Adjuvant effects of vagus nerve stimulation on post-stroke rehabilitation: a systematic review and meta-analysis

Abstract

Objective

Previous meta-analysis only focused on the safety and effectiveness of vagus nerve stimulation (VNS) in upper extremities motor function in stroke patients. The aim of this study was to systematically evaluate the efficacy and safety of VNS for more comprehensive functional rehabilitation in stroke patients.

Design

This systematic review and meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines. The study protocol was registered in PROSPERO (CRD42024552624) and included randomised controlled trials (RCTs) investigating VNS-assisted rehabilitation in patients with stroke.

Data sources

Searches were conducted in PubMed, Web of Science, Cochrane Library, Embase, China National Knowledge Infrastructure (CNKI), and Wanfang Database, covering the period up to October 2025.

Eligibility criteria

We included studies in which the intervention group received VNS, while the control group received either sham stimulation or conventional rehabilitation alone. Studies were required to report efficacy outcomes (such as upper extremity motor function and swallowing function) as well as safety outcomes.

Data extraction and synthesis

Two researchers independently performed literature screening, data extraction and quality assessment. A meta-analysis was conducted using RevMan V.5.4 (The Cochrane Collaboration, London, United Kingdom) and STATA V.18.0(StataCorp LLC, College Station, TX, USA), using the standardised mean difference (SMD) for continuous outcomes and applying a random-effects model. To explore potential sources of heterogeneity, sensitivity and subgroup analyses were performed, while publication bias was assessed using funnel plots and Egger's test.

Results

Overall, 18 RCTs involving 954 participants were included in this study. This meta-analysis indicated that VNS could improve Fugl-Meyer Assessment of Upper Extremity (SMD=0.89, 95% CI 0.59 to 1.20, I²=64%, p<0.00001), Wolf Motor Function Test (WMFT; SMD=1.07, 95% CI 0.32 to 1.83, I²=81%, p=0.005), Motor Activity Log (SMD=0.44, 95% CI 0.18 to 0.70, I²=0%, p=0.0008), swallowing function (SMD=0.62, 95% CI 0.12 to 1.11, I²=0%, p=0.01), extensor carpi radialis muscle strength (SMD=1.07, 95% CI 0.67 to 1.47, I²=0%, p<0.00001), Functional Independence Measure (SMD=1.42, 95% CI 0.61 to 2.23, I²=48%, p≤0.0006), Modified Barthel Index (SMD=0.95, 95% CI 0.48 to 1.42, I²=75%, p<0.0001) and shortened Motor Evoked Potential latency (SMD=−0.76, 95% CI −1.20 to –0.31, I²=0%, p=0.00009) in stroke patients compared with the control group. Quality assessment using the Cochrane Risk of Bias tool indicated that most studies had a low risk of bias. Publication bias was low, as indicated by symmetric funnel plots and Egger's test results (p>0.05 for all key outcomes). According to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, five outcomes (eg, swallowing function) were rated as high quality, while two (WMFT and spasticity) were rated as low quality. Sensitivity analysis confirmed that no single study significantly influenced the pooled results. Subgroup analysis identified stimulation modality, treatment duration, region and age as the main sources of heterogeneity in this study.

Conclusion

The evidence supporting the use of VNS to improve function in stroke patients demonstrates reasonable reliability, a satisfactory degree of consistency and applicability, and suggests a potentially favourable clinical impact.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• A dual-independent operation was used for literature screening, data extraction and quality assessment, with disputes resolved through discussion or third-researcher adjudication to minimise subjective bias.

• Study quality was evaluated using the Cochrane Risk of Bias tool, and the certainty of evidence was assessed via the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system.

• Statistical approaches were used to quantify interstudy heterogeneity, and potential sources were explored through subgroup analyses (eg, stimulation modality and treatment duration).

• The included studies exhibited geographical skewing (8 out of 18 were from Asia), which may limit the methodological representativeness across different regions.

