The Time Is Brain Review

Scalp acupuncture coupled with speech-language therapy modulates the ventral-dorsal structural brain network in patients with post-stroke anomia: a randomized clinical and DTI neuroimaging trial

· DOI: 10.1186/s13020-026-01501-1

stroke deep-dive aphasia rehabilitation acupuncture

Reconstructed on October 14, 2026 for the missed October 9 publication slot. This is a replacement article, not a recovered original. Source selection: this paper appeared in the archived October 9 briefing; its selection is reconstructed from that briefing, not recovered from the failed original run.

Clinical Question (PICO)

In adults with first-ever left-hemisphere stroke and post-stroke anomia, did adding scalp acupuncture (SA) to conventional stroke care and speech-language therapy (SLT), compared with the same care and SLT alone, improve naming after four weeks and alter diffusion-tensor imaging (DTI) measures of language-network structure?[1]

Bottom Line

In a single-center randomized trial, 57 participants in the per-protocol clinical/MRI analysis (27 control, 30 SA) had greater immediate improvement in Western Aphasia Battery (WAB) naming total score with SA plus SLT than with SLT alone (median change 21 [IQR 14.75–41.50] vs 14 [8–21]; P=.009; reported effect size r=.342)[1]. Communication-domain SAQOL-39 change also favored SA (P=.036), while total SAQOL-39 did not (P=.183)[1]. The result is promising but not confirmatory: treatment was unblinded, there was no sham acupuncture, 28 of 85 randomized participants were not in the per-protocol analysis, follow-up ended immediately after treatment, and imaging findings are exploratory rather than proof of mechanism[1].

Design

  • Trial type: Single-center, 1:1 randomized controlled trial; allocation used sealed opaque envelopes[1]. Outcome assessors and statisticians were blinded; participants and therapists were not[1].
  • N: 85 patients were randomized (43 control, 42 SA, as shown in Figure 1)[1]. The per-protocol analysis included 27 control and 30 SA participants with complete clinical/MRI data[1].
  • Setting and enrollment: Rehabilitation Hospital Affiliated with Fujian University of Traditional Chinese Medicine; recruited July 2022–July 2023[1].
  • Analysis: Per-protocol on T1−T0 change scores. Between-group continuous tests were independent-samples t or Mann–Whitney U; within-group tests were paired t or Wilcoxon, with chi-square for categorical data. Imbalanced imaging measures used ANCOVA with baseline values as covariates. Two-sided P<.05; SPSS 26.0[1].
  • Primary outcome: WAB naming-subtest total score (maximum 100), assessed immediately after the four-week course[1]. The paper does not report a longer-term clinical follow-up[1].

Population

Inclusion Criteria

  • Left-hemisphere ischaemic or haemorrhagic stroke; WAB aphasia quotient <93.8 and BNT-30 spontaneous naming <20; age 45–85; first stroke 2 weeks to 6 months before enrollment; premorbid right-handedness, MRI eligibility, and ability to participate[1].

Exclusion Criteria

  • Aphasia from another disease, dysarthria-related speech difficulty, or severe auditory, visual or cognitive impairment preventing assessment or MRI[1].

Baseline Characteristics (per-protocol groups)

  • Mean age: 58.85±13.20 years control and 56.60±15.38 SA[1]. Median disease duration: 36 days (IQR 25–68) and 41.5 days (29.5–84.5), respectively[1].
  • Stroke type was ischaemic/haemorrhagic in 21/6 control and 21/9 SA; aphasia type was fluent/non-fluent in 6/21 and 8/22. Baseline WAB naming total was 8 (0–24) control and 9 (4.5–33) SA; these differences were not statistically significant in Table 1[1].
  • Baseline WAB aphasia quotient was 24.6 (13.5–45.4) control and 33.7 (19.7–49.2) SA (P=.237). Participants were 24 men/3 women in control and 22/8 in SA; median education was 9 years in each group. Figure 1 shows 43 control and 42 SA participants randomized, with 16 and 12 discontinuations, respectively; the 27/30 complete-data groups formed the per-protocol analysis[1].

