Specific mode acupoint electrical stimulation combined with NGF for neurological recovery in ischemic stroke convalescence: A randomized controlled study

Abstract

Background

Nerve growth factor (NGF) has neurorestorative potential after ischemic stroke, but its efficacy may be limited by the blood-brain barrier (BBB). Preclinical studies suggested that specific mode acupoint electrical stimulation (SMES) may transiently increase BBB permeability and facilitate neurotrophic factor delivery. This study evaluated the efficacy and safety of SMES combined with NGF in ischemic stroke and explored treatment-associated functional brain changes.

Methods

In this multicentre, randomized, placebo-controlled trial, adults aged 30–80 years with first-ever ischemic stroke, 14 days to 6 months after onset, modified Rankin Scale (mRS) score 3–4, and MMSE score < 27 were randomly assigned (1:1:1:1) to acupuncture + placebo, acupuncture + NGF, SMES + placebo, or SMES + NGF. NGF or placebo was administered by daily intramuscular injection for 28 days, followed by acupuncture or SMES at GV20 and GV26. The primary efficacy outcome was longitudinal change in mRS score. Secondary outcomes included the Modified Barthel Index, Fugl-Meyer Assessment, Tinetti Performance-Oriented Mobility Assessment, Montreal Cognitive Assessment, and Loewenstein Occupational Therapy Cognitive Assessment. Outcomes were assessed at baseline, week 2, and week 4, with mRS and MoCA further assessed during follow-up. Exploratory rs-fMRI and fNIRS assessments were performed at baseline and week 4.

Results

Among 324 randomized participants, 296 completed treatment and follow-up. Repeated-measures ANOVA showed significant time and group-by-time effects for mRS, indicating different recovery trajectories among groups. In factorial analysis, NGF had a significant main effect on mRS improvement (P = 0.001), whereas the SMES main effect and SMES-by-NGF interaction were not significant. Cognitive outcomes showed a clearer combined-treatment pattern: the SMES + NGF group had the greatest improvements in MoCA and LOTCA at week 4, and the SMES-by-NGF interaction for MoCA was significant at week 4 (P = 0.019) and during follow-up (P < 0.001). Exploratory neuroimaging analyses suggested treatment-associated functional brain changes. Six adverse events occurred, with no serious treatment-related adverse events.

Conclusions

NGF-containing interventions were associated with improved neurological disability after ischemic stroke, while SMES combined with NGF showed additional benefits for cognitive recovery. SMES combined with NGF may represent a promising adjunctive strategy for post-stroke rehabilitation.

Trial Registration

The protocol was registered in the Clinical Trials repository under NCT05231694.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12916-026-05073-7.

Background

Ischemic stroke can lead to accelerated brain aging and neurological dysfunction, and in severe cases, even death [1, 2]. Stroke has become a major threat to human health. Failure to adequately reverse neurological damage in the early stages of ischemic stroke often leads to neurological deficits such as motor dysfunction and/or vascular dementia [3, 4]. Current treatment options for nerve injuries include vasodilation, neurotrophic interventions, neurological rehabilitation, and electroacupuncture [5–10]. Although progress has been made in neurological rehabilitation therapies, many stroke patients continue to experience motor or cognitive impairments during rehabilitation [11]. Therefore, enhancing the clinical efficacy of neurological recovery in stroke patients during the recovery period is imperative.

Nerve growth factor (NGF) is a neurotrophic substance that plays crucial roles in nerve repair and protection, cholinergic neurone nourishment, neuronal survival [12, 13]. NGF has shown promise in cognitive rehabilitation; However, its effectiveness in clinical practice has proven unsatisfactory [14]. The primary obstacle is that its large molecular weight (13.4 KD) makes it challenging to cross the blood-brain barrier (BBB) and achieve optimal therapeutic concentrations within the central nervous system [15]. This limitation directly matched the rationale for using specific mode acupoint electrical stimulation (SMES) as a potential non-invasive approach to facilitate neurotrophic factor delivery across the BBB, as suggested by preclinical evidence. NGF was chosen rather than other neurotrophic factors such as BDNF, NT-3, or NT-4 because the present study was designed as a clinical translation of our previous NGF-based preclinical work. In previous animal studies, the SMES stimulation parameters were specifically optimized using NGF as the target macromolecular agent. Therefore, this trial focused on NGF rather than comparing multiple neurotrophins.

Current methods for opening the blood-brain barrier are often complex, typically requiring rigorous and cumbersome technical procedures, and carry risks such as cerebral edema and neuronal apoptosis [16–20]. There remains a significant gap before they can be put into practical clinical use. Electroacupuncture, a technique developed from traditional Chinese medicine, has been used alongside Western medicine in ischaemic stroke recovery. Preliminary study indicates that a specific electroacupuncture stimulation method applied to the Baihui and Shuikou acupoints can downregulate the expression of tight junction proteins ZO-1 and Occludin, thereby reversibly modulating BBB permeability, increasing the penetration of NGF into the brain, and enhancing spatial learning and memory abilities in rats. Meanwhile, this method did not induce adverse reactions such as cerebral edema, activation of glial cells (astrocytes and microglia), or cell apoptosis [15, 21–23].

These preclinical findings provided the rationale for testing whether SMES combined with NGF could improve clinical recovery after ischemic stroke. However, direct clinical evidence regarding the efficacy of this combined intervention remains limited, and it remains unclear whether the potential benefits are reflected mainly in global disability, motor recovery, activities of daily living, balance and gait, cognitive function, or treatment-associated functional brain changes. Therefore, a randomized clinical trial with multidimensional clinical and exploratory neuroimaging outcomes is needed.

Therefore, this study aims to evaluate the clinical efficacy and safety of SMES combined with NGF, and to investigate whether clinical improvement is accompanied by functional brain changes. The trial hypothesis is that for patients in the rehabilitation phase of ischemic stroke, SMES combined with NGF will produce more significant improvements in neurological function compared to control interventions and induce treatment-related functional brain alterations. To comprehensively assess post-stroke disability and functional dependence, the Modified Rankin Scale (mRS) was selected as the primary clinical outcome measure; secondary outcome measures included clinically relevant recovery domains such as activities of daily living, motor balance function, and post-stroke cognitive function. For exploratory neurophysiological outcome measures, we utilized techniques such as functional near-infrared spectroscopy (fNIRS) and resting-state functional magnetic resonance imaging (rs-fMRI) to determine whether clinical improvement was associated with cortical hemodynamic responses and functional brain network changes.

Methods

Study design

This study is a multicenter, randomized, placebo-controlled, prospective clinical trial, which was registered on ClinicalTrials.gov prior to subject enrollment (Registration Number: NCT05231694), and the relevant study protocol has been publicly released [24]. Ethical approval was obtained from the ethics committees of the participating institutions (approval numbers: ZSLL-KY-2022-002-01, EJ-NBEY-KY-2022-027-01, and 20220210-1). The trial was conducted in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants before any study procedure [25]. Participants were free to withdraw from the trial at any time, and adverse events were monitored throughout the study period. The study consisted of a 4-week treatment phase, with clinical assessments at baseline, week 2, and week 4, followed by follow-up assessments at 4 and 12 weeks after treatment completion.

Participants

Participants were recruited from three tertiary hospitals in Zhejiang, China, between March 2022 and April 2024. Inclusion criteria were: (1) Age 30 to 80 years; (2) First-ever ischemic stroke confirmed by Computed Tomography (CT) or Magnetic Resonance Imaging (MRI); (3) 14 days to 6 months from stroke onset to screening; This time window was prespecified in the published protocol and was intended to capture the post-acute/convalescent recovery period of ischemic stroke, during which functional and cognitive rehabilitation remains clinically relevant; (4) Modified Rankin Scale (mRS) score of 3 to 4 at screening; (5) Mini-Mental State Examination (MMSE) score < 27 at screening; (6) Ability and willingness to comply with the treatment schedule and outcome assessments, with written informed consent obtained before enrollment.