• The small sample sizes in individual randomised controlled trials constrained the statistical power for between-group comparisons.

Introduction

Strokes, or cerebrovascular accidents, occur when blood flow in the brain is disrupted, either via blockage or intracranial bleeding, leading to compromised neurological functions. According to the Global Burden of Disease study, stroke is the leading cause of death among adults in China and ranks as the second-most common cause of mortality worldwide.^{1 2} Stroke survivors frequently face a diverse range of challenges, including motor deficits, cognitive impairments, speech and swallowing difficulties, and psychological issues, all of which greatly impact their ability to carry out activities of daily living (ADL).^{3 4}

Vagus nerve stimulation (VNS) has gained attention as a neuromodulatory technique with roots traced back to the late 19th century when James Corning first explored its potential. Over the decades, both invasive and non-invasive forms of VNS have proven effective in treating conditions like depression, epilepsy and rehabilitation of motor functions post stroke.⁵ Various stimulations are used in VNS, including: (1) electrical stimulation through implanted device, transcutaneous cervical VNS (tcVNS), transcutaneous auricular VNS (taVNS) and transcranial direct current stimulation over the vagus nerve (VN-tDCS); (2) repetitive transcranial magnetic stimulation (rTMS); and (3) traditional Chinese therapies like acupuncture and auricular point seed-pressing. Recent literature increasingly supports the role of VNS in enhancing motor skills, hand function, cognitive abilities, swallowing capacities, mood and ADLs in stroke rehabilitation. In a pivotal study by Dawson et al, 106 participants took part in a randomised, sham-controlled, triple-blind trial in which all underwent a 6-week outpatient programme with a VNS device, followed by exercises at home. The findings revealed significant clinical advantages of VNS, particularly in enhancing motor abilities in patients with moderate to severe arm deficits after ischaemic stroke.⁶ In a long-term study by Kimberley et al, patients receiving paired VNS were found to maintain improvements in upper extremity impairment, activity ability, participation and quality of life (QOL) at 1 year. Paired VNS, which is approved by the US Food and Drug Administration as a treatment option, has been shown to confer sustained benefits in individuals with chronic upper extremity dysfunction following ischaemic stroke.⁷

Recently, a team of Chinese researchers carried out a double-blind, sham-controlled, parallel pilot study involving 40 acute stroke patients experiencing dysphagia. These patients were randomised to either a sham taVNS treatment or taVNS in addition to traditional rehabilitation training. When compared with the sham group, the group receiving taVNS showed significantly better swallowing performance after 3 weeks, and these improvements lasted for at least four weeks after treatment. This suggests that integrating taVNS with standard rehabilitation may offer enhanced benefits for improving swallowing abilities in stroke patients with dysphagia.⁸

While a meta-analysis examining the safety and effectiveness of VNS in stroke patients has been published, its main focus was on motor function in the upper extremities.⁹ Another meta-analysis, while broadening its scope to incorporate outcomes such as QOL, level of independence and adverse events, was limited by the inclusion of only seven studies.¹⁰

To date, a comprehensive review evaluating the effect of VNS on a wider spectrum of functional outcomes—including hand function, swallowing ability and muscle strength—is still lacking. We undertake a systematic review and meta-analysis in this study to try and close this gap. By updating and integrating the latest literature with previously available data, we provide the most up-to-date and thorough information on the safety and effectiveness of VNS in post-stroke rehabilitation.

Materials and methods

Literature search

This research adhered to the guidelines set by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (2020),¹¹ and the study protocol was registered with the International Prospective Registry of Systematic Reviews (PROSPERO: CRD42024552624).

Investigators JZ and MS formulated the search strategy, independently creating search terms and keywords for a thorough exploration of several databases: PubMed, Web of Science, Embase, Wanfang, China National Knowledge Infrastructure and Cochrane Library. The search targeted randomised controlled trials (RCTs) and spanned from the inception of each database to October 2025. Search terms included ‘vagus nerve stimulation’, ‘stroke & cerebrovascular accident & cerebrovascular stroke’, as well as ‘random*’. The online supplemental table S1 presents a comprehensive strategy for searching the literature.