The authors report no statistically significant baseline differences in age, sex, education, disease duration, stroke type, aphasia type, WAB AQ or BNT spontaneous naming (all P>.05); several tract and network MRI measures nevertheless differed[1].

Interventions

  • Control: Conventional stroke treatment, health education and SLT for 30 minutes once daily, 5 days per week for 4 weeks[1].
  • SA + SLT: Same conventional care and SLT, with scalp acupuncture during the speech session. A professional therapist used sterile 0.25×40-mm needles at the left lower 2/5 anterior oblique temporal line and anterior temporal line (15–30° angle, about 15–25 mm deep); initial twirling was about 200/min for 1 minute to elicit Deqi. Needles were retained 30 minutes and manually twisted twice more at 10-minute intervals while SLT proceeded. Treatment was once daily, 5 days/week for 4 weeks[1].

Outcomes

Between-group results below are changes from baseline to the immediate post-treatment assessment, reproduced as reported; values are median (IQR) unless marked mean±SD[1]. Effect size and two-sided P value are included[1]. The article does not provide confidence intervals for these clinical between-group effects[1].

Outcome Control change SA change Effect size P
WAB naming total 14 (8–21) 21 (14.75–41.50) r=.342 .009
Object naming 9 (4–12) 12.5 (8.5–24.5) r=.336 .011
Word fluency 2 (0–5) 3.5 (0–6) r=.082 .541
Sentence completion 2 (0–2) 4 (1.5–4.25) r=.349 .008
Responsive speech 2 (0–2) 4 (0.75–6.25) r=.328 .013
BNT spontaneous naming 3 (1–4) 3.5 (2.75–6) r=.268 .043
BNT selective naming 1 (0–4) 1.5 (−0.25–3) r=.065 .629
BNT prompt naming 1 (0–3) 1 (−1.25–3.25) r=.013 .926
Misnamed items −7 (−10 to −3) −5 (−9.25 to −2.5) r=.051 .705
SAQOL-39 physical .94 (.35–1.47) 1.15 (source IQR malformed) r=.132 .322
SAQOL-39 communication .92±.63 1.29±.66 Hedges’ g=.564 .036
SAQOL-39 psychosocial .99±.75 1.37±.75 Hedges’ g=.509 .057
SAQOL-39 total 36.30±22.30 45.13±26.70 Hedges’ g=.353 .183

Secondary nominal gains were limited to sentence completion, responsive speech, BNT spontaneous naming and SAQOL communication; total SAQOL-39 and several other clinical outcomes were nonsignificant. With many comparisons and no reported multiplicity adjustment, treat these secondary signals as exploratory[1].

BNT-30 uses 30 drawings; SAQOL-39 covers physical, communication and psychosocial domains. WAB AQ was listed as monitored, but no between-group WAB AQ result appears in the main table. End-of-course clinical assessment and MRI leave durability unknown[1].

The sample required first stroke, left-hemisphere lesion, premorbid right-handedness, MRI eligibility and disease duration of 2 weeks to 6 months. That selection limits extrapolation to chronic or recurrent stroke and to patients unable to complete MRI-based assessment[1].

The SA group also had smaller changes in small-worldness (P=.008) and normalized clustering coefficient (P=.017), and greater changes in clustering coefficient (P=.011) and local efficiency (P<.001)[1]. Within SA, change in small-worldness correlated with WAB naming change (r=−.451, P=.027)[1]. For axial diffusivity, between-group changes favored smaller SA-group changes in left IFOF (P<.001), ILF (P=.001), UF (P=.003) and SLF (P=.002), with Bonferroni-corrected P values; left UF axial-diffusivity change correlated with WAB naming change (r=−.440, P=.031)[1].

For mean diffusivity, changes were lower in SA in left IFOF and SLF (both P<.001) and left ILF (P=.001); for radial diffusivity, left IFOF (P<.001), ILF (P=.006) and SLF (P<.001). These comparisons were reported as Bonferroni corrected; adjusted correlations with naming were not significant. In exploratory regression, SA assignment was associated with WAB naming improvement (B=12.931, standardized β=.378, P=.005); adding selected imaging changes raised R² from .269 to .292, but the increment was not significant (ΔR²=.023; F-change=.487; P=.693), and no imaging measure was independently associated with improvement[1].