Exclusion criteria were: (1) Transient ischemic attack, cerebral haemorrhage, and other cerebrovascular diseases; (2) Unstable medical conditions, including myocardial infarction and severe pulmonary disease; (3) Brain tumours, recurrent stroke-related cerebral infarction, or previous brain surgery or trauma; (4) Visual or auditory impairment, aphasia, agnosia, severe hemiplegia, or affected limb dysfunction before stroke; (5) Pre-stroke cognitive impairment, including Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) ≥ 3.19; (6) Pregnancy or lactation; (7) Skin infection at the acupuncture site, needle sickness, or needle phobia; (8) Implantable cardioverter-defibrillators, pacemakers, or conductive metallic foreign bodies; (9) History of NGF allergy; (10) Participation in another clinical trial within the past 3 months or at the time of screening. Additional detailed eligibility criteria are provided in the published protocol [24]. Sex, race, and ethnicity were self-reported, with options for sex limited to “male” or “female”.

Criteria for discontinuation included: (1) Intolerable severe adverse effects related to the intervention; (2) A voluntary request by the participant to withdraw from the study. Participants who discontinued treatment were retained in follow-up and included in the intention-to-treat analysis whenever possible.

Randomisation and masking

The random allocation sequence was generated by an independent researcher who was not involved in participant recruitment, intervention delivery, outcome assessment, or data analysis. After written informed consent had been obtained and eligibility confirmed, participants were assigned in a 1:1:1:1 ratio to one of four groups—acupuncture plus placebo, acupuncture plus NGF, SMES plus placebo, or SMES plus NGF-through the eBalance randomisation system (Randomization and Trial Supply Management, Taimei Technology, Shanghai, China). Allocation concealment was maintained until assignment, as investigators responsible for screening and enrollment had no access to the randomization sequence or to upcoming group allocations. Treatment assignments were disclosed only after enrollment and randomization had been completed, and only to the operators responsible for administering the assigned intervention. This included the trained nurses who administered the intramuscular NGF or placebo injections and the operators who delivered conventional acupuncture or SMES. Participants, outcome assessors, independent statisticians, and study staff involved in data analysis remained blinded to treatment allocation throughout the trial.

Procedures

For each dose, 20 µg of NGF (NO. S20060051, Jinlujie, Hiteck Biopharmaceutical Co., Ltd., Wuhan, China) was dissolved in 2 mL sterile water for injection. Placebo consisted of 2 mL normal saline. During hospitalization, all interventions were delivered at the patient’s bedside in the participating study hospitals. Intramuscular injections into the gluteus maximus were administered once daily for 28 days by professionally trained nurses according to a standardized study protocol [24]. This duration was selected because it corresponded to one complete 28-day course of NGF administration according to the drug instructions and the published trial protocol. During this period, NGF or placebo was administered once daily, followed immediately by acupuncture or SMES, thereby ensuring a consistent intervention sequence and schedule across the four groups.

Conventional acupuncture or SMES was performed immediately after NGF or placebo injection. For conventional acupuncture, two stainless steel needles (Hua Tuo brand, Suzhou Medical Supplies Company Ltd., Jiangsu, China) were inserted into GV20 (Baihui) and GV26 (Shuigou), and manual stimulation was applied until the participant experienced deqi sensations such as distension, heaviness, or soreness.

For SMES, the same acupoints (GV20 and GV26) and basic acupuncture procedure were used, after which an electroacupuncture instrument (HANS-200 A, Nanjing Jisheng, Ltd., China) was connected to the needles to deliver stimulation at 3 mA and 2/100 Hz for 40 min (a self-made relay cycled power to the electrode for 6s on and off). These stimulation parameters were selected based on previous preclinical studies and the published trial protocol, and were delivered by trained operators under standardized bedside monitoring to ensure patient safety and tolerability [24].

All participants received standard post-stroke rehabilitation during hospitalization and follow-up, as clinically indicated. Rehabilitation programs were prescribed by rehabilitation physicians and delivered by trained rehabilitation therapists according to local clinical practice and each participant’s neurological deficits, motor function, balance, activities of daily living, and cognitive status. Routine rehabilitation generally included physical therapy, limb function training, balance and gait training, activities-of-daily-living training, and cognitive rehabilitation when clinically indicated. These rehabilitation procedures were not determined by trial allocation and were provided similarly across the four treatment groups. Participants were advised to maintain their usual rehabilitation and exercise routines during follow-up, and concomitant rehabilitation procedures were recorded during the study period.

Following completion of the 4-week intervention, follow-up visits were conducted at 4 and 12 weeks to assess the durability of treatment effects and safety. Participants were monitored for adverse events throughout the study period. Participants were free to discontinue treatment or withdraw from the study at any time. To minimize loss to follow-up, participants who discontinued treatment were invited to continue follow-up assessments if they were willing to do so [21].

Outcomes

According to the published protocol, the primary endpoint was the basic cure rate, defined as the proportion of participants with mRS ≤ 2 [24]. In the present report, the primary efficacy analysis focused on longitudinal changes in mRS score, which provide a more granular assessment of neurological recovery over time. The protocol-defined basic cure rate was additionally analyzed as a supportive outcome. The secondary outcome indicators were the MoCA, Loewenstein Occupational Therapy Cognitive Assessment (LOTCA), Modified Barthel Index (MBI), Fugl-Meyer Assessment of Motor Function Score (FMA), and Tinetti Performance-Oriented Mobility Assessment (POMA). These secondary outcomes were prespecified in the published protocol and were selected to provide a multidimensional assessment of post-stroke recovery. MBI was used to assess independence in activities of daily living. FMA to quantify motor function. POMA to evaluate balance and gait performance, which is clinically relevant to mobility, fall risk, and functional ambulation after stroke and is not fully captured by mRS, MBI, or FMA alone. MoCA and LOTCA were selected to assess cognitive domains relevant to post-stroke cognitive impairment. MoCA and LOTCA, rather than MMSE, were prespecified as cognitive efficacy outcomes because they provide a broader assessment of cognitive domains relevant to post-stroke cognitive impairment. MMSE was administered only at screening/baseline for eligibility assessment and baseline cognitive characterization; Therefore, it was not included in the longitudinal outcome table. Data were collected by an independent outcome evaluator at baseline, week 2, and week 4 (end of treatment). During the follow-up phase, outcome assessors evaluated changes in patients’ conditions using the mRS scale and MoCA scale.

Resting-state functional magnetic resonance imaging (rs-fMRI) and functional near-infrared spectroscopy (fNIRS) were performed at baseline and week 4 to evaluate changes in brain function. The rs-fMRI assessment was conducted under resting-state conditions; Participants were instructed to remain still with their eyes closed, stay awake and relaxed, and minimize structured thinking. Resting-state BOLD and 3D-T1WI sequences were acquired, and ALFF, ReHo, and FC were calculated.

The fNIRS assessment was task-based. During the cognitive task, participants performed a color-word Stroop task consisting of neutral, congruent, and incongruent conditions. Oxyhaemoglobin (HbO₂) responses were analyzed separately for these Stroop conditions, with the incongruent condition regarded as the principal cognitive-control condition and the neutral and congruent conditions used as comparison conditions. The fNIRS protocol also included upper-limb and lower-limb movement tasks, as described in the published protocol [24].

Brain imaging assessment

fNIRS assessment and preprocessing

Task-based functional near-infrared spectroscopy (fNIRS) was performed at baseline and week 4 in a subset of participants to assess cortical haemodynamic responses during the Stroop task. The Stroop paradigm included neutral, consistent, and inconsistent conditions, with the inconsistent condition considered to impose higher cognitive-control demand. Participants were seated comfortably and instructed to complete the task while minimizing head and body movement. fNIRS optodes were positioned to cover frontal, prefrontal, and sensorimotor cortical regions relevant to cognitive control, executive processing, and sensorimotor integration. HbO₂ was used as the primary haemodynamic index for subsequent analyses.

