Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Authors: Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan, Yuguang Guan Published: 2026 ABSTRACT Background:Disorders of consciousness (DoC) have limited effective therapies. Implantable vagus nerve stimulation (VNS) is established for epilepsy but remains underexplored for DoC with comorbid epilepsy.Objectives:To evaluate whether VNS is associated with improved consciousness recovery in DoC patients with epilepsy and to explore factors related to response.Design:Retrospective propensity score-matched cohort study.Methods:Ninety DoC patients with epilepsy received implantable VNS or conservative management. The primary outcome was clinically meaningful improvement at 1 year (Coma Recovery Scale-Revised (CRS-R) increase ⩾3 points). Propensity score matching (1:1) balanced baseline characteristics (23 per group). Longitudinal CRS-R trajectories were assessed using linear mixed-effects models with Type III ANOVA. Logistic regression within the VNS cohort explored factors associated with responder status; adverse events were extracted from operative and follow-up records.Results:The 1-year responder rate was higher with VNS than with conservative management (34.78% vs 4.35%; two-sided Fisher’s exact test, p = 0.022). Longitudinal analyses showed a significant group × time interaction (χ2 = 43.535, p < 0.001) with greater CRS-R gains from 3 months onward in the VNS group. Within the VNS cohort, responder status was associated with baseline minimally conscious state (aOR = 9.750, 95% confidence interval (CI) 1.592–59.695; p = 0.014) and better seizure control (McHugh classification; aOR = 22.667, 95% CI 3.140–163.629; p = 0.002). Traumatic etiology was not associated with 12-month net CRS-R improvement after adjustment for baseline CRS-R (β = 0.893, 95% CI −1.583 to 3.369; p = 0.466). Five patients reported stimulation-related hoarseness/dysphonia, and two had surgical-site complications; no device removal occurred.Conclusion:Implantable VNS was associated with higher 1-year clinically meaningful improvement and greater longitudinal CRS-R gains than conservative management in DoC patients with epilepsy. Prospective controlled studies are warranted. Introduction Traumatic brain injury (TBI) often impairs consciousness, with severity-dependent outcomes. 1 Large observational studies indicate that 0.63%–7.33% of severe TBI patients presenting in coma progress to a vegetative state (VS) within 6 months, though incidence varies substantially across treatment centers.2,3 Disorders of consciousness (DoC) comprise a clinical spectrum after severe brain injury, including coma, unresponsive wakefulness syndrome (UWS)/VS (wakefulness without behavioral evidence of awareness), and minimally conscious state (MCS; inconsistent but reproducible signs of awareness), with emergence from MCS (EMCS) defined by functional communication and/or object use. Because diagnosis is primarily behavioral, repeated standardized assessments (e.g., Coma Recovery Scale-Revised and CRS-R) are recommended to reduce misclassification. In parallel, recent EEG/fMRI studies have highlighted cognitive motor dissociation (covert consciousness), underscoring preserved cognitive capacity in some patients despite limited motor output.4 –9 Current therapeutic approaches include behavioral interventions, pharmacotherapy, and neuromodulation, with surgical options primarily limited to spinal cord stimulation (SCS) and deep brain stimulation (DBS)—the former being more frequently employed. While SCS has demonstrated efficacy in MCS patients, 10 its therapeutic benefits, like other interventions, require further validation. Seizures frequently complicate post-traumatic DoC, though their potential role in consciousness impairment remains unclear. These complex clinical presentations create substantial societal burdens and pose significant management challenges. Vagus nerve stimulation (VNS), approved by the FDA for epilepsy treatment in 1997, shows particular promise for refractory cases despite incomplete mechanistic understanding. A comprehensive meta-analysis of 74 studies (n = 3321) demonstrated that VNS reduces seizure frequency by ⩾50% in half of treated patients, with maximal benefits emerging beyond 1 year post-implantation. 