These DTI results do not form a simple “more connectivity is better” pattern: SA showed smaller changes in small-worldness and normalized clustering but greater changes in clustering coefficient and local efficiency. The reported correlations with naming were within the SA group for small-worldness and left UF axial diffusivity; they were not independent predictors in the adjusted regression. Read the imaging plots as treatment-associated correlates, not as a validated neural mediator or response biomarker[1].

Adverse Events / Safety: No adverse-event or safety results are reported in the article[1]. This is an absence of reporting, not evidence that acupuncture was safe or risk-free[1].

Figures

Study assessment and participant-flow diagram for the randomized scalp-acupuncture trial
Figure 1. Study assessment/MRI schedule and participant flow; the diagram shows 85 randomized participants and the 57-person per-protocol imaging sample[1]. Source: Song et al., Chinese Medicine (2026), CC BY 4.0.
Between-group changes in structural network topology and correlation with WAB naming change
Figure 2. Group comparisons of small-worldness, normalized clustering, clustering coefficient and local efficiency, plus the within-SA association between small-worldness and WAB naming change[1]. Source: Song et al., Chinese Medicine (2026), CC BY 4.0.
Axial-diffusivity comparisons in four left-hemisphere language tracts, with left uncinate-fasciculus correlation and tract reconstructions
Figure 3. Axial-diffusivity change comparisons in left UF, SLF, ILF and IFOF; the figure also shows the left-UF/WAB naming correlation and tract reconstructions[1]. Source: Song et al., Chinese Medicine (2026), CC BY 4.0.
Mean- and radial-diffusivity changes in left SLF, ILF and IFOF, with tract reconstructions
Figure 4. Mean- and radial-diffusivity change comparisons in left SLF, ILF and IFOF, with representative tract reconstructions[1]. Source: Song et al., Chinese Medicine (2026), CC BY 4.0.

Criticisms

  • The 85 randomized participants were analyzed per protocol; the immediate MRI/clinical sample was 57[1]. Without an intention-to-treat estimate, attrition could affect the apparent treatment contrast[1].
  • Figure 1 lists discontinuation reasons as transfers/lost visits, disease aggravation, or refusal of the second MRI: 3/1/12 in control and 4/3/5 in SA. No attrition-adjusted sensitivity analysis is reported[1].
  • No sham intervention or participant/therapist blinding leaves expectation, performance and added-attention effects unresolved, despite blinded outcome assessors and statisticians[1].
  • The trial is small and single-center, and measured outcomes only at the end of four weeks[1]. The durability and clinical importance of the score difference are unknown; the paper says no WAB naming minimal clinically important difference is established[1].
  • Many clinical endpoints were tested at a nominal P<.05; the article does not report a multiplicity adjustment for these clinical comparisons[1]. Several secondary results are near .05, while total SAQOL-39 was not statistically significant[1].
  • Imaging findings are surrogate/exploratory: they did not significantly improve regression fit, and no imaging measure independently predicted naming improvement[1]. Baseline FA differed in left IFOF, ILF and UF, and network clustering/local efficiency also differed; ANCOVA addressed imbalanced imaging metrics, but residual uncertainty remains in this small complete-case sample[1].
  • Table 2 prints a malformed SA-group physical SAQOL-39 change IQR (“0.47.1.47”); no correction is inferred here[1]. The P value and effect size are reproduced as printed[1].

Funding

Supported by Fujian Province Science and Technology Planning (2023Y0035), Fujian Province Natural Science Foundation Young Investigator Project (2024J010033), Fujian Province innovation joint funds (2025Y9601), Fujian Province Young Eagle Program, and National Natural Science Foundation of China (82074512)[1]. Authors declared no competing interests[1].