Raw fNIRS data were processed using MATLAB and NIRS_KIT. Bad channels were inspected and excluded, and preprocessing included detrending, motion correction using the temporal derivative distribution repair method, and band-pass filtering with a third-order Butterworth IIR filter. For each channel and Stroop condition, HbO₂ responses were segmented from − 0.5 to 2 s relative to task onset, with the − 0.5 to 0 s interval used for baseline correction. The mean HbO₂ response during the 0 to 2 s task period was extracted as the channel-wise statistical index. Participants with incomplete pre- or post-treatment recordings or poor data quality were excluded from the fNIRS analysis.

rs-fMRI acquisition and preprocessing

rs-fMRI was performed at baseline and week 4 in a selected imaging subgroup from the acupuncture + NGF and SMES + NGF groups. To minimize lesion-related distortion of cortical BOLD signals, only participants with subcortical ischemic lesions and no cortical involvement were included. Participants were instructed to keep their eyes closed, remain awake and relaxed, and avoid structured thinking during scanning.

Images were acquired using a 1.5-T MRI scanner (uMR586, United Imaging Healthcare, Shanghai, China). Resting-state BOLD images were acquired with an echo-planar imaging sequence: repetition time/echo time = 2000/30 ms, slice thickness = 4.0 mm, no interslice gap, 27 slices, flip angle = 90°, field of view = 240 × 240 mm², and matrix = 64 × 64. High-resolution 3D T1-weighted gradient-echo images were acquired with repetition time/echo time/flip angle = 1900/2.52 ms/15°, field of view = 240 mm, slice thickness = 1.0 mm, no interslice gap, 172 slices, and matrix = 256 × 256.

DICOM images were converted to NIfTI format using dcm2niix. Preprocessing was performed using MATLAB, SPM12, and DPABI, including removal of the first 10 volumes, slice-timing correction, realignment, co-registration, segmentation, DARTEL-based normalization to Montreal Neurological Institute space, resampling to 3 × 3 × 3 mm³, smoothing with a 6-mm Gaussian kernel, linear detrending, and nuisance regression of white matter signal, cerebrospinal fluid signal, and 24 head-motion parameters. Participants with head motion greater than 3 mm translation or 3° rotation were excluded.

ALFF, ReHo, and seed-based functional connectivity were calculated to assess spontaneous neural activity, regional synchronization, and functional connectivity, respectively. ALFF was calculated in the 0.01–0.08 Hz band, ReHo was calculated using Kendall’s coefficient of concordance and standardized by the global mean, and FC was calculated using predefined AAL atlas seed regions, including the precentral gyrus, hippocampus, and postcentral gyrus. These seed regions were selected because of their relevance to motor function, somatosensory integration, and post-stroke cognitive and memory recovery. Correlation coefficients were Fisher z-transformed for group-level analyses. Statistical maps were thresholded at voxel-level P < 0.005 and cluster-level P < 0.05 with family-wise error correction.

Statistical analysis

Sample size was calculated according to the published protocol using the protocol-defined primary endpoint, the basic cure rate. The anticipated cure rates were 35%, 40%, 50%, and 75% in the acupuncture + placebo, acupuncture + NGF, SMES + placebo, and SMES + NGF groups, respectively. With a two-sided α level of 0.05, 80% power, and an assumed 20% dropout rate, the planned sample size was 72 participants per group, or 288 participants in total. The final randomized sample exceeded this target.

The primary efficacy analysis was performed on the full analysis set according to the intention-to-treat principle, with participants analyzed according to randomized assignment. Missing values for the primary mRS outcome were imputed using last observation carried forward; secondary outcomes were analyzed using available cases. Longitudinal change in mRS score was analyzed using repeated-measures ANOVA, with treatment group as the between-participant factor and assessment time point as the within-participant factor. Group, time, and group-by-time effects were reported with F and P values. When the interaction was significant, corrected post-hoc comparisons were performed at post-baseline time points. Because mRS is ordinal, a non-parametric sensitivity analysis was performed using Kruskal-Wallis tests and Dunn’s Bonferroni-corrected post-hoc tests for mRS change scores. Secondary outcomes, including MoCA, LOTCA, MBI, FMA, and POMA, were analyzed using the same repeated-measures framework when appropriate, and the protocol-defined basic cure rate was analyzed as a supportive categorical outcome.

Exploratory fNIRS and rs-fMRI analyses were performed to assess treatment-associated functional brain changes. fNIRS analyses focused on Stroop task-evoked HbO₂ responses and were interpreted as exploratory because no channel-wise findings survived FDR correction. rs-fMRI analyses assessed ALFF, ReHo, and seed-based functional connectivity, with statistical maps thresholded at voxel-level P < 0.005 and cluster-level P < 0.05 with family-wise error correction. All tests were two-sided, and P < 0.05 was considered statistically significant. Analyses were performed using SAS version 9.3 and SPSS version 26.0. Data generated or analyzed during the study are available from the corresponding author by request.

Results

In total, 583 patients were assessed for eligibility, and 324 eligible patients were randomly assigned to the four treatment groups. During treatment, 28 participants withdrew or discontinued the intervention, leaving 296 participants who completed the intervention and follow-up assessments: 73 in the acupuncture + placebo group, 74 in the acupuncture + NGF group, 75 in the SMES + placebo group, and 74 in the SMES + NGF group (Fig. 1).

Fig. 1.

Study flow diagram

The baseline characteristics of these participants are shown in Table 1. Most demographic and clinical characteristics were comparable among the four groups, including sex, age, educational attainment, blood pressure, injury side, baseline neurological and cognitive assessment scores, laboratory test results, and comorbidities. Cerebral infarction focus differed significantly among the groups (P = 0.012). In addition, although time since stroke onset did not differ significantly when analyzed as a continuous variable (P = 0.164), the distribution of participants enrolled within versus beyond 90 days after stroke onset differed among the groups (P = 0.019). These variables were considered in the interpretation of the outcome analyses.

Table 1.