Study selection

The process for selecting RCTs included a rigorous evaluation of potential studies through titles, abstracts and full texts based on predefined inclusion and exclusion criteria.

Inclusion criteria:

1. Participants: adults aged 18 years or older diagnosed with stroke based on clinical criteria, regardless of gender or race.

2. Intervention and control: the intervention group received VNS, whereas the control group was given either sham stimulation or conventional rehabilitation treatment.

3. Outcomes: the outcomes of interest encompassed both the efficacy and safety outcomes of VNS. Efficacy outcomes included upper limb motor function recovery, hand function, swallowing function, Motor Evoked Potential (MEP) latency, muscle strength, spasticity, QOL and ADL. Safety outcomes were evaluated based on the incidence of adverse events. Upper limb motor function recovery was primarily assessed using the Fugl-Meyer Assessment of Upper Extremity (FMA-UE), Wolf Motor Function Test (WMFT) and Motor Activity Log (MAL). Hand function was measured via the Box and Block test (BBT). Swallowing function was evaluated with the Australian Therapy Outcome Measures for Swallowing and the Mann Assessment of Swallowing Ability. Wrist extensor muscle strength was quantified using surface electromyography, with a specific focus on the root mean square (RMS) value. QOL was assessed by the Stroke Impact Scale (SIS), while ADL were measured using the Functional Independence Measure (FIM) and the Modified Barthel Index (MBI).

Exclusion criteria:

1. Studies that were not RCTs, such as retrospective studies, animal research or reviews.

2. Reports with incomplete or incorrect outcome data, where contact with original authors failed to retrieve sufficient information.

3. Duplicate publications.

4. Researchers JZ and MS conducted independent screenings of titles and abstracts, evaluated full texts and determined study eligibility. Discrepancies were resolved through consensus discussions.

Data extraction

Researchers JZand MS independently performed data extraction based on the established search strategy and inclusion criteria. The information collected included:

1. Publication details: title, lead author and year of publication.

2. Study characteristics: type of study, duration of treatment.

3. Participant demographics: sample size, age, gender and stroke duration.

4. Intervention specifics: stimulation mode, duration, frequency, intensity and basic rehabilitation treatments.

5. Outcomes: for continuous variables, the means and standardised mean difference (SMD) were recorded, while for categorical variables, the number of events and total occurrences were noted.

Any disagreements during data extraction were resolved by a third researcher, XG, who made the final decision.

Quality assessment

The quality of the RCTs included in this review was independently evaluated by researchers JZ and XS using the Cochrane Risk of Bias tool.¹² This assessment covered seven domains: random sequence generation, allocation concealment, participants and personnel blinding, reporting bias, completeness of outcome data and other potential biases. Studies were categorised as having a ‘high’, ‘low’ or ‘unclear’ risk of bias. Any differences in assessment were addressed through discussion, with researcher GXM providing the final decision if necessary.

Statistical analysis

Literature management was performed using EndNote X9(Clarivate Analytics, Philadelphia, PA, USA), data organization was conducted with Excel(Microsoft Corporation, Redmond, WA, USA), and statistical analyses were carried out with RevMan V.5.4(The Cochrane Collaboration, London, United Kingdom) and Stata V.18.0(StataCorp LLC, College Station, TX, USA). For binary outcomes, OR were calculated, while SMDs were used for continuous outcomes. Heterogeneity was assessed with 95% CIs and I² tests, with p<0.05 or I²≥50% indicating significant heterogeneity. All analyses used random-effects models with a significance level of α=0.05. Sensitivity analyses were conducted using the leave-one-out method. Publication bias was evaluated through funnel plots and Egger’s test. Subgroup analysis of the WMFT, FMA-UE and MBI was conducted based on stimulation modality, transcutaneous stimulation modalities, treatment duration, study region and mean age to explore result stability and potential sources of heterogeneity. Additionally, the Grading of Recommendations Assessment, Development and Evaluation(GRADE) framework was used to assess the certainty of the evidence for each outcome, which was then classified as high, moderate, low or very low.¹³

Patient and public involvement

No patients or members of the public were involved in the design, conduct, or reporting of this research. However, they will be engaged in the dissemination of the study findings.