The paper

  • Authors. Jian Song, Ruiming Nan, Jinglei Ni, Linsong Chai, Lanlan Zhang, Mengxue Wang, Nizhuan Wang, Bingbing Lin and Jia Huang.
  • Title. Scalp acupuncture coupled with speech-language therapy modulates the ventral-dorsal structural brain network in patients with post-stroke anomia: a randomized clinical and DTI neuroimaging trial.
  • Journal. Chinese Medicine.
  • Year. 2026.
  • DOI. 10.1186/s13020-026-01501-1
  • PMCID / PMID. PMC13647535 / 42850647.
  • Registration. Chinese Clinical Trial Registry, ChiCTR2200062146.
  • License. Creative Commons Attribution 4.0 International (CC BY 4.0); the paper states third-party material is included unless a credit line says otherwise.
  • Primary source. [1] https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13647535/fullTextXML — Scalp acupuncture coupled with speech-language therapy modulates the ventral-dorsal structural brain network in patients with post-stroke anomia: a randomized clinical and DTI neuroimaging trial.

Deep Dive — click to expand

What this is

This RCT tests scalp acupuncture as an SLT add-on after stroke. Its per-protocol naming signal is favorable; DTI findings are secondary correlates, not proof of mechanism[1].

1. Shadow Audit

The clean headline is a median WAB naming change of 21 with SA versus 14 in control, P=.009[1]. The less visible denominator is 57 analyzed from 85 randomized, with 27 control and 30 SA participants in the complete-data per-protocol sample[1]. Randomization protects the assignment comparison only if the analysis preserves it; a per-protocol subset can reintroduce selection[1].

The second shadow is the control condition[1]. Assessors and statisticians were blinded, but participants and therapists knew whether needles were used, and there was no sham procedure[1]. The trial therefore estimates the effect of offering this acupuncture-plus-SLT package under these conditions—not a needle-specific physiological effect isolated from expectation, attention or performance differences[1].

2. Inversion Engine

Confidence would fall if a larger intention-to-treat trial with a credible sham, blinded outcome assessment and a prespecified patient-important endpoint failed to reproduce the naming advantage or showed that it faded after treatment[1]. This study supplies no long-term estimate and no established minimal clinically important difference for its WAB naming subtest; the reported P value alone cannot define a clinically meaningful threshold[1].

3. Second-Order Catalyst

The immediate next step is a better confirmatory trial, not routine adoption as an established aphasia treatment[1]. Keep structured SLT as the common foundation, match contact time and therapist attention, document adherence and adverse events, and assess communication and naming after a meaningful follow-up interval[1]. The present report provides a four-week signal but not durability or a complete safety estimate[1].

4. Asymmetric Leverage

The primary score and several naming subtests favored SA, but results were not uniform: word fluency, selective and prompt naming, misnamed items, physical and psychosocial quality-of-life domains, and total SAQOL-39 were not statistically significant[1]. Communication-domain SAQOL-39 favored SA (P=.036), while the total score did not (P=.183)[1]. The narrower favorable outcomes are hypothesis-generating within a small set of many comparisons, not proof of broad language or quality-of-life benefit[1].

5. Paradigm Destroyer

The study does not show that network-imaging changes explain the clinical result[1]. The imaging variables added only ΔR²=.023 and the increment was not significant (P=.693); none independently predicted naming improvement after adjustment[1]. The habit to retire is treating an attractive DTI correlation or group-difference plot as mechanism validation[1].

MVP — Minimum Viable Proof

A minimally persuasive confirmation would reproduce the primary naming result in an intention-to-treat population, with a credible sham and prespecified clinical significance, while documenting harms and showing whether benefit persists beyond the immediate post-course test[1]. To elevate DTI from correlate to mechanism or predictor would require independent validation beyond this sample[1].

Best Combination

The strongest design element to retain is the shared SLT in both groups: it frames acupuncture as an adjunct rather than a replacement[1]. Combine that with concealed randomization, matched expectations/contact, complete follow-up, patient-centered communication measures, and preregistered imaging analyses[1]. This paper's exploratory regression already cautions that its measured imaging changes did not improve prediction beyond clinical covariates[1].

Overvalue Warning

  1. Do not generalize the per-protocol 57-person result to all post-stroke aphasia, chronic stroke, other acupuncture protocols or acupuncture without SLT[1].
  2. Do not call the intervention proven safe: the article reports no adverse-event results, and lack of reporting cannot establish absence of harm[1]. Do not call DTI changes proof of how the clinical signal occurred[1].

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.