Baseline Characteristics of the Patients

Characteristic	Acupuncture+placebo	Acupuncture+NGF	SMES+placebo	SMES+NGF	P value
	(n=73)	(n=74)	(n=75)	(n=74)
Sex, n (%)					0.254
Male	41(56.2)	49(66.2)	51(68.0)	41(55.4)
Female	32(43.8)	25(33.8)	24(32.0)	33(44.6)
Mean age (SD), y					0.094
	64.59(10.9)	68.36(8.9)	65.35(11.2)	64.72(10.5)
Educational attainment, n (%)					0.883
Illiterate / no formal education	19 (26.0)	25 (33.8)	26 (34.7)	23 (31.1)
Elementary school	14 (19.2)	17 (23.0)	17 (22.7)	15 (20.3)
Junior high school	14 (19.2)	9 (12.2)	9 (12.0)	11 (14.9)
High school	16 (21.9)	10 (13.5)	10 (13.3)	10 (13.5)
Bachelor’s degree	10 (13.7)	13 (17.6)	13 (17.3)	15 (20.3)
Mean blood pressure (SD)					0.779
Systolic BP, mmHg	143.5(18.2)	144.6 (17.6)	143.2 (18.4)	143.8 (18. 1)
Diastolic BP, mmHg	79.5 (11.4)	75.8 (13. 1)	76. 1 (12.4)	78.3 (11.7)
Mean heart rate (SD), bpm	73. 1(11.4)	71.9(12.3)	74.3(11.8)	71.7(12.7)
Cerebral infarction focus, n (%)					0.012*
Cerebral cortex	38(46.9)	41(50.6)	35(43.2)	34(42.0)
Basal ganglia	23(28.4)	27(33.3)	29(35.8)	30(37.0)
Cerebellum	3(3.7)	1(1.2)	5(6.2)	3(3.7)
Brain stem	17(21.0)	12(14.8)	12(14.8)	14(17.3)
Injury side, n (%)					0.698
Left	38(52. 1)	34(46.0)	37(49.3)	41(55.4)
Right	35(48.0)	40(54. 1)	38(50.7)	33(44.6)
Time since stroke onset, days, median (IQR)					0.164
Sample size(n)	51	55	54	57
time span(d)	14–176	14–125	14–173	14–147
median (IQR)	49.0 (38.0–89.5)	40.0 (29.0–68.5)	51.0 (33.0–81.8)	48.0 (32.0–66.0)
Time since stroke onset category, n (%)					0.019*
<90d n (%)	54 (74.0%)	69 (93.2%)	62 (82.7%)	62 (83.8%)
≥90d n (%)	19 (26.0%)	5 (6.8%)	13 (17.3%)	12 (16.2%)
Evaluation scale characteristics of screening, mean (SD)
MMSE	18.93(4.9)	18.72(5.0)	18.91(5.7)	18. 14(6.0)
IQCODE	3.054(0. 1)	3.041(0. 1)	3.048(0. 1)	3.072(0. 1)
Comprehensive assessment scale, mean (SD)
mRS	3.51(0.5)	3.62(0.5)	3.53(0.5)	3.64(0.5)	0.302
Barthel Index	38.27(16.7)	35.85(17.4)	40. 13(17.7)	33.91(17.6)	0.14
Cognitive/Motor Function assessment scale (SD)
MoCA	14.63(3.7)	14. 18(4.7)	13. 17(4.6)	14.49(4.3)	0.168
LOTCA	42.42(12.8)	41.2(12.3)	43.61(14.7)	40.05(12.4)	0.381
FMA	36.51(13.94)	34.41(14.85)	38.27(15.49)	33.77(15.04)	0.212
POMA	3.59(2.88)	4.26(3.25)	4.24(3.22)	3.74(3.3)	0.437
Mean laboratory test results (SD)
Glucose, mg/dL	6.9(0.6)	6.5(0.7)	6.2(0.9)	6.7(0.7)
Blood homocysteine,μmol/L	11.7(1.3)	11.3(1.4)	10.8(1.2)	11.6(1.3)
Glutamicpyruvic transaminase, U/L	9.45(0.7)	10. 17(0.6)	9.75(0.7)	9.93(0.6)
Creatinine, μmol/L	41.3(0. 1)	42.7(0.2)	41.5(0. 1)	42. 1(0. 1)
Urea nitrogen, mg/dl	4.76(2.7)	5.24(2.5)	5. 18(2.8)	4.97(2.2)
Creatine kinase, U/L	17.4(3.6)	18.2(3.7)	17.7(3.2)	17.9(3. 1)
Associated diseases n (%)
Hypertension	44(60.3)	39(52.7)	40(53.3)	42(56.8)	0.779
Diabetes	27(37.0)	30(40.5)	33(44.0)	26(35. 1)	0.695
Coronary heart disease	20(27.4)	22(29.7)	15(20.0)	19(25.7)	0.569

Note: Data are presented as mean (SD) or n (%), as appropriate. Between-group comparisons were performed using one-way ANOVA or Kruskal-Wallis tests for continuous variables, χ² tests or Fisher’s exact tests for categorical variables, and rank-based non-parametric tests for ordinal variables, as appropriate. P values indicate overall comparisons among the four treatment groups

MBI = the Modified Barthel Index; mRS=modified Rankin Scale; MMSE = Mini- Mental State Examination; IQCODE = Informant Questionnaire on Cognitive Decline in the Elderly; MoCA = Montreal Cognitive Assessment; LOTCA = Loewenstein Occupational Therapy Cognitive Assessment; FMA = Fugl-Meyer Assessment of Motor Function  Score; POMA = Tinetti Performance Oriented Mobility Assessment

MBI (range, 0 to 100), 100 points is considered normal; ≥ 60 indicates that life is basically self-care; 41 to 59 means moderate dysfunction, life needs help; 21 to 40 points indicates severe dysfunction, life dependence is obvious; ≤ 20 points means that life is completely dependent

mRS(range, 0 to 6), with lower scores representing better neurological function. MMSE (range, 0 to 30), with higher scores representing better global cognitive functioning. MoCA (range, 0 to 30), with higher scores representing better global cognitive functioning. LOTCA (range, 0 to 119), with higher scores representing better global cognitive functioning. FMA (range, 0 to 100), with higher score representing smaller movement disorder. POMA (range, 0 to 28), with higher the score representing better balance and walking ability. MMSE and MoCA scores were interpreted in the context of stroke status and educational background and were not used alone to diagnose dementiaundefined

Primary outcome: Changes in mRS score

The primary outcome was the longitudinal change in modified Rankin Scale (mRS) score from baseline to week 2, week 4, 1-month follow-up, and 3-month follow-up. Because lower mRS scores indicate better neurological function, a reduction in mRS score represents clinical improvement. The repeated-measures ANOVA revealed a significant main effect of time (F = 180.899, P < 0.001) and a significant group-by-time interaction (F = 7.389, P < 0.001), indicating that mRS scores changed over time and that recovery trajectories differed among the four treatment groups. The main effect of treatment group did not reach statistical significance (F = 2.533, P = 0.057).

Given that mRS is an ordinal outcome, time-point-specific non-parametric sensitivity analyses were further performed using mRS change scores from baseline. Kruskal-Wallis tests showed no significant between-group difference in mRS change at week 2 (H = 1.664, P = 0.645), whereas significant between-group differences were observed at week 4 (H = 20.750, P < 0.001), 1-month follow-up (H = 24.249, P < 0.001), and 3-month follow-up (H = 20.706, P < 0.001). Dunn’s rank-based post-hoc tests with Bonferroni correction showed that both NGF-containing groups had greater mRS improvement than the acupuncture + placebo group at week 4 and during follow-up. However, the SMES + NGF group did not show significantly greater mRS improvement than the acupuncture + NGF group at any post-baseline time point.

In the 2 × 2 factorial analysis of mRS change, the main effect of NGF was significant (P = 0.001), whereas the main effect of SMES and the SMES-by-NGF interaction were not significant (both P = 0.323). These findings suggest that improvement in global neurological disability, as measured by mRS, was mainly associated with NGF-containing treatment conditions rather than a statistically significant synergistic effect between SMES and NGF (Fig. 2). Detailed time-point-specific post-hoc comparisons are shown in Additional file 3: Table 1.

Fig. 2.

Change in mRS score and MoCA score. (A) In the box and violin plots, the horizontal line in the middle of the box represents the median, the upper side of the box represents the upper quartile, and the lower side of the box represents the lower quartile. If there is a black spot above and below the box, it means that the sample has outliers. (B) All of the means with 95% CIs were estimated from a repeated-measures, mixed linear model with time and group. The error bars indicate 95% CIs. mRS = modified Rankin Scale. (C1) Under the SMES condition, the simple effect of NGF was significant (P < 0.001) at week2. Under the condition of NGF, the simple effect of SMES treatment was significant (P = 0.030). (C2) Under the condition of SMES, the simple effect of NGF was significant (P < 0.001) at week4. Under the condition of NGF, the simple effect of SMES treatment was significant (P < 0.001). (C3) After 1 month of follow-up, the main effect of SMES was significant (P = 0.008), the main effect of NGF was significant (P < 0.001), and the interaction effect between SMES and NGF was significant (P < 0.001). Under the condition of NGF, the simple effect of SMES treatment was significant (P < 0.001). (C4) At the third month of follow-up, the simple effect of NGF was significant under SMES conditions (P < 0.001). Under the condition of NGF, the simple effect of SMES treatment was significant (P < 0.001). MoCA = Montreal Cognitive Assessment. (D) All of the means with 95% CIs were estimated from a repeated-measures, mixed linear model with time and group. The error bars indicate 95% CIs

Secondary Outcomes

Statistical details for secondary outcomes are provided in Table 2 and Additional file 3: Table 2, and the overall changes in secondary outcomes are shown in Fig. 3. For cognitive outcomes, repeated-measures ANOVA showed significant time effects and group-by-time interactions for both MoCA and LOTCA. For MoCA, there was a significant main effect of treatment group (F = 9.992, P < 0.001), a significant main effect of time (F = 285.112, P < 0.001), and a significant group-by-time interaction (F = 11.523, P < 0.001). For LOTCA, the main effect of time (F = 171.186, P < 0.001) and the group-by-time interaction (F = 14.144, P < 0.001) were significant, whereas the main effect of treatment group was not significant (F = 2.227, P = 0.085).

Table 2.