Results

Selection of eligible studies

The database search initially identified 467 studies that appeared relevant. After eliminating 213 duplicates, 254 records were screened for titles and abstracts, resulting in the exclusion of 193 studies. Of the 61 studies subjected to full-text reviews, 18 were ultimately included in the analysis.68 14,29 The literature search and selection process are presented in figure 1.

Figure 1. Flow chart of literature screening. RCT, randomised controlled trial.

Study characteristics

The meta-analysis encompassed 954 stroke patients from 18 studies. The mean age of participants ranged from 56.60±13.23 to 72.9±12.2 years, with a mean stroke duration spanning from 17.3±7.1 days to approximately 3.3±2.6 years. Of these participants, 539 were male, and the cohort comprised cases of both ischaemic and haemorrhagic stroke. Treatment durations varied between 2 weeks and 3 months.

Regarding participant recruitment phases, one study enrolled participants in the acute phase,⁸ eight in acute and subacute phases,1416 17 20,22 24 29 three in the subacute phase,^{18 19 23} one in the subacute and chronic phases,¹⁵ and the remaining studies in the chronic phase.625,28 Intervention methodologies included taVNS in eleven studies, VNS device implantation in three studies, and each of tcVNS, rTMS, VN-tDCS and auricular point seed-pressing stimulation in the remaining four studies (online supplemental table S2).

Methodological quality and risks of bias

The quality of the selected studies was evaluated using the Cochrane Risk Assessment Tool. All studies appropriately used random number tables for allocation, designating them as low risk. Fifteen studies used allocation concealment methods such as the envelope method or centre randomisation, categorised as low risk, while the remaining studies lacked specification of these methods and were rated as having an unclear risk. Five studies were considered high risk as investigators, participants and therapists were aware of the treatment assignments. The remaining studies demonstrated low risk in this domain. Eleven studies reported blinded outcome assessment (low risk), while the others did not mention blinding procedures (unknown risk). All studies presented complete outcome data and showed no evidence of bias (low risk). Eleven studies were pre-registered and were rated as low risk for selective reporting. Conversely, studies without registration numbers could not be assessed for selective reporting and were classified as having an unknown risk. Figure 2 provides a visual representation of the risk of bias assessment.

Figure 2. Risk of bias graph and summary.

Meta-analysis results

The studies assessed various outcomes, including upper limb motor function, hand function, swallowing function, spasticity, muscle strength, MEP latency, QOL, ADL and adverse events.

Change in Fugl-Meyer Assessment for Upper Extremity

Data for the FMA-UE were derived from 12 studies. The findings showed that the upper extremity motor performance in the VNS group was significantly better than that of the control group (SMD=0.89, 95% CI 0.59 to 1.20, I²=64%, p<0.00001; figure 3A). To explore potential heterogeneity, subgroup analyses were conducted based on stimulation modality, transcutaneous stimulation types, treatment duration, geographic region and average age (online supplemental table S3). Notably, studies conducted in America did not significantly outperform the control group, (SMD=0.35, 95% CI −0.03 to 0.73, I²=16%, p=0.07).

Figure 3. (A) FMA-UE forest plot; (B) WMFT forest plot; and (C) MAL forest plot. FMA-UE, Fugl-Meyer Assessment of Upper Extremity; MAL, Motor Activity Log; VNS, vagus nerve stimulation; WMFT, Wolf Motor Function Test.