11 Beyond seizure control, VNS yields multiple ancillary benefits including mood stabilization,12,13 reduced daytime somnolence, enhanced alertness, improved slow-wave sleep architecture, 14 and cognitive/memory enhancement, 15 collectively improving quality of life. 13 The therapeutic scope of VNS has expanded to include depression, 16 and post-stroke movement disorders, 17 with Shi et al. 18 first reporting its potential for consciousness recovery in severe TBI in 2013. Clinical evidence supporting VNS for DoC remains limited and heterogeneous. Implanted VNS has primarily been documented in a landmark single-case report and small open-label series, while non-invasive transauricular VNS has advanced to feasibility cohorts and sham-controlled randomized trials in DoC populations.19,20 The emerging literature indicates a potential therapeutic benefit, albeit with substantial variability across different patient subtypes and etiologies. Prior studies have seldom focused on DoC patients with comorbid epilepsy or systematically reported seizure outcomes, leaving this important clinical subgroup largely underexplored. This investigation retrospectively analyzed outcomes in consecutive DoC patients receiving either VNS or conservative management at our institution. We evaluated long-term treatment effects, examined potential correlations between seizure control and consciousness improvement, and explored preoperative predictive factors to optimize patient selection for precision neuromodulation therapy. Methods Patient selection This retrospective cohort study enrolled patients with prolonged disorders of consciousness (pDoC) and comorbid epilepsy who underwent VNS therapy or conservative management at Sanbo Brain Hospital, Capital Medical University (Beijing, China) between March 2020 and March 2024. Investigators documented baseline characteristics including sex, diagnosis, etiology, surgical age, and DoC duration, along with epilepsy features such as injury-to-seizure latency, seizure type and frequency, and epileptiform discharge localization via video electroencephalography (VEEG). Inclusion criteria required: (1) age ⩾5 years; (2) pDoC diagnosis confirmed by two neurologists using MRI/CT, clinical presentation, EEG, and CRS-R scores (including patients aged 5–18) 21 ; (3) Epilepsy was defined as a documented diagnosis and/or ⩾2 unprovoked clinical seizures, with at least one documented seizure within the preceding 6 months and epileptiform activity on EEG; (4) ⩾3 months of stable DoC (⩾6 months for traumatic cases) without consciousness improvement for >4 weeks (>8 weeks for trauma); (5) VNS or conservative management; and (6) absence of surgical contraindications. Exclusion criteria comprised: (1) pre-existing conditions (e.g., deafness, intellectual disability) precluding accurate CRS-R assessment; (2) receipt of other neuromodulation/neurostimulation interventions (e.g., tDCS, TMS, DBS, or SCS) within 6 months prior to baseline assessment or during the early post-baseline period; and (3) follow-up <12 months or missing 12-month CRS-R assessment. Assessment Two independent neurosurgeons conducted evaluations at admission, pre-VNS activation (⩾1 week post-admission), and follow-ups (1, 3, 6, and 12 months post-implantation). Radiological assessment included CT (2-mm axial, Philips Medical Systems, Best, the Netherlands) and, when clinically indicated, MRI (1.5-T MAGNETOM Avanto, Siemens Healthcare, Erlangen, Germany, or 3.0-T Discovery MR750, GE Healthcare, Waukesha, WI, USA). All patients underwent ⩾16-h VEEG monitoring, supplemented where possible by auditory brainstem responses, somatosensory evoked potentials, and mismatch negativity. Baseline consciousness state was determined prior to treatment through repeated CRS-R assessments, with patients categorized according to established behavioral diagnostic criteria; this classification relied on CRS-R item-level behaviors rather than a single total-score cutoff. Specifically, pDoC severity was classified as: (1) coma, characterized by absent