Clinical outcomes and changes from baseline by treatment group

Measure	Acupuncture + placebo (n=73)	Acupuncture + NGF (n=74)	SMES + placebo (n=75)	SMES + NGF (n=74)	Change from baseline, median (IQR)				Effect size	Overall P value for change
					Acupuncture + placebo	Acupuncture + NGF	SMES + placebo	SMES + NGF	partial η2
Primary Outcome: mRS score
Baseline	3.51(0.5)	3.62(0.49)	3.53(0.5)	3.64(0.48)	-	-	-	-	-	-
2 wk	3.21(0.67)	3.41(0.57)	3. 17(0.58)	3.36(0.56)	0.00(-1.00,0.00)	0.00(-1.00,0.00)	0.00(-1.00,0.00)	0.00(-1.00,0.00)	0.005	0.645
4 wk	3. 16(0.75)	2.81(0.77)	3.00(0.62)	2.81(0.77)	0.00(-1.00,0.00)	-1.00(-1.00,0.00)	0.00(-1.00,0.00)	-1.00(-1.00,0.00)	0.061	<0.001
8 wk	3.08(0.72)	2.61(0.82)	2.76(0.57)	2.59(0.76)	0.00(-1.00,0.00)	-1.00(-2.00,0.00)	-1.00(-1.00,0.00)	-1.00(-2.00,0.00)	0.073	<0.001
16 wk	2.89(0.79)	2.38(0.84)	2.64(0.67)	2.53(0.83)	-1.00(-1.00,0.00)	-1.00(-2.00, -1.00)	-1.00(-1.00,0.00)	-1.00(-2.00,0.00)	0.060	<0.001
Secondary outcomes
MBI
Baseline	38.27(16.74)	35.85(17.36)	40.13(17.71)	33.91(17.55)	-	-	-	-	-	-
2 wk	41.9(16.75)	43.04(16.59)	47.41(18.16)	51.09(19.63)	2.00 (0.00,7.00)	4.00 (2.00, 11.00)	4.00 (1.00, 14.00)	17.00 (6.00,23.00)	0.174	<0.001
4 wk	45.63(17.2)	51.05(17.67)	52(18.3)	61.22(19.27)	4.00 (0.50, 14.00)	12.05(5.00,24.25)	10.00 (2.00,21.00)	26(16.75,37.25)	0.219	<0.001
FMA
Baseline	36.51(13.94)	34.41(14.85)	38.27(15.49)	33.77(15.04)	-	-	-	-	-	-
2 wk	39.75(14.05)	38.86(13.8)	42.31(15.29)	41.86(17.46)	3.00 (1.00,5.00)	3.00 (2.00,6.00)	4.00 (2.00,6.00)	4.00 (2.00, 10.25)	0.012	<0.001
4 wk	43.16(14.97)	42.65(13.32)	46.61(14.95)	49. 16(18. 19)	6.00 (2.00,9.50)	7.00 (3.75, 12.00)	7.00 (4.00, 12.00)	12.00 (6.00,21.25)	0.098	<0.001
POMA
Baseline	3.59(2.88)	4.26(3.25)	4.24(3.22)	3.74(3.3)	-	-	-	-	-	-
2 wk	4.4(2.97)	5. 12(3. 19)	5.37(3. 17)	5.68(3.4)	1.00 (0.00, 1.00)	1.00 (0.00, 1.00)	1.00 (0.00,2.00)	2.00 (1.00,3.00)	0.096	<0.001
4 wk	5(3.2)	5.95(3.45)	6.35(3.26)	7.95(3.4)	1.00 (1.00,2.00)	1.00 (0.00,3.00)	2.00 (1.00,3.00)	4.00 (2.00,6.25)	0.154	<0.001
LOTCA
Baseline	42.42(12.78)	41.2(12.25)	43.61(14.68)	40.05(12.41)	-	-	-	-	-	-
2 wk	47.12(11.65)	47.03(13.37)	48.97(14.98)	50. 12(16. 17)	2.00(0.00,8.00)	3.00(0.00,8.25)	2.00(0.00, 13.00)	10.50(3.00, 18)	0.061	0.024
4 wk	48.71(12.05)	52.3(13.25)	52.88(14. 13)	62. 19(15.2)	3.00(0.00, 11.50)	8.00(2.75, 18.25)	7.00(2.00, 16.00)	22(15.75,30.00)	0.226	<0.001
MoCA
Baseline	14.63(3.74)	14. 18(4.67)	13. 17(4.64)	14.49(4.29)	-	-	-	-		-
2 wk	16.75(3.84)	16.68(4.38)	15.52(4.44)	18. 16(3.90)	2.00 (0.00,4.00)	2.00 (0.00,3.00)	2.00 (0.00,4.00)	3.00 (2.00,6.00)	0.045	0.016
4 wk	17.05(3.86)	17.69(4.09)	17.29(4.67)	20. 19(3.87)	2.00 (0.00,4.00)	2.00 (1,50.25.00)	3.00 (1.00,6.00)	5.00 (3.00,8.00)	0.096	<0.001
8 wk	17.63(3.78)	18.24(3.95)	17.03(4.61)	21.32(3.66)	3.00 (2.00,4.00)	3.00 (1.00,6.00)	4.00(1.00,6.00)	7.00(4.00,8.00)	0.169	<0.001
16 wk	17.86(3.67)	18.76(3.65)	17.25(4.75)	21.99(3.56)	3.00(2.00,4.00)	4.00(1.00,7.00)	4.00(2.00,6.00)	7.00(5.00, 10.00)	0.180	<0.001

Fig. 3.

Change in secondary outcomes. Note: Top. Effects of acupuncture + placebo and acupuncture + NGF on the continuous variables. Median differences and Q25, Q75 are shown. Bottom. Effects of SMES + placebo and SMES + NGF on the binary variables. Odds ratios and 95%CIs are shown. mRS = modified Rankin Scale; MBI = the simplified Modified Barthel Index; LOTCA = Loewenstein Occupational Therapy Cognitive Assessment; MoCA = Montreal Cognitive Assessment; FMA = Fugl-Meyer Assessment of Motor Function Score; POMA = Tinetti Performance Oriented Mobility Assessment

Time-point-specific post-hoc analyses showed that the SMES + NGF group demonstrated the most prominent cognitive improvement. For MoCA change scores, overall between-group differences were significant at week 2, week 4, 1-month follow-up, and 3-month follow-up. Bonferroni-corrected pairwise comparisons showed that the SMES + NGF group had greater MoCA improvement than the acupuncture + placebo group at week 2 and week 4, and greater improvement than all other groups at 1-month and 3-month follow-up. In the 2 × 2 factorial analysis at week 4, the main effects of SMES and NGF were significant, and the SMES-by-NGF interaction was also significant (P = 0.019), suggesting that the combined intervention had a more apparent effect on cognitive recovery than either component alone. For LOTCA, the SMES + NGF group showed greater improvement than the acupuncture + placebo group at week 2 and greater improvement than all other groups at week 4 after correction for multiple comparisons.

For functional and motor outcomes, repeated-measures ANOVA also showed significant time effects and group-by-time interactions for MBI, FMA, and POMA. For MBI, there was a significant time effect (F = 249.863, P < 0.001) and group-by-time interaction (F = 20.069, P < 0.001), while the main effect of treatment group did not reach statistical significance (F = 2.602, P = 0.052). Similar patterns were observed for FMA, with a significant time effect (F = 319.508, P < 0.001) and group-by-time interaction (F = 13.069, P < 0.001), but no significant main effect of treatment group (F = 0.997, P = 0.395). For POMA, the time effect (F = 333.599, P < 0.001) and group-by-time interaction (F = 31.285, P < 0.001) were significant, whereas the group effect was not significant (F = 2.367, P = 0.071).

Bonferroni-corrected post-hoc comparisons indicated that the SMES + NGF group showed greater improvement in MBI, FMA, and POMA than all other groups at week 4. Specifically, for MBI and FMA change scores, the SMES + NGF group showed significantly greater improvement than the acupuncture + placebo, acupuncture + NGF, and SMES + placebo groups at week 4. For POMA, the SMES + NGF group also showed greater improvement than all other groups at week 4, and the SMES + placebo group showed greater improvement than the acupuncture + placebo group. Detailed time-point-specific corrected post-hoc comparisons are presented in Additional file 3: Table 3.