Change in Wolf Motor Function Test

Five studies were included in the WMFT analysis, showing that VNS significantly outperformed the control in enhancing motor function (SMD=1.07, 95% CI 0.32 to 1.83, I²=81%, p=0.005; figure 3B). Subgroup analysis by stimulation modality revealed that invasive VNS resulted in notably superior motor function improvement compared with non-invasive VNS (SMD=0.76, 95% CI 0.40 to 1.12, I²=0%, p<0.00001). In contrast, no significant improvement was observed in the non-invasive VNS group (SMD=1.42, 95% CI −0.03 to 2.87, I²=88%, p=0.05). When analysing treatment duration, patients undergoing treatment for more than 6 weeks showed greater motor function gains (SMD=0.76, 95% CI 0.40 to 1.12, I²=0%, p<0.00001). Regional subgroup analysis indicated no significant improvement in American studies compared with control (SMD=0.54, 95% CI −0.13 to 1.22, I²=64%, p=0.11), whereas significant enhancements were documented in studies from Asia and Europe (p<0.00001). Moreover, participants with a mean age below 60 years did not significantly differ from the control group in terms of motor function improvement (SMD=0.21, 95% CI −0.37 to 0.79, I²=0%, p=0.48; online supplemental table S3).

Change in Motor Activity Log

The assessment of real-world arm use in terms of quantity and quality was conducted using the MAL. Three studies contributed data to this analysis, and the findings indicated that arm use in stroke patients treated with VNS was much better than in the control group (SMD=0.44, 95% CI 0.18 to 0.70, I²=0%, p=0.0008; figure 3C).

Change in Box and Block test

The BBT was used to evaluate hand function and manual dexterity. The present analysis was based on two investigations, the findings of which showed no statistically significant variation in the improvement of hand function between control and experimental groups (SMD=0.41, 95% CI −0.01 to 0.84, I²=0%, p=0.06; figure 4A).

Figure 4. (A) Box and Block test forest plot; (B) swallowing function test forest plot; (C) MEP latency forest plot; (D) root mean square of the extensor carpi radialis brevis forest plot; (E) spasticity forest plot; and (F) SIS forest plot. MEP, Motor Evoked Potential; SIS, Stroke Impact Scale; VNS, vagus nerve stimulation.

Change in swallowing function

Regarding swallowing function, the pooled analysis of three studies indicated that the VNS group demonstrated significantly greater improvement than the control group (SMD=0.68, 95% CI 0.33 to 1.04, I²=0%, p=0.0002; figure 4B).

Change in Motor Evoked Potential latency

Analysis of two studies on MEP latency demonstrated that VNS was superior to the control group in reducing MEP latency (SMD=−0.76, 95% CI −1.20 to –0.31, I²=0%, p=0.00009; figure 4C).

Change in muscle strength

Muscle strength of the wrist extensor was measured using surface electromyography, specifically focusing on the RMS, indicative of muscle strength. Two studies showed that VNS significantly enhanced the strength of the extensor carpi radialis brevis compared with controls (SMD=1.07, 95% CI 0.67 to 1.47, I²=0%, p<0.00001; figure 4D).

Change in spasticity

An analysis encompassing two studies found no statistically significant variation in spasticity reduction between participants undergoing VNS and those in the (SMD=−1.50, 95% CI −3.25 to 0.26, I²=92%, p=0.1; figure 4E).

Change in quality of life

Two studies assessed QOL using the SIS. The improvement in QOL did not show a statistically significant difference between the VNS and control groups (SMD=0.04, 95% CI −0.31 to 0.39, I²=0%, p=0.83; figure 4F).

