups. Linear mixed-effects models included group, time, and group-by-time interaction, with participant as a random intercept. Significant interactions were followed by pairwise change comparisons with Bonferroni adjustment. fNIRS interaction tests were corrected across regions or connections using the Benjamini–Hochberg false-discovery-rate (FDR) procedure. The sample-size target was 28 per arm, based on the manual-acupuncture-versus-sham FMA-UE comparison.
Primary outcome: Change in FMA-UE score from baseline to the four-week endpoint.

Population

Inclusion Criteria

Right-handed adults aged 30–85 years with imaging-confirmed ischemic stroke within the prior three months.
Clinically stable, conscious, and able to understand instructions and complete treatment and assessments.
Unilateral upper-limb motor dysfunction, Brunnstrom stage II or higher, and affected-arm Manual Muscle Testing grade 4 or lower; Mini-Mental State Examination score within the normal range.

Exclusion Criteria

Pre-existing upper-limb dysfunction unrelated to the index stroke; intracranial hemorrhage or hemorrhagic transformation; prior traumatic brain injury or another central nervous system disorder.
Severe systemic disease, active infection, malignancy, coagulation disorder, or skin condition at a planned needling site; prior needle-related syncope.

Baseline Characteristics (manual-acupuncture arm)

Mean age 67.14 ± 9.73 years; 17/28 (60.7%) male.
Median time since stroke 2.00 months (IQR 1.00–2.75); baseline FMA-UE 27.07 ± 10.10 and MBI 41.39 ± 12.24.
Deep hemispheric lesions were reported in 15/28 (53.6%). The authors reported no statistically significant baseline differences among groups (all p > 0.05).
Of 96 patients assessed, 12 were excluded before randomization: needle-related syncope history (2), coagulation disorder (5), hemorrhage/hemorrhagic transformation (4), or a local skin lesion (1).

Interventions

Manual acupuncture + conventional rehabilitation: All participants received routine ischemic-stroke medical management and the same standardized rehabilitation program: 20 sessions over four weeks, five days per week, approximately 60 minutes per session. Sessions included affected-arm positioning, range-of-motion work, task-oriented upper-limb practice, and progressive strengthening or endurance exercises. Licensed, trial-trained acupuncturists used disposable stainless-steel needles at GV20, GV14, LI11, PC6, ST36, and SP6; limb acupoints were needled bilaterally. Needles were retained for 30 minutes. Manipulation was applied after insertion and every 10 minutes to elicit de qi.
Sham acupuncture + conventional rehabilitation: The same rehabilitation dose was paired with superficial needling at six nearby but distinct nonacupoint sites (0.2 body-cun), without manipulation or an attempt to elicit de qi. Retention time, treatment schedule, setting, and practitioner contact were matched; needle handles were lightly touched every 10 minutes.
Conventional rehabilitation alone: The same 20-session, four-week rehabilitation program without needling.

Outcomes

Primary Outcome — FMA-UE change from baseline to the four-week endpoint:

A significant group-by-time interaction was reported (Wald χ² = 26.54, p < 0.001). Mean changes (95% CI) were 18.04 (14.19–21.88) points with manual acupuncture, 9.46 (6.50–12.43) with sham, and 8.39 (6.39–10.40) with rehabilitation alone. The manual-acupuncture advantage was 8.57 points over sham (95% CI, 4.55–12.59; Bonferroni-adjusted p < 0.001) and 9.64 over rehabilitation alone (95% CI, 5.62–13.66; adjusted p < 0.001). Sham and rehabilitation alone did not differ (adjusted p > 0.999). The authors cite published FMA-UE clinically important difference estimates ranging roughly from 5 to 13 points, depending on stroke stage, severity, and anchor method; the trial did not establish one universal threshold.