Table 3.

Adverse Events, According to Group, Among Participants With Available Data

Adverse Event*	Participants Experiencing Event, n/N (%)
	Acupuncture + placebo	Acupuncture + NGF	SMES + placebo	SMES + NGF
0–4 wk
Discontinued participation because of adverse events	7/80(8.8)	9/83(10.8)	4/79(5. 1)	8/82(9.6)
Any adverse event (related or unrelated)	26/73(35.6)	36/74(48.6)	17/75(22.7)	29/74(39.2)
Nonserious adverse events(unrelated)	15/80(18.8)	8/83(9.6)	4/79(5.0)	6/82(7.3)
pulmonary infection	12/73(16.4)	6/74(8. 1)	3/75(4)	2/74(2.7)
Infection with coronavirus disease 2019	3/80(3.8)	2/83(2.4)	1/79(1.3)	4/82(4.9)
Nonserious adverse events(related)	11/80(13.8)	28/83(33.7)	13/79(16.5)	23/82(28.0)
Could not stand acupuncture treatment	2/80(2.5)	1/83(1.2)	1/79(1.3)	1/82(1.2)
It was considered that the treatment effect is poor	7/80(8.8)	3/83(3.6)	2/79(2.5)	1/82(1.2)
Small subcutaneous hematoma	2/73(2.7)	3/74(4. 1)	10/75(13.3)	8/74(10.8)
Hip muscle pain	0/73(0.0)	21/74(28.4)	0/75(0.0)	13/74(17.6)

fNIRS Examination

Task-based fNIRS was performed at baseline and week 4 to assess cortical haemodynamic responses during the Stroop task. HbO₂ concentration change was used as the primary haemodynamic index. The Stroop task included neutral, consistent, and inconsistent conditions, with the inconsistent condition representing the highest cognitive-control demand.

After 4 weeks of treatment, between-group differences in HbO₂ responses were observed in the left frontal cortex, particularly in channel 5. The SMES + NGF group showed lower HbO₂ values than the other treatment groups, suggesting reduced frontal cortical activation during task performance after treatment. No significant difference was observed between the neutral and consistent conditions. (Fig. 4)

Fig. 4.

The concentration of HbO₂ in the SMES + NGF group of left frontal cortex (channel 5). Note: (A) Enrollment and completion of the test. (B) Near infrared spectral imaging eight pairs of electrode channel distribution. (C) Three types of color-word matching tasks. (D) Correcting and dedrifting. (E) After post-test, it was found that HbO₂ in the SMES + NGF group was significantly smaller than that in the other groups (P < 0.05)

Condition-specific analyses further showed that treatment-related HbO₂ changes varied across Stroop conditions and cortical channels. (Fig. 5) Under the inconsistent condition, the SMES + NGF group showed reduced HbO₂ responses after treatment compared with pre-test values in the right frontal pole (channel 3) and primary motor cortex (channel 10). Under the consistent condition, reduced HbO₂ responses after treatment were observed in the right primary motor cortex and dorsolateral prefrontal cortex region (channel 20). These findings suggest that SMES combined with NGF was associated with condition-dependent changes in task-evoked cortical haemodynamic responses, particularly in frontal and motor-related cortical regions during the Stroop task. Because the fNIRS results did not survive FDR correction and were based on a small exploratory sample, they should be interpreted cautiously.

Fig. 5.

The changes of HbO₂ in different tasks of right frontal pole (channel 3) and primary motor cortex (channel 10) and dorsolateral prefrontal cortex (channel 20) before and after treatment in the four groups were observed. Note: (A) The comparison of HbO₂ concentration data of each channel during the Stroop task after treatment showed that the interaction effect of SMES + NGF group on channel 3 after intervention (P < 0.05). (B) The time main effect of No.10 channel was significant (P < 0.05). The interaction effect of channel 10 was significant (P < 0.05). (C) The interaction effect of channel 20 was significant (P < 0.05)

rs-fMRI examination

Exploratory rs-fMRI analyses were performed in a selected subgroup of participants from the acupuncture + NGF and SMES + NGF groups who had subcortical ischemic lesions without cortical involvement. In this imaging subgroup, 10 participants in the acupuncture + NGF group and 10 participants in the SMES + NGF group completed both baseline and post-treatment rs-fMRI assessments. Therefore, these findings should be interpreted as exploratory functional imaging results in a selected subcortical-stroke subgroup.

Exploratory rs-fMRI analyses assessed z-transformed amplitude of low-frequency fluctuation (zALFF), standardized regional homogeneity (zReHo), and seed-based functional connectivity (FC). For spontaneous neural activity, zALFF analysis showed treatment-associated changes after SMES + NGF treatment in several cortical regions, mainly involving frontal, parietal, occipital, and sensorimotor-related areas. These regions included the middle frontal gyrus, dorsolateral superior frontal gyrus, prefrontal cortex, precuneus, superior and middle occipital gyri, cuneus, and supplementary motor area (Fig. 6). The cluster-wise plots showed different pre- to post-treatment zALFF trajectories between the SMES + NGF and acupuncture + NGF groups, suggesting divergent changes in spontaneous neural activity after treatment.

Fig. 6.

Differences in ALFF main effect between groups and within groups. Note: (A, B) Main effect results of the zALFF analysis by group; (C-G) show significant differences in zALFF values in the classical frequency band (0.01–0.08 Hz) after treatment in the middle frontal gyrus, dorsolateral superior frontal gyrus, prefrontal lobe, parietal lobe, precuneus, superior occipital gyrus, middle occipital gyrus, cuneus, and supplementary motor area. (F) Significant differences in zALFF-Classical were observed in the parietal lobe, precuneus, cuneus, superior occipital gyrus, and middle occipital gyrus before and after SMES + NGF treatment. (H) Enrollment and completion of the test. The main effect of the zReHo analysis by group was voxel P < 0.005, cluster P < 0.05, FWE-corrected. ^*P < 0.05, ^**P < 0.005, ^***P < 0.001

For regional synchronization, standardized ReHo analysis showed treatment-associated changes in frontal and motor-related regions, including the dorsolateral superior frontal gyrus, superior frontal gyrus, premotor cortex, and supplementary motor area. These ReHo differences were observed both between the SMES + NGF and acupuncture + NGF groups after treatment and within the SMES + NGF group before and after treatment (Fig. 7), suggesting altered local synchronization of spontaneous neural activity after combined SMES + NGF treatment.

Fig. 7.

Differences of ReHo main effect between groups and within groups Note: (A, B) ReHo analysis group main effect results; (C) Classical frequency band (0.01–0.08 Hz), ReHo values in the dorsolateral superior frontal gyrus, frontal lobe, superior frontal gyrus, premotor cortex and supplementary motor area were significantly different between the two groups after treatment and before and after treatment in the SMES + NGF group. The main effect of szReHo analysis group was voxel P < 0.005, cluster P < 0.05, FWE corrected, ^*P < 0.05, ^**P < 0.005

Seed-based functional connectivity analyses further showed treatment-associated changes in connectivity between predefined seed regions and whole-brain voxels. The predefined seed regions included the precentral gyrus, hippocampus, and postcentral gyrus. In the SMES + NGF group, functional connectivity between the precentral gyrus and posterior cingulate gyrus, middle temporal gyrus, and parahippocampal gyrus decreased after treatment, whereas the acupuncture + NGF group showed an opposite direction of change. Similarly, hippocampus-based connectivity with the cerebellum, middle temporal gyrus, and precuneus decreased after SMES + NGF treatment but increased after acupuncture + NGF treatment. Postcentral gyrus-based connectivity with the precentral gyrus, prefrontal cortex, cuneus, and visual cortex also showed divergent pre- to post-treatment changes between the two groups (Fig. 8).

Fig. 8.