Change in activities of daily living

ADL was assessed using the FIM and MBI. Two studies analysing FIM scores showed that VNS significantly outperformed the control group (SMD=1.42, 95% CI 0.61 to 2.23, I²=48%, p<=0.0006; figure 5A). Five studies analysing MBI scores also demonstrated the superiority of VNS as an adjunctive treatment (SMD=0.95, 95% CI 0.48 to 1.42, I²=75%, p<0.0001; figure 5B). Subgroup analysis based on geographical region revealed no statistically significant improvement in American studies compared with the control group (SMD=0.37, 95% CI −0.08 to 0.82, I²=NA, p=0.11). To assess the impact of study quality, we performed a subgroup analysis after excluding articles with high or unclear risk of bias. Notably, the pooled estimate for the MBI changed significantly after this exclusion (SMD=0.95, 95% CI 0.48 to 1.42, I²=75%, p<0.0001 vs SMD=0.56, 95% CI 0.09 to 1.02, I²=32%, p=0.02; online supplemental table S3). These findings suggest that the quality of the articles may affect the stability of the results, and further research is needed to confirm this.

Figure 5. (A) FIM forest plot; (B) MBI forest plot; (C) adverse effects forest plot. FIM, Functional Independence Measure; MBI, Modified Barthel Index; VNS, vagus nerve stimulation.

Change in adverse effects

Across five studies assessing the frequency of adverse events attributable to the treatment, no substantial difference was observed between the VNS and control groups regarding the occurrence of adverse effects (OR=1.12, 95% CI 0.79 to 1.60, I²=0%, p=0.53; figure 5C).

Sensitivity analysis and publication bias

The sensitivity analysis, performed by case-by-case exclusion across FMA-UE, WMFT, MAL, swallowing function, MBI and adverse events, revealed no significant sensitivity sources (figure 6). To assess publication bias, funnel plots and Egger’s test were used. The funnel plots demonstrated symmetry, suggesting a lack of considerable publication bias in the meta-analysis concerning the FMA-UE, WMFT, MAL, swallowing function, MBI and adverse events (figure 7). Egger’s test corroborated this, showing no significant publication bias for FMA-UE (Egger’s test p=0.630), WMFT (Egger’s test p=0.650), MAL (Egger’s test p=0.341), swallowing function (Egger’s test p=0.435), MBI (Egger’s test p=0.900) and adverse events (Egger’s test p=0.174). However, due to having fewer than three studies available for other outcomes, it was not possible to conduct publication bias assessments for them.

Figure 6. Sensitivity analysis of (A) FMA-UE, (B) WMFT, (C) MAL, (D) MBI, (E) adverse events and (F) swallowing function. FMA-UE, Fugl-Meyer Assessment of Upper Extremity; MAL, Motor Activity Log; MBI, Modified Barthel Index; WMFT, Wolf Motor Function Test.

Figure 7. Funnel plot of (A) FMA-UE, (B) WMFT, (C) MAL, (D) MBI, (E) adverse events and (F) swallowing function. FMA-UE, Fugl-Meyer Assessment of Upper Extremity; MAL, Motor Activity Log; MBI, Modified Barthel Index; RR, risk ratio; SMD, standardised mean difference; WMFT, Wolf Motor Function Test.

GRADE rating

This study analysed all outcome measures, revealing that MAL, swallowing function test, MEP latency, RMS and FIM were of high quality; FMA-UE, MBI, adverse events, SIS and BBT were of moderate quality; and WMFT and spasticity were of low quality. Detailed grading results are presented in online supplemental table S4.

Discussion

Stroke is a dominant cause of mortality and chronic disability among middle-aged and older adults in China. Despite advances in medical technology, the incidence and prevalence of stroke continue to rise, leading to a high disability rate and an increasing need for effective rehabilitation solutions.³⁰ As neurorehabilitation concepts evolve, central intervention techniques have become integral to stroke rehabilitation, establishing themselves as key modalities in neurorehabilitation.¹⁹ Among these, VNS is an emerging brain stimulation technique that uses various stimulation forms to target the vagus nerve network. Recent studies have increasingly focused on the impact of VNS on functional recovery in post-stroke patients. For instance, Wang et al conducted a double-blind RCT demonstrating the efficacy of VNS combined with task-oriented training, promoting recovery of task-specific functions through modulation of bilateral cortical excitability.¹⁵ Another double-blind, randomised, placebo-controlled trial on post-stroke depression demonstrated that a combined approach of taVNS with conventional therapy significantly enhanced both efficacy and tolerability.¹³