Secondary Outcomes:

MBI: Significant group-by-time interaction (Wald χ² = 14.95, p < 0.001). Mean changes were 22.82 (95% CI, 18.19–27.45) with manual acupuncture, 11.54 (8.07–15.00) with sham, and 15.50 (10.60–20.40) with rehabilitation alone. Manual acupuncture exceeded rehabilitation alone by 7.32 points (95% CI, 1.52–13.13; adjusted p = 0.040) and sham by 11.29 (95% CI, 5.48–17.09; adjusted p < 0.001). Sham versus rehabilitation alone was not significant (adjusted p = 0.542).
Surface EMG during passive stretch: Group-by-time interactions were reported for triceps brachii RMS (Wald χ² = 8.33, p = 0.016) and wrist-flexor RMS (Wald χ² = 19.75, p < 0.001), but not biceps or wrist-extensor RMS. Triceps change was +12.17 μV (95% CI, 6.68–17.66) with rehabilitation, +8.32 (3.94–12.70) with sham, and +2.70 (−1.98–7.39) with manual acupuncture; rehabilitation exceeded manual acupuncture by 9.47 μV (95% CI, 3.00–15.93; adjusted p = 0.012). Wrist-flexor change was −7.44 μV (−11.10 to −3.79), −4.98 (−7.91 to −2.05), and +2.37 (−1.18 to 5.93), respectively. Manual acupuncture differed from rehabilitation by +9.81 μV (95% CI, 5.31–14.32; adjusted p < 0.001) and from sham by +7.35 (2.84–11.85; adjusted p = 0.004). These are task- and muscle-specific RMS changes, not a uniform reduction in tone or demonstrated antispastic effect.
Surface EMG during voluntary contraction: Interactions were reported for biceps brachii (Wald χ² = 10.33, p = 0.006) and triceps brachii RMS (Wald χ² = 29.94, p < 0.001), but not wrist flexor (p = 0.086) or wrist extensor (p = 0.817). Biceps changes were +39.09 μV (95% CI, 22.63–55.55) with manual acupuncture, +10.51 (−4.55 to 25.57) with sham, and +8.37 (−7.40 to 24.13) with rehabilitation. Manual acupuncture exceeded rehabilitation by 30.72 μV (95% CI, 9.80–51.64; adjusted p = 0.012) and sham by 28.58 (7.66–49.50; adjusted p = 0.022). Triceps changes were +57.64 μV (38.37–76.91), +9.24 (−3.14 to 21.62), and +8.31 (−4.80 to 21.42), respectively; manual acupuncture exceeded rehabilitation by 49.33 μV (29.12–69.54) and sham by 48.40 (28.19–68.61), both adjusted p < 0.001. RMS was analyzed without normalization.
Task-related fNIRS activation: After FDR correction, group-by-time interactions remained for ipsilesional PFC (Wald χ² = 40.56, p < 0.001; FDR-adjusted p < 0.001) and ipsilesional M1 (Wald χ² = 9.07, p = 0.011; FDR-adjusted p = 0.043), not other ROIs. In the table’s ΔHbO units (10⁻² mmol/L·mm), manual acupuncture exceeded rehabilitation for PFC change by 7.80 (95% CI, 5.13–10.48; Bonferroni-adjusted p < 0.001) and sham by 7.21 (4.54–9.89; p < 0.001). For M1, the corresponding differences were 3.62 (1.01–6.23; p = 0.020) and 3.31 (0.70–5.92; p = 0.039). Treat these as exploratory hemodynamic correlates, not evidence that a specific neural mechanism caused the functional gains.
Resting-state fNIRS connectivity: Four ROI-pair interactions survived FDR correction. In Fisher-z units scaled by 10⁻², changes in ipsilesional PFC–contralesional PFC connectivity were −41.49 with manual acupuncture, +33.59 with rehabilitation, and +21.09 with sham; the rehabilitation and sham changes exceeded the manual-acupuncture change by 75.09 (95% CI, 49.39–100.78) and 62.58 (36.89–88.28), respectively (both adjusted p < 0.001). For ipsilesional PFC–contralesional premotor cortex, the corresponding manual-acupuncture change was −10.33 versus +40.60 and +41.90; comparator-minus-acupuncture differences were 50.93 (25.20–76.66) and 52.22 (26.49–77.96), both adjusted p < 0.001. For contralesional PFC–contralesional premotor cortex, changes were −19.91, +30.41, and +20.81; differences were 50.32 (28.91–71.73) and 40.72 (19.31–62.12), both adjusted p < 0.001. Ipsilesional M1–S1 connectivity changed by +74.14 with manual acupuncture, +1.57 with rehabilitation, and +34.82 with sham; the manual-acupuncture advantage over rehabilitation was 72.58 (31.68–113.47; adjusted p = 0.002), while its comparison with sham was not significant (39.32; −1.58 to 80.21; adjusted p = 0.179). These mixed-direction connectivity findings do not, by themselves, establish “normalization” or a causal recovery pathway.