consciousness and no sleep–wake cycles; (2) UWS, defined by wakefulness without reproducible evidence of awareness; (3) MCS, marked by inconsistent but reproducible evidence of awareness; or (4) EMCS, indicated by functional communication and/or functional object use. MCS was further stratified into MCS− and MCS+ based on the presence of language-related behaviors, such as command following, versus non-language behaviors only, such as visual pursuit or localization to noxious stimulation.7,22 Epilepsy characterization followed International League Against Epilepsy 2017 criteria, with baseline established using seizure frequency/type during the 3-month pre-admission period. Assessments incorporated inpatient, outpatient, and telemedicine evaluations. As pDoC precluded aura observation, this parameter was excluded. Observers documented stimulation parameters (amplitude (V), frequency (Hz), pulse width (μs)) and adverse events. These included surgical-site complications such as infection, hematoma, seroma, or wound dehiscence. Lead- or generator-related complications were also noted. Stimulation-related side effects encompassed hoarseness, dysphonia, cough, throat discomfort, dysphagia, and dyspnea. In addition, cardiovascular events such as bradycardia or arrhythmia were documented, as were instances of seizure worsening or status epilepticus. Surgical procedures Following successful general anesthesia, the patient was placed in the supine position with the head turned approximately 30° to the right. A 4-cm transverse incision was made on the left side of the neck along the anterior border of the sternocleidomastoid muscle. After sharp and blunt dissection through the skin, subcutaneous tissue, and platysma, the carotid sheath was opened. The left cervical vagus nerve was identified between the carotid artery and internal jugular vein and was carefully isolated over an approximately 3-cm segment. An implantable VNS system was used (pulse generator model G112; PINS Medical Co., Ltd., Beijing, China; or model VNS201A; Rishena Medical Co., Ltd, Changzhou, Jiangsu, China). The helical lead/electrode was wrapped around the vagus nerve and secured to adjacent soft tissue to reduce traction on the lead. A second 4-cm incision was made in the left infraclavicular/anterior axillary region. A generator pocket (approximately 5 × 5 cm) was created in the subcutaneous/subfascial plane superficial to the pectoralis major fascia, and the pulse generator was positioned within the pocket. A subcutaneous tunnel was then created from the chest pocket cranially over the clavicular region toward the cervical incision, and the lead was passed through the tunnel to connect with the generator. Intraoperative system checks were performed, including confirmation of device connectivity and acceptable lead impedance, and the VNS system was activated according to the manufacturer’s recommendations. Hemostasis was achieved, and both incisions were closed in layers and covered with sterile dressings. Programming strategy of VNS After the postoperative period (at least 7 days later), the start-up parameters were set to a current of 0.5 mA with a pulse width of 250 µs and a frequency of 30 Hz. After 2 weeks, they were adjusted to a current of 0.7 mA with a pulse width of 250 µs and a frequency of 30 Hz, and, after 4 weeks, to a current of 1.0 mA with a pulse width of 250 µs and a frequency of 30 Hz. Then, the stimulation intensity was adjusted according to the patient’s recovery and tolerance level, with a maximum current intensity of 3.0 mA. During initial activation and each parameter titration, patients were monitored for vital signs (heart rate, blood pressure, oxygen saturation, and respiratory rate) under clinical supervision. Stimulation was paused, or the output current was reduced if clinically significant intolerance occurred (e.g., bradycardia/arrhythmia, hypotension, desaturation, respiratory distress, or marked seizure worsening). Conservative management All patients received standard-of-care management for pDoC and epilepsy according to their clinical status. In the conservative group, this standard management was provided without VNS