The relationship between seed regions and the whole brain Note: The main effect results of group were analyzed by FC in (A) precentral gyrus, (B) hippocampus and (C) postcentral gyrus, respectively. In (D, E) SMES + NGF group, the connection between precentral gyrus and posterior cingulate gyrus, middle temporal gyrus and parahippocampal gyrus decreased after treatment, while in acupuncture + NGF group, the connection between precentral gyrus and posterior cingulate gyrus, middle temporal gyrus and parahippocampal gyrus increased after treatment. (F) After treatment, the connection between hippocampus and cerebellum, middle temporal gyrus and precuneus decreased in SMES + NGF group, while the connection between hippocampus and cerebellum, middle temporal gyrus and precuneus increased in acupuncture + NGF group. (G, H) SMES + NGF group, the connection between the postcentral gyrus and the precentral gyrus, prefrontal lobe, cuneus, visual cortex decreased after treatment, while the connection between the postcentral gyrus and the precentral gyrus, prefrontal lobe, cuneus, visual cortex increased after treatment in the acupuncture + NGF group. The main effect of FC analysis group was voxel P < 0.005, cluster P < 0.05, FWE corrected, ^*P < 0.05, ^**P < 0.005, ^***P < 0.001

Overall, these exploratory rs-fMRI findings suggest that SMES combined with NGF was associated with changes in spontaneous neural activity, regional synchronization, and functional connectivity within cognitive, sensorimotor, and visual networks. However, because the rs-fMRI analyses were exploratory, involved a small selected subgroup with subcortical lesions only, and did not assess structural brain changes or direct BBB opening, these findings should be interpreted as hypothesis-generating functional imaging evidence.

As shown in Fig. 8, in the SMES + NGF group, FC decreased in several brain regions, including the limbic lobe, lingual gyrus, middle temporal gyrus, posterior cingulate gyrus, inferior temporal gyrus, precuneus, cuneus, parietal lobe, and posterior parietal cortex, particularly within the visual association cortex and hippocampal network. Additionally, FC decreased in the middle occipital gyrus, precentral gyrus, cuneus, and networks associated with visual processing, prefrontal cortex, premotor cortex, and postcentral gyrus. In contrast, FC increased in the acupuncture + NGF group after treatment.

Adverse Reaction Events

Six adverse events were observed (Table 3). Two moderate adverse events requiring hospitalisation were unrelated to the intervention. In the reported cases of mild adverse events, 33 were unrelated, and 75 were related to the intervention, mainly caused by a local small haematoma after acupuncture treatment and hip muscle pain following hip injection.

Discussion

This multicentre randomized trial evaluated the clinical effects and safety of SMES, NGF, and their combined use in patients with ischemic stroke during the convalescent phase. The primary mRS analysis showed significant improvement over time and a significant group-by-time interaction, indicating that the recovery trajectories of global disability differed among the four treatment groups. However, the 2 × 2 factorial analysis suggested that mRS improvement was mainly associated with the NGF component, whereas the SMES main effect and the SMES-by-NGF interaction were not significant. Therefore, the primary mRS findings should be interpreted as evidence of NGF-related improvement in global neurological disability rather than definitive evidence of a synergistic effect between SMES and NGF on mRS.

In contrast, the evidence for a combined SMES + NGF effect was more apparent for cognitive outcomes. The SMES + NGF group showed the most prominent improvement in MoCA and LOTCA, and factorial analyses indicated a significant SMES-by-NGF interaction for MoCA. These findings suggest that the combined intervention may be particularly relevant to post-stroke cognitive recovery. This domain-specific pattern may also explain why the combined effect was clearer for cognitive measures than for mRS. The mRS is a coarse ordinal scale of global disability and may be less sensitive to changes in specific cognitive domains, motor control, balance, or daily functional performance [26]. In addition, baseline mRS scores were restricted to 3 or 4 by the eligibility criteria, limiting the measurable range of improvement and reducing the ability of mRS to detect subtle functional changes.

The apparent difference in mRS trajectories between the acupuncture + NGF and SMES + NGF groups should therefore be interpreted cautiously. Although both NGF-containing groups showed greater mRS improvement than the acupuncture + placebo group, SMES + NGF did not show significantly greater mRS improvement than acupuncture + NGF in the non-parametric post-hoc sensitivity analyses. Several factors may have contributed to this pattern, including the limited sensitivity of mRS, the restricted baseline mRS range, sample size and variability in subgroup comparisons, and potential short-term fatigue or adaptation related to electrical stimulation [27, 28]. Accordingly, the mRS results do not support a definitive synergistic effect of SMES and NGF on global disability, although they do not exclude potential benefits in more specific functional domains.

The most consistent evidence supporting a combined SMES + NGF effect was observed in cognitive outcomes. Compared with global disability measured by mRS, cognitive measures appeared more sensitive to the combined intervention. The SMES + NGF group showed the largest improvements in MoCA and LOTCA, and the factorial analyses indicated a significant SMES-by-NGF interaction for MoCA. These findings suggest that the combined intervention may be particularly relevant to post-stroke cognitive recovery. The persistence of MoCA improvement during follow-up further suggests that the cognitive benefits may extend beyond the 4-week treatment period, although longer follow-up is needed to determine the durability of this effect.

Secondary outcomes also suggested broader functional improvement. Improvements in MBI, FMA, and POMA were generally greater in the SMES + NGF group, particularly at week 4, indicating potential benefits in activities of daily living, motor function, balance, and gait. However, these findings should be interpreted as supportive rather than definitive, because multiple secondary outcomes were assessed and the trial was not primarily powered to establish treatment effects for each functional domain separately. Taken together, the clinical results suggest that the combined intervention may have broader rehabilitative value, with the strongest evidence observed for cognitive recovery.

The exploratory fNIRS findings provide preliminary neurophysiological support for treatment-associated changes in cortical haemodynamic responses. In the SMES + NGF group, reduced task-evoked HbO2 responses were observed in selected frontal and motor-related channels, including the right frontal pole, primary motor cortex, and right primary motor cortex/dorsolateral prefrontal cortical regions under specific Stroop conditions. One possible interpretation is that the combined intervention may alter cortical haemodynamic recruitment during cognitive-control processing, potentially reflecting reduced compensatory activation or improved neural efficiency after treatment [29, 30]. However, this interpretation remains speculative. Because the fNIRS analyses were exploratory, involved a small sample, and the channel-wise findings did not survive correction for multiple comparisons, these results should be regarded as hypothesis-generating rather than confirmatory mechanistic evidence.

Similarly, the exploratory rs-fMRI findings suggested that clinical improvement may be accompanied by changes in spontaneous neural activity, regional synchronization, and functional connectivity within cognitive and sensorimotor networks. These functional imaging findings are consistent with the possibility that post-treatment recovery may involve reorganization of task-evoked haemodynamic responses and resting-state functional networks [31]. Nevertheless, the present trial did not directly demonstrate BBB opening, NGF delivery into the human brain, or structural brain repair. Therefore, the neurophysiological findings should be interpreted cautiously and require confirmation in larger imaging cohorts with longer follow-up.

In the present study, SMES + NGF was associated with treatment-related changes in zALFF, standardized ReHo, and seed-based FC in regions involved in cognitive, sensorimotor, and visual networks. zALFF changes were mainly observed in frontal, parietal, occipital, and sensorimotor-related areas, including the prefrontal cortex, dorsolateral superior frontal gyrus, precuneus, cuneus, occipital cortex, and supplementary motor area. ReHo analyses showed altered regional synchronization in frontal and motor-related regions, including the dorsolateral superior frontal gyrus, premotor cortex, and supplementary motor area. Seed-based FC analyses further showed divergent pre- to post-treatment connectivity patterns between the SMES + NGF and acupuncture + NGF groups, involving connectivity between the precentral gyrus, hippocampus, postcentral gyrus, posterior cingulate cortex, middle temporal gyrus, parahippocampal gyrus, cerebellum, precuneus, prefrontal cortex, cuneus, and visual cortex.