Adverse Events / Safety:

Three participants had mild subcutaneous bleeding at a needling site: 2/28 (7.14%) in the manual-acupuncture group, 1/28 (3.57%) in the sham group, and 0/28 in rehabilitation alone (3/84 overall, 3.57%). All resolved with routine local management and without sequelae. No serious adverse events, infection, fainting, stuck or broken needles, recurrent stroke, clinically significant deterioration, or adverse-event withdrawals were reported. With only 84 participants and four weeks of observation, this does not establish the absence of uncommon harms.

Figures

fNIRS setup, probe montage, and recording sequence — *Figure 1.* fNIRS acquisition setup, probe montage, and experimental procedure: a 10-second baseline, 180-second resting-state recording, 60-second rest interval, 10-second task baseline, and four cycles alternating 30 seconds of rest with 30 seconds of the composite affected-arm movement task.

CONSORT participant flow diagram for the three trial groups — *Figure 2.* CONSORT flow diagram. Ninety-six people were assessed, 12 were excluded before randomization, and 84 were randomized (28 per group); all randomized participants contributed endpoint data and were analyzed, including three assessed before the scheduled week-four visit after early discharge.

Surface EMG RMS before and after treatment during stretch and contraction — *Figure 3.* sEMG-derived RMS amplitudes before and after treatment: biceps, triceps, wrist flexor, and wrist extensor during passive stretch (A–D) and voluntary contraction (E–H). Asterisks denote significance thresholds: * p < 0.05; ** p < 0.01; *** p < 0.001.

Task-related cortical activation maps and ROI-level HbO changes — *Figure 4.* Cortical activation before and after treatment in the three groups. Panel A shows three-dimensional activation maps; panel B shows ROI-level ΔHbO values. PFC, prefrontal cortex; M1, primary motor cortex; PM, premotor cortex; S1, primary somatosensory cortex; ipsilesional and contralesional labels were recoded by lesion side.

Resting-state functional-connectivity networks and ROI-to-ROI heatmaps — *Figure 5.* Resting-state functional-connectivity network maps and ROI-to-ROI heatmaps. Panels A–D show manual acupuncture, E–H sham acupuncture, and I–L conventional rehabilitation at the two time points; panels M–N compare manual acupuncture with sham, and O–P compare manual acupuncture with rehabilitation, using Bonferroni-corrected treatment-effect comparisons. Red and blue scales show connectivity values or between-group treatment-effect differences.