implantation. In the VNS group, standard management was continued alongside the implanted VNS therapy. Standard management included: (1) routine arousal-promoting pharmacotherapy used in our center—amantadine (1 tablet twice daily) and madopar (1 tablet twice daily); for pediatric patients, doses were reduced according to body weight; (2) optimization of antiseizure medications and treatment of acute seizures as clinically indicated; and (3) comprehensive supportive and rehabilitative care, including airway/respiratory care, nutritional support, prevention and treatment of infection, thromboprophylaxis, pressure injury prevention, and individualized rehabilitation therapies when applicable. No additional neuromodulation/neurostimulation interventions (e.g., tDCS, TMS, DBS, and SCS) were administered during the study period. Selection for VNS was based on routine clinical decision-making after multidisciplinary evaluation by the treating team and discussion with the patients’ families/legal guardians. Patients underwent VNS after confirmation of eligibility, absence of surgical contraindications, and informed consent from the family/legal guardian. Outcome evaluation The recovery outcome was assessed using the CRS-R at baseline and at 1, 3, 6, and 12 months during follow-up as part of routine clinical care. Trained clinicians performed these assessments according to a standardized protocol, and the scores were extracted from medical records for this retrospective analysis. In this real-world retrospective setting, evaluators were not routinely blinded to treatment allocation, and inter-rater reliability was not formally quantified. A CRS-R increase of ⩾3 points was considered a clinically meaningful improvement, consistent with prior DoC studies.23,24 Seizure prognosis was evaluated using the VNS efficacy criteria established by McHugh et al., which primarily track changes in seizure frequency during follow-up. These outcomes were categorized into five grades: Grade I (>80% reduction), Grade II (50%–79% reduction), Grade III (<50% reduction), Grade IV (response only to magnet use), and Grade V (no improvement). Grades I and II were classified as effective, while Grades III through V were considered ineffective. Statistical analysis Statistical analyses were conducted using SPSS 26.0 (IBM Corp., Armonk, NY, USA), Prism 10.1.2 (GraphPad Software, LLC, Boston, MA, USA), and R 4.5.0 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were reported as mean ± standard deviation (SD) or median with interquartile range (IQR) based on their distribution. Normality was evaluated using the Shapiro–Wilk test. Non-normally distributed variables were compared using Wilcoxon rank-sum tests. Categorical variables were assessed with Chi-square or Fisher’s exact tests as appropriate. Propensity scores were estimated via logistic regression, modeling treatment assignment as a function of the pre-treatment covariates age, gender, pretreatment CRS-R score, consciousness status, etiology, duration of impaired consciousness, and monthly seizure frequency. We then performed 1:1 nearest-neighbor matching without replacement, applying a caliper width of 0.3 SDs of the logit propensity score. Covariate balance was assessed by comparing standardized mean differences (SMDs) before and after matching, where an absolute SMD below 0.1 indicated adequate balance. Treatment efficacy and other binary outcomes were examined through doubly robust-adjusted logistic regression. Longitudinal CRS-R score changes were modeled with linear mixed-effects analyses incorporating random intercepts; group × time interactions and main effects were tested via Type III ANOVA with Chi-square approximation. The relationship between epilepsy control and CRS-R improvement was examined through linear regression with 95% confidence intervals (CIs), supported by scatter plot visualization. When multiple univariable hypothesis tests were performed within the same analysis, p-values were adjusted using the Benjamini–Hochberg false discovery rate (BH-FDR) procedure, and the corresponding q values are reported where applicable. Statistical significance was set at p < 0.05 (two-tailed); for FDR-adjusted results, q < 0.05 was considered statistically significant. Results Baseline characteristics This study enrolled 90 patients with DoC complicated by epilepsy. Supplemental Material 1 presents the baseline characteristics before propensity score matching (PSM), which accounted for age, gender, pretreatment CRS-R scores, consciousness status, etiology, duration of impaired consciousness, and monthly seizure frequency to balance intergroup differences. After PSM, 46 patients qualified for comparative analysis, with the enrollment process detailed in Figure 1. Flowchart of patient enrollment. Following PSM, baseline covariates were well balanced between the VNS and conservative treatment groups, with all matched covariates achieved |SMDs| <0.1. This result confirms adequate post-matching balance, as detailed in Supplemental Material 2, while Table 1 summarizes the matched cohort’s characteristics. Table 1.Baseline characteristics of patients after PSM.VariableVNS (n = 23)Conservative (n = 23)SMDAge, years, median (IQR)35.0 (22.0–48.0)36.0 (25.0–47.0)0.045Gender (%) Male13 (56.52%)13 (56.52%)0.087 Female10 (43.48%)10 (43.48%)State of consciousness (%) MCS11 (47.83%)11 (47.83%)0.000 VS/UWS12 (52.17%)12 (52.17%)Pathogeny (%) Trauma8 (34.78%)8 (34.78%)0.043 Stroke7 (30.43%)7 (30.43%)0.043 Ischemia-hypoxia8 (34.78%)8 (34.78%)0.087Duration of DoC at admission (%) 3–5 months6 (26.09%)6 (26.09%)0.087 6–11 months9 (39.13%)9 (39.13%)0.000 ⩾12 months8 (34.78%)8 (34.78%)0.087Monthly seizure frequency >1 time/month11 (47.83%)11 (47.83%)0.087 ⩽1 time/month12 (52.17%)12 (52.17%)DoC, disorders of consciousness; IQR, interquartile range; MCS, minimally conscious state; PSM, propensity score matching; SMD, standardized mean difference; UWS, unresponsive wakefulness syndrome; VNS, vagus nerve stimulation; VS, vegetative state. Baseline characteristics of patients after PSM. DoC, disorders of consciousness; IQR, interquartile range; MCS, minimally conscious state; PSM, propensity score matching; SMD, standardized mean difference; UWS, unresponsive wakefulness syndrome; VNS, vagus nerve stimulation; VS, vegetative state. The median ages were 35.0 and 36.0 years in the respective groups. Males predominated in both cohorts (56.52%, n = 26), and VS/UWS represented the most frequent consciousness impairment (52.17%, n = 24). Etiologies included trauma (34.78%, n = 16), stroke (30.43%, n = 14), and ischemia-hypoxia (34.78%, n = 16). Most patients (39.13%, n = 18) had experienced DoC for 6–11 months. Seizure frequency was ⩽1 time/month in 52.17% (n = 24) of cases and >1 time/month in 47.83% (n = 22). Improvement in DoC after treatment Patient characteristics were examined descriptively according to 1-year clinical response, defined as a CRS-R increase of ⩾3 points (Table 2). Nine of 46 patients (19.57%) met this response criterion. Responders presented with MCS more frequently than non-responders. Other univariable comparisons—including age, sex, etiology, DoC duration, and baseline seizure frequency—showed no clear differences between groups. As these comparisons were exploratory and involved multiple testing, they were interpreted with caution and were not used as the primary basis for inferring treatment effects. In the PSM cohort (n = 46; 23 per group), the 1-year responder rate (CRS-R increase ⩾3 points) was higher in the VNS group than in the conservative group (8/23 (34.78%) vs 1/23 (4.35%); two-sided Fisher’s exact test, p = 0.022). Furthermore, we analyzed how treatment method and follow-up duration (1, 3, 6, and 12 months) jointly influenced clinical improvement in 46 patients. Type III ANOVA demonstrated significant main effects of time (χ2 = 125.416, df = 4, p < 0.001, Figure 2) and group × time interaction (χ2 = 43.535, df = 4, p < 0.001, Figure 2). The linear mixed-effects model (LMM, Supplemental Material 3) indicated poorer CRS-R score improvement in the conservative group versus the VNS group at 3 months (β = −1.391, SE = 0.433, 95% CI −2.240 to −0.542, p = 0.001), 6 months (β =−1.913, SE = 0.433, 95% CI −2.762 to −1.064, p < 0.001), and 12 months (β = −2.565, SE = 0.433, 95% CI −3.414 to −1.716, p < 