These findings suggest that SMES combined with NGF may be associated with functional reorganization of cognitive and sensorimotor networks after stroke. However, this interpretation should remain cautious. Decreased or increased ALFF, ReHo, or FC values cannot be directly interpreted as better or worse neural function without additional mechanistic validation. In particular, reduced connectivity after treatment may reflect reduced compensatory overactivation, improved network efficiency, or altered task-independent functional organization, but these possibilities cannot be distinguished in the present study. Therefore, the rs-fMRI findings should be regarded as exploratory functional imaging evidence rather than definitive proof of neural repair or treatment mechanism.

The biological rationale for combining SMES with NGF was derived from our previous animal studies, in which SMES applied at GV20 and GV26 was shown to transiently increase BBB permeability and facilitate NGF entry into brain tissue under specific stimulation parameters, including 2/100 Hz frequency, 3 mA current, alternating 6s on/off cycles, and 40-min stimulation duration. These preclinical findings provided the basis for translating this intervention to the present clinical trial. Nevertheless, the present study did not directly measure BBB permeability, NGF concentration in the central nervous system, or tracer-based NGF delivery in human participants. Thus, although the clinical and exploratory imaging findings are consistent with the preclinical rationale, they do not directly demonstrate SMES-mediated BBB opening or NGF delivery into the human brain.

Regarding safety, the most common adverse events were local hematoma related to acupuncture and mild gluteal muscle pain after intramuscular injection. These events occurred mainly during the treatment period, were generally mild, and resolved without complications. No serious treatment-related adverse events were observed. These findings suggest that SMES combined with NGF was generally well tolerated under the present study protocol. However, larger trials with longer follow-up are needed to further evaluate rare adverse events and the long-term safety of repeated treatment courses.

This study has several limitations. The treatment duration was relatively short, with only one course of NGF administration, making it unclear whether repeated treatments would yield better or more sustained results. Additionally, the follow-up period was limited to 16 weeks, and longer-term studies are required to fully assess the sustained impact of SMES + NGF treatment. On the other hand, previous animal studies have demonstrated that SMES can transiently open the BBB and facilitate the entry of NGF into the brain to exert therapeutic effects. In the present trial, rs-fMRI and fNIRS were used as exploratory tools to investigate treatment-associated functional brain changes. However, the present clinical trial did not directly assess BBB opening or NGF delivery into the human brain using BBB-specific imaging, tracer methods, or direct biological measurements. Therefore, whether SMES facilitates NGF delivery across the BBB in humans remains to be directly confirmed.In addition, the rs-fMRI analysis was performed only in participants with subcortical ischemic lesions and no cortical involvement. This selection was made to reduce lesion-related distortion of cortical BOLD signals and to improve the interpretability of ALFF, ReHo, and functional connectivity analyses. However, because patients with cortical lesions were not included in the rs-fMRI subgroup, these functional imaging findings cannot be fully generalized to patients with cortical infarcts. Future studies should include larger imaging cohorts with both cortical and subcortical lesions and perform lesion-stratified analyses. Another limitation is that participants were not stratified by time since stroke onset, lesion location, or cerebral blood supply territory during randomization or outcome analysis. Because stroke stage, lesion site, and vascular territory may influence spontaneous recovery, neuroplastic potential, and treatment response, future trials should consider stratified randomization and prespecified subgroup analyses based on these factors. Although BDNF, NT-3, and NT-4 are also relevant to post-stroke neural repair, the present trial focused on NGF because it followed our previous NGF-based preclinical framework and approved clinical protocol. Future studies may examine whether SMES can be combined with other neurotrophic factors or macromolecular therapeutic agents. Current limitations include equipment constraints, ethical considerations, and patient consent issues. Nevertheless, we continue to pursue deeper clinical validation of this mechanism. A further limitation is that NIHSS was not collected as a prespecified baseline or follow-up measure. Although baseline mRS, MBI, FMA, POMA, MoCA, and LOTCA were used to characterize disability, motor function, mobility, and cognition, the absence of NIHSS limited our ability to quantify acute neurological deficit severity and to adjust for its potential influence on recovery. Future trials should include NIHSS, particularly for participants enrolled in the early post-stroke phase. An additional point that warrants emphasis is the relatively low baseline MMSE and MoCA scores across all groups, indicating that the enrolled population had substantial post-stroke cognitive impairment at study entry. This was consistent with our eligibility strategy, which aimed to include patients with cognitive dysfunction during the convalescent stage of ischemic stroke. However, these findings may limit the generalizability of the results to stroke survivors with milder or no cognitive impairment. Because education may influence performance on cognitive screening scales, future studies should further stratify or adjust for educational level. Additionally, we retrospectively extracted the reperfusion therapy status from the clinical record, but the trial was not stratified by recanalization, nor was subgroup analysis performed based on thrombotic or endovascular thrombectomy status. Since reperfusion therapy may affect post-stroke recovery, future trials should proactively record and adjust recanalization history.

Conclusions

In conclusion, NGF-containing interventions were associated with improvement in global neurological disability in patients with ischemic stroke during the convalescent phase, while SMES combined with NGF showed additional benefits for cognitive recovery. The intervention was generally well tolerated, with no serious treatment-related adverse events observed. These findings suggest that SMES combined with NGF may serve as a promising adjunctive therapeutic strategy for post-stroke cognitive and functional recovery.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ALFF

Amplitude of Low-Frequency Fluctuations

BBB

Blood-Brain Barrier

CT

Computed Tomography

FMA

Fugl-Meyer Assessment of Motor Function Score

FC

Functional Connectivity

fNIRS

functional Near-Infrared Spectroscopy

LOTCA

Loewenstein Occupational Therapy Cognitive Assessment

MRI

Magnetic Resonance Imaging

MMSE

Mini-Mental State Examination

MBI

Modified Barthel Index

mRS

modified Rankin Scale

MoCA

Montreal Cognitive Assessment

NGF

Nerve Growth Factor

POMA

Performance-Oriented Mobility Assessment

rs-fMRI

resting-state functional Magnetic Resonance Imaging

ReHo

Regional homogeneity

SMES

Specific mode acupoint electrical stimulation

Author contributions

Xianming Lin, Congcong Ma and Mengyuan Dai had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: Xianming Lin, Congcong Ma, Mengyuan Dai, Yibin Zhao, Zhaoxing Jia, Ziqi Xuan, Junhong Liu. Acquisition, analysis, or interpretation of data: Mengyuan Dai, Zhaoxing Jia, Ziqi Xuan, Junhong Liu, Shuijin Zhang and Yiqing Zhang. Drafting of the manuscript: Mengyuan Dai, Zhaoxing Jia, Ziqi Xuan, Junhong Liu, Xuewen Wu, Yiqing Zhang, Yibin Zhao, Shuijin Zhang. Critical revision of the manuscript for important intellectual content: Xianming Lin, Congcong Ma and Mengyuan Dai. Statistical analysis: Xuewen Wu, Mengyuan Dai, Zhaoxing Jia. Obtained funding: Xianming Lin. Administrative, technical, or material support: Xianming Lin, Mengyuan Dai, Congcong Ma, Zhaoxing Jia, Xuewen Wu, Yiqing Zhang, Yibin Zhao, Shuijin Zhang, Jianxun Liu, Lixiu Wu, Huimei Zhu, Mengming Shao, Caijun Ma, Shiting Xu, Nuo Xu, Weiwei Han. Supervision: Xianming Lin, Mengyuan Dai, Congcong Ma, Zhaoxing Jia, Xuewen Wu, Yibin Zhao, Yiqing Zhang.

Funding

The trial is funded by the Special Project of Modernization of Traditional Chinese Medicine in Zhejiang Province in 2022 (No.2022ZX009), the National Natural Science Foundation of China (No. 82174502) and Zhejiang Chinese Medical University Scientific Research Project-key research project (No. 742219A00203).

Data availability

Deidentified participant data and data directory will be made available upon request. The authors will share all data that support published results of the trial. Data will be made available as required for specific, approved analyses and will be provided from a locked, cleaned, and deidentified database.

Declarations

Ethics approval and consent to participate

Ethical approval was obtained from the ethics committees of the participating institutions (approval numbers: ZSLL-KY-2022-002-01, EJ-NBEY-KY-2022-027-01, and 20220210-1).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.