Criticisms

Small, single-center, short-term trial: 84 participants at one Chinese hospital and four weeks of treatment cannot establish generalizability, durability, or uncommon harms. No follow-up beyond the intervention endpoint was reported.
Early/subacute population: Participants were within three months of stroke. The findings should not be carried over automatically to chronic stroke, other impairment severities, or people outside the age and eligibility criteria.
Blinding was incomplete: Practitioners were unblinded; the no-needling group necessarily knew it received no needling. Sham credibility and treatment expectancy were not formally assessed. Allocation guesses were not associated with actual assignment in the two needling groups (χ² = 3.40, p = 0.182), but that does not establish effective masking or eliminate expectation/performance bias.
Endpoint timing varied for three participants: One sham participant was assessed after three weeks and two manual-acupuncture participants on days 22 and 25 after early discharge. All were analyzed without imputation, but the endpoint was not exactly four weeks for those participants.
Many exploratory secondary comparisons: fNIRS interaction tests used FDR correction and post-hoc comparisons used Bonferroni adjustment; nevertheless, sEMG and neuroimaging findings are secondary and should be treated as hypothesis-generating, not as equivalent to the primary clinical endpoint.
Biomarker limitations: sEMG RMS was not normalized, leaving electrode-placement and skin–electrode variation as possible sources of between-session noise. sEMG and fNIRS do not fully measure lesion anatomy, corticospinal tract integrity, or fine hand kinematics; the two systems also lacked a shared hardware trigger/time base for temporally aligned analysis.
Clinical-importance threshold is context-dependent: The FMA-UE advantage is statistically clear in this sample, but the authors cite a 5–13-point range for clinically important differences depending on context. The manual-acupuncture-versus-sham confidence interval (4.55–12.59) includes values below the lower end of that range.
Safety inference is limited: Three mild bleeding events and no serious events in 84 people over four weeks are reassuring for this protocol, but cannot rule out rare events.

Funding

Supported by the Key Medical and Health Project of Guangzhou Panyu District Science, Technology, Industry, Commerce and Information Technology Bureau, grant 2022-Z04-107. The funding statement does not specify the funder's role. Trial registration: International Traditional Medicine Clinical Trial Registry, ITMCTR2025002121; registered November 3, 2025, before enrollment.

The paper

Authors. Yu Min, Zhendong Huang, Ju Sun, Zetong Peng, Hongli Guan, Yuying Gong, and Kaifeng Guo.
Title. Clinical and neurophysiological effects of manual acupuncture for upper-limb motor dysfunction after ischemic stroke: a three-arm randomized controlled trial.
Journal. Frontiers in Neurology.
Publication date / issue. August 25, 2026; volume 17, article 1886740.
DOI. 10.3389/fneur.2026.1886740
PMCID. PMC13549818
PMID. 42712438
License. Creative Commons Attribution 4.0 International (CC BY 4.0): license terms. Attribution to the authors and original publication is required.

Deep Dive — click to expand

What this is

This is a small, registered, single-center randomized trial asking whether a specific six-point manual acupuncture protocol adds short-term benefit to a standardized rehabilitation program after ischemic stroke. Its strongest evidence is the four-week FMA-UE difference versus both superficial nonacupoint needling and rehabilitation alone: 8.57 and 9.64 points, respectively. The clinical outcome supports further testing of this adjunct; the sEMG and fNIRS findings are exploratory correlates, not proof of how the benefit occurred.

1. Shadow Audit

The headline is not “acupuncture reverses stroke” and the biomarker figures do not prove neuroplasticity. All groups improved during early/subacute recovery and structured rehabilitation; the randomized comparisons estimate the additional short-term change associated with the manual-acupuncture protocol in this sample. Sham acupuncture did not differ significantly from rehabilitation alone, but shallow nonacupoint needling is not guaranteed to be physiologically inert, and the study did not formally establish sham credibility.

The fNIRS story is also mixed, not a single “more connectivity is better” result. Manual acupuncture increased task-related ipsilesional PFC and M1 activation, while some resting PFC–PFC and PFC–premotor connections changed less or declined relative to controls; ipsilesional M1–S1 connectivity increased versus rehabilitation but not significantly versus sham after correction. Those are exploratory patterns, not validated surrogate endpoints. Likewise, absolute unnormalized RMS changes—particularly during passive stretch—cannot be translated directly into stronger movement or lower spasticity.