Our study concluded that the adjunctive use of VNS substantially improved functional rehabilitation in stroke patients, encompassing motor impairment, motor function, real-world arm use quantity and quality, hand function, manual dexterity, swallowing function, ADL, electrical muscle activity and MEP. In addition, there was no significant difference in adverse events between the VNS and control groups. Sensitivity analysis indicated no significant sensitivity sources and no significant publication bias in this study. The previous meta-analysis¹⁰ was consistent with our findings, indicating that VNS significantly improves FMA-UE and WMFT scores. However, it included only six and two articles, respectively, representing a limited literature base. Our study addressed the limitations of their meta-analysis. We substantiated the efficacy of VNS in enhancing upper limb motor function post-stroke through a larger dataset and reported no significant differences in adverse events between groups, reinforcing the position of VNS as a safe and effective approach to enhance recovery of upper extremity function in stroke patients. Furthermore, we further found that VNS improved the quantity and quality of actual arm use, swallowing function and ADL in stroke patients. We identified significant improvement in the strength of wrist extensor muscles and a notable reduction in MEP latency within the VNS group, indicative of enhanced nerve conduction from the cerebral cortex motor areas to the hand.

Subgroup analysis suggested that invasive VNS may be associated with greater improvement in upper limb motor function than non-invasive VNS in stroke patients, although this finding was based on a limited number of studies. The auricular branch of the vagus nerve (ABVN), being the only superficial branch, predominantly innervates the auricular concha, with full innervation in the cymba conchae and partial innervation (45%) in the cavum concha, the remainder being served by great auricular nerve.³¹ Among the three non-invasive VNS articles included in the analysis, two targeted the cymba conchae, while one stimulated the auricular concha. The central projections of the ABVN are in agreement with established central vagal pathways, allowing non-invasive access through the external ear³² and reaching regions such as brainstem afferent vagal nuclei in rats with stroke,³³ and the motor cortex, insula, precentral gyrus and thalamus in healthy individuals.^{34 35} This suggests that taVNS might use mechanisms or pathways similar to those of conventional VNS. However, Ay et al³³ observed that while taVNS significantly reduced infarct volume in rats, its effect size was less pronounced compared with VNS (28% vs 50%). This disparity could stem from the less potent activation of central vagal pathways by taVNS, as evidenced by the reduced c-Fos expression in the nucleus tractus solitarius (NTS). Additionally, individual variability in the innervation patterns of sensory nerves within the cavum concha could contribute to differences in effect size, with significant variability reported in external ear innervation.

Interestingly, stimulation of cortical areas innervated by the trigeminal nerve, which supplies much of the outer ear, including regions adjacent to the vagal area,^{31 36} did not elicit the typical activation pattern observed with vagal stimulation. This finding underscores the necessity of precise targeting of the vagal cortical area for consistent electrophysiological responses. Consequently, it is essential to consider whether taVNS in stroke patients may inadvertently stimulate other nerves, potentially diminishing the efferent effect and impacting therapeutic outcomes. Future research should prioritise more precise localisation techniques and compare the effects of electrical stimulation across different auricular concha regions on the vagal pathway activation in stroke patients. To conclusively determine whether a significant difference between invasive and non-invasive VNS in improving motor function post stroke exists, larger scale multicentre studies with robust sample sizes are warranted.