0.001), though no differences emerged at 1 month. These findings indicate greater cumulative improvement in CRS-R scores over time in patients receiving VNS. Table 2.Clinical information of patients according to their outcome.VariablesEffective (n = 9)Unchanged/died (n = 37)p Value (effective vs unchanged/died)FDR qAge, years, median (IQR)34.0 (14.0–39.5)34.0 (23.0–47.0)0.2730.382Gender (%)0.1520.266 Male7 (77.78%)19 (51.35%) Female2 (22.22%)18 (48.65%)State of consciousness (%)0.0440.154 MCS7 (77.78%)15 (40.54%) VS/UWS2 (22.22%)22 (59.46%)Pathogeny (%)0.6580.658 Trauma4 (44.44%)12 (32.43%) Stroke3 (33.33%)11 (29.73%)Ischemia-hypoxia2 (22.22%)14 (37.84%) Duration at admission (%)0.3700.432 3–5 months4 (44.44%)8 (21.62%) 6–11 months2 (22.22%)16 (43.24%) ⩾12 months3 (33.33%)13 (35.14%)Monthly seizure frequency0.1300.266 ⩽1 time/month4 (44.44%)15 (40.54%) >1 time/month5 (55.56%)22 (59.46%)Treatment (%)0.0220.154 VNS8 (88.89%)15 (40.54%) Conservation1 (11.11%)22 (59.46%)All p-values compare the Effective group versus the Unchanged/died group. Age was compared using the Wilcoxon rank-sum test. Categorical variables were compared using Fisher’s exact test or χ2 test as appropriate. For “Treatment,” the p value was derived from Fisher’s exact test assessing the association between treatment assignment (VNS vs conservative management) and outcome group (effective vs unchanged/died). BH-FDR q values were calculated across all univariable tests reported in Table 2.BH, Benjamini–Hochberg; DoC, disorders of consciousness; FDR, false discovery rate; IQR, interquartile range; MCS, minimally conscious state; UWS, unresponsive wakefulness syndrome; VNS, vagus nerve stimulation. Clinical information of patients according to their outcome. All p-values compare the Effective group versus the Unchanged/died group. Age was compared using the Wilcoxon rank-sum test. Categorical variables were compared using Fisher’s exact test or χ2 test as appropriate. For “Treatment,” the p value was derived from Fisher’s exact test assessing the association between treatment assignment (VNS vs conservative management) and outcome group (effective vs unchanged/died). BH-FDR q values were calculated across all univariable tests reported in Table 2. BH, Benjamini–Hochberg; DoC, disorders of consciousness; FDR, false discovery rate; IQR, interquartile range; MCS, minimally conscious state; UWS, unresponsive wakefulness syndrome; VNS, vagus nerve stimulation. Improvement in CRS-R scores over time in patients in conservative and VNS groups. CRS-R scores gradually increased over time, with VNS group demonstrated a more rapid rate of improvement. p < 0.05 indicates a significant difference; **p < 0.001. CRS-R, Coma Recovery Scale-Revised; VNS, vagus nerve stimulation. Subgroup analysis of patients receiving VNS We next evaluated which patient subgroups derived the greatest benefit from VNS therapy. Subgroup analyses of 1-year consciousness improvement outcomes (Figure 3) revealed no significant efficacy differences based on age (p = 0.959), sex (p = 0.229), DoC duration (p = 0.769), or epilepsy frequency (p = 0.192). However, treatment response varied significantly by etiology (Trauma: aOR = 6.600 (95% CI: 1.236–35.234); non-trauma: aOR = 1.891 (95% CI: 0.425–8.406); p = 0.027) and consciousness state (MCS: aOR = 12.800 (95% CI: 2.395–68.407); VS/UWS: aOR = 1.771 (0.137–4.355); p = 0.003). Subgroup analysis for the improvement in consciousness at 1 year using multivariate regression analysis adjusted for age, gender, pathogeny, DoCs, state of consciousness, and frequency of epilepsy. p < 0.05 indicates a significant difference. DoCs, duration of consciousness; MCS, minimally conscious state; VNS, vagus nerve stimulation; VS/UWS, vegetative state/unresponsive wakefulness syndrome. Analysis of related factors of the therapeutic effects of VNS We further analyzed the potential factors associated with the therapeutic effects of VNS in patients with DoCs. Multivariate analysis identified consciousness state (MCS: aOR = 9.750 (1.592–59.695), p = 0.014) and epilepsy control efficacy (McHugh score: 22.667 (3.140–163.629), p = 0.002) were significantly associated with