2. Inversion Engine

The clinical interpretation would weaken if a larger, multicenter, adequately masked replication found that the manual-acupuncture advantage over sham was small and excluded a clinically meaningful effect. A concrete benchmark is the authors’ cited context-dependent FMA-UE meaningful-change range of about 5–13 points: the current manual-acupuncture-versus-sham estimate is 8.57 points, but its 95% CI runs from 4.55 to 12.59 and therefore includes an effect below five points. Replication with a point estimate below five and a confidence interval excluding benefits of five points or more would challenge the claim of a clinically important incremental effect. Loss of benefit at longer follow-up would separately overturn any claim of durable recovery.

3. Second-Order Catalyst

If the effect is real, the first group to act should be stroke-rehabilitation teams and investigators treating patients who resemble this cohort—clinically stable adults in the first three months after ischemic stroke with unilateral upper-limb impairment—not acute reperfusion services or chronic-stroke programs. The defensible immediate step is a monitored, consent-based adjunct within a research or service-evaluation pathway that preserves usual rehabilitation, uses the protocol as studied, and measures FMA-UE before and after treatment. This trial alone does not justify replacing rehabilitation, making acupuncture routine for every stroke survivor, or changing an acute-stroke protocol.

4. Asymmetric Leverage

The possible leverage is an added 8.57 FMA-UE points over a sham procedure and 9.64 over rehabilitation alone after four weeks, alongside a 7.32–11.29-point MBI advantage. If reproduced and durable, an adjunct that improves both arm impairment and daily activity could matter to patients. The cost side is also real: a 30-minute needling procedure, five-day-per-week treatment context, trained staff, and needling-related bleeding occurred in 3 of 56 participants assigned to a needling arm. The trial is too small and short to establish a population-level benefit–harm balance or service burden.

5. Paradigm Destroyer

The habit this paper should challenge is treating a statistically significant imaging or EMG change as mechanistic confirmation—or equating an RMS increase with clinical recovery. It should not displace the reflex to deliver structured rehabilitation: keep usual rehabilitation in place, and regard manual acupuncture only as a candidate adjunct whose clinical effect still needs replication and follow-up.

A cautious two-sentence protocol note would read: “Continue the planned task-oriented rehabilitation program. For an eligible patient considering adjunctive manual acupuncture, explain that one small four-week trial found a short-term FMA-UE advantage but did not establish long-term benefit; record baseline and endpoint motor function and monitor needling adverse events.”

MVP — Minimum Viable Proof

The minimum result worth testing at the bedside is the randomized four-week comparison: manual acupuncture plus the same rehabilitation produced 8.57 more FMA-UE points than sham plus rehabilitation (95% CI, 4.55–12.59; Bonferroni-adjusted p < 0.001; 28 participants per arm). To make this a practice-changing claim rather than a promising signal, a larger independent trial should reproduce a clinically important advantage with credible masking and show that it persists beyond the immediate intervention period.

Best Combination

Use this trial as one component of the broader post-stroke acupuncture evidence, not as a stand-alone practice standard. Its useful contribution is the three-arm design: the manual protocol was compared both with matched shallow needling and with rehabilitation alone, while sham and rehabilitation did not differ on the primary outcome. A synthesis should keep the protocol, stroke stage, rehabilitation dose, comparator credibility, outcome timing, and risk of bias visible rather than pooling different acupuncture approaches as though they were interchangeable. The next trial should be multicenter, prospectively registered, adequately powered for a prespecified clinically meaningful FMA-UE difference, and include longer follow-up and standardized adverse-event reporting.

Overvalue Warning

Do not describe the fNIRS or sEMG results as proof that acupuncture caused neuroplastic recovery. They are secondary exploratory measures, and no validated mediation or surrogate analysis established that they explain the FMA-UE change.
Do not call the result a proven antispastic or universally safe treatment. Passive-stretch RMS effects differed by muscle, RMS was unnormalized, and three mild bleeding events in a small four-week trial cannot rule out uncommon harms.

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.