Subgroup analysis indicated that VNS treatment extending beyond 6 weeks demonstrated superior efficacy in improving upper limb motor function compared with shorter treatment durations. Chang et al¹³ observed weaker improvement in motor function within their study compared with two other investigations.^{8 37} They attributed this to the limited treatment regimen, which consisted of only undergoing nine training sessions (three sessions/week for 3 weeks) with approximately 250 stimulated movements per session. In contrast, Dawson et al⁶ reported a more extensive protocol where patients received 18 sessions (three sessions/week for 6 weeks), with each session including more than 300 stimulation exercises at 0.8 mA, 30 Hz. This protocol also incorporated a 30 min daily home exercise regime and continuous VNS delivery every 10 s for 30 min, extending to a 90-day follow-up. Similarly, Capone et al³⁷, using taVNS, used a high-dose stimulation regimen. Their protocol involved 10 consecutive daily taVNS treatments, each lasting 30 s at 0.8 mA, 30 Hz, every 5 min for 1 hour before wrist or shoulder/elbow exercises. Therefore, Chang et al¹⁴ concluded that their study participants were likely undertreated. These findings suggest that prolonged stimulation periods and increased stimulation frequency may enhance the efficacy of VNS in improving upper limb motor function post stroke. Nevertheless, additional research is necessary to definitively establish the optimal treatment duration and intensity.

The ABVN passes through the vagal trunk, culminating in the NTS. The NTS serves as a critical integration centre, relaying sensory information to higher brain regions, including the thalamus, hypothalamus, cerebral cortex and limbic system. This neural pathway is crucial for the enhancement of neurological function following a stroke.³⁴ VNS has been shown to promote norepinephrine release from the locus coeruleus, stimulating cholinergic neurons to produce acetylcholine.^{38 39} The presence of norepinephrine and acetylcholine fosters cortical disinhibition, enhances cortical facilitation and induces extensive reorganisation of the brain tissue, affecting sensory and motor networks.^{40 41} This neuroplasticity forms the foundation for recovery across various neurological domains, including motor function, swallowing, cognitive and other higher brain functions.^{9 23}

VNS exerts multifaceted neuroprotective and neuroregenerative effects in stroke recovery. It upregulates the expression of neurotrophic factors, including brain-derived nerve growth factor and basic fibroblast growth factor, promoting nerve regeneration and synaptic remodelling. This facilitates motor cortex reorganisation and supports the recovery of motor function.42,44 Additionally, VNS modulates the immune response by inhibiting immune cell activation and inflammatory factor release, thereby attenuating neuroinflammation and mitigating neuronal damage.⁴⁵ Early application of VNS also helps protect the blood–brain barrier by decreasing inflammatory factors expression, thus safeguarding nerve cells.⁴⁶

This study comprehensively analysed 18 RCTs investigating VNS as a therapeutic intervention for stroke patients. The analysis demonstrated the safety and efficacy of VNS as an adjunctive therapy in stroke rehabilitation. However, several limitations warrant consideration. First, the included RCTs had relatively small sample sizes and lacked large-scale, multicentre trials. Most studies originated from Asia (8/18), which may introduce publication bias and regional selection bias. Consequently, generalising the global applicability of VNS is premature, as original studies from other regions are not yet available. Second, some data exhibited high heterogeneity, such as outcomes related to the FMA-UE, WMFT, spasticity and MBI. But subgroup analyses identified sources of some of the heterogeneity, with FMA-UE variability linked to stimulation modality and region, and WMFT heterogeneity associated with stimulation modality, treatment duration, region and mean age. The low GRADE rating of the WMFT outcome (due to imprecision) indicates that its observed positive effect should be interpreted with caution and requires further verification. In contrast, the high-quality evidence for both the MAL and swallowing function outcomes supports more reliable conclusions.

Conclusion

This meta-analysis indicates that VNS can enhance upper limb motor function, the quantity and quality of arm use, swallowing function, extensor carpi radialis muscle strength, ADL and MEP latency in stroke patients, without significant adverse effects. Invasive VNS demonstrated a numerically greater improvement in FMA-UE scores; however, only two invasive studies were included. This difference is preliminary and requires validation through large-scale RCTs. Given the small sample sizes, regional limitations and data heterogeneity in this study, future large-scale studies, multicentre trials, especially in non-Asian populations, are necessary to confirm the precise efficacy and safety of VNS for functional recovery after stroke. Particular emphasis should be placed on comparing the effects of invasive VNS and non-invasive VNS modalities, as well as elucidating the optimal treatment duration.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.