therapeutic response to VNS (Table 3). Table 3.Logistic analysis of factors related to improvement in consciousness after VNS.VariableMultivariate analysisResponders (N = 11, aOR (95% CI))p ValueAge0.949 (0.899–1.002)0.059Gender, male0.338 (0.068–1.678)0.184State of consciousness, MCS9.750 (1.592–59.695)0.014*Pathogeny Ischemia-hypoxia–– Trauma3.750 (0.342–41.081)0.279 Stroke3.333 (0.276–40.287)0.344Duration at admission ⩾12 month–– 3–5 month7.000 (0.861–56.895)0.069 6–11 month1.167 (0.151–9.006)0.882Epilepsy control effect (McHugh)22.667 (3.140–163.629)0.002*Responders were defined as CRS-R increase ⩾3 points at 12 months among VNS-treated patients; n = 11 indicates the number of responders (events) used in the logistic regression. Total VNS-treated sample size included in this analysis was n = 30..*p < 0.05.CI, confidence interval; CRS-R, Coma Recovery Scale-Revised; MCS, minimally conscious state; OR, odds ratio; VNS, vagus nerve stimulation. Logistic analysis of factors related to improvement in consciousness after VNS. Responders were defined as CRS-R increase ⩾3 points at 12 months among VNS-treated patients; n = 11 indicates the number of responders (events) used in the logistic regression. Total VNS-treated sample size included in this analysis was n = 30.. p < 0.05. CI, confidence interval; CRS-R, Coma Recovery Scale-Revised; MCS, minimally conscious state; OR, odds ratio; VNS, vagus nerve stimulation. To further examine whether etiology was associated with the magnitude of consciousness improvement within the VNS-treated cohort, we performed a linear regression analysis with 12-month net CRS-R improvement as the dependent variable and traumatic etiology (trauma vs non-trauma) as the independent variable, adjusting for baseline CRS-R. Traumatic etiology was not significantly associated with net improvement (β = 0.893, 95% CI −1.583 to 3.369, p = 0.466). Within the VNS group, the consciousness state × time interaction significantly influenced treatment outcomes (Type III ANOVA: χ2 = 70.738, df = 4, p < 0.001; Figure 4), with both time (χ2 = 213.654, df = 4, p < 0.001) and group effects (χ2 = 8.273, df = 1, p = 0.004) showing significance. LMM analysis demonstrated that VS patients exhibited poorer recovery than MCS patients at 3 months (β = −1.667, SE = 0.588, 95% CI −2.819 to −0.515, p = 0.005), 6 months (β = −2.600, SE = 0.588, 95% CI −3.752 to −1.448, p < 0.001), and 12 months (β = −4.600, SE = 0.588, 95% CI −5.752 to −3.448, p < 0.001; Supplemental Material 4). Improvement in CRS-R scores over time in VNS patients according to state of consciousness. CRS-R scores gradually increased over time, with MCS patients demonstrating a more rapid rate of improvement. p < 0.05 indicates a significant difference; *p < 0.05; **p < 0.001. CRS-R, Coma Recovery Scale-Revised; MCS, minimally conscious state; VNS, vagus nerve stimulation; VS/UWS, vegetative state/unresponsive wakefulness syndrome. Patients with ineffective seizure control showed significantly less consciousness improvement beginning 1 month postoperatively (Supplemental Material 5). Meanwhile, the model indicated that starting from the first month after surgery, the patients’ McHugh scores were significantly correlated with the improvement in consciousness state (the p-values were 0.011 at 1 month, <0.001 at 3 months, <0.001 at 6 months, and <0.001 at 12 months), and this correlation continued to strengthen as time elapsed after surgery (Figure 5). Correlation between CRS-R and Mchugh scores in VNS patients across follow-up intervals. Postoperative CRS-R and Mchugh scores exhibited significant temporal correlations, with the association strengthening over longer follow-up durations. CRS-R, Coma Recovery Scale-Revised; VNS, vagus nerve stimulation; Effect of VNS on CRS-R subscale scores VNS significantly improved scores across all four subscales (motor: p < 0.001; audio: p = 0.039; communication: p = 0.005; awake: p = 0.009; Wilcoxon rank-sum test, Figure 6(a)). Among MCS patients receiving VNS, improvements reached significance for all six subscales (oral: p = 0.016; visual: p = 0.043; motor: p < 0.001; audio: p < 0.001; communication: p = 0.041; awake: p < 0.001; Wilcoxon rank-sum test, Figure 6(b)). Similarly, VNS-treated patients with refractory seizures showed significant improvements in five subscales (oral: p = 0.001; motor: p < 0.001; audio: p < 0.001; communication: p = 0.009; awake: p < 0.001; Wilcoxon rank-sum test, Figure 6(c)). Radar chart for improvement in CRS-R subscores. (a) VNS group and conservative group in overall patients. (b) Patients with MCS and VS/UWS underwent VNS. (c) Patients underwent VNS with effective/ineffective epilepsy improvement. CRS-R, Coma Recovery Scale-Revised; MCS, minimally conscious state; VNS, vagus nerve stimulation; VS/UWS, vegetative state/unresponsive wakefulness syndrome. Safety and adverse events Adverse events were documented from operative records and follow-up visits. Stimulation-related hoarseness or dysphonia occurred in five patients and resolved following parameter adjustment and conservative management. Surgical-site complications, comprising infection or hematoma, developed in two patients and were managed successfully with standard care, without requiring device removal. No other stimulation- or procedure-related adverse events, such as cardiopulmonary events or sustained seizure exacerbation attributable to stimulation, were observed. Discussion Consciousness arises from distributed neural networks encompassing sensory, motor, and associative cortical regions. The thalamus mediates consciousness through thalamocortical loops that integrate signals between subcortical and cortical structures. 25 As the primary neural pathway for brain-body communication, the vagus nerve regulates metabolic homeostasis and modulates pathological processes via reflex arcs.26,27 Its projections extend throughout the brainstem, terminating in the nucleus tractus solitarius, thalamus, amygdala, and hippocampus. 28 Experimental evidence demonstrates VNS enhances motor learning through GABAergic mechanisms 29 and increases metabolic activity in prefrontal, thalamic, and reticular structures in humans. 30 Noradrenergic fibers from the locus coeruleus show enhanced activity during VNS, releasing norepinephrine in thalamocortical circuits critical for arousal. 31 PET imaging reveals VNS-induced activation of the ventromedial prefrontal cortex, a key dopaminergic projection zone. 32 Clinical observations suggest VNS may benefit DoC by improving cerebral perfusion. Animal models demonstrate that VNS enhances spatial memory, accelerates forelimb recovery, and reduces post-traumatic edema.33 –37 Corazzol et al. 38 reported cortical signal propagation and metabolic improvements in a chronic VS patient after VNS, with behavioral gains measurable on the CRS-R scale. Combined VNS and rehabilitation triples corticospinal tract synaptic density compared to rehabilitation alone, 39 potentially through spike-timing-dependent plasticity mechanisms. 40 Dawson et al. 41 documented significant upper limb functional recovery in post-stroke patients receiving VNS, with sustained Fugl-Meyer improvements at 90-day follow-up. Against this mechanistic background and early clinical evidence, we examined the therapeutic effects of implanted VNS in DoC patients with comorbid epilepsy. We observed superior motor-domain CRS-R scores in VNS-treated patients compared with controls (p < 0.001), although verification in larger cohorts is warranted. Patients in MCS responded better than those in VS/UWS, which may reflect relatively preserved functional connectivity and network capacity in MCSs. 33 Exploratory subgroup analyses suggested a larger estimated treatment effect in patients with traumatic etiology. However, within the VNS-treated cohort, traumatic etiology was not significantly associated with 12-month net CRS-R improvement after adjustment for baseline CRS-R, highlighting uncertainty regarding etiology-specific responsiveness and the need for validation in larger cohorts. However, the mechanisms underlying etiology-specific responses remain uncertain and warrant further investigation. Notably, we also observed an association between seizure control and CRS-R improvement, which warrants cautious interpretation. Epileptiform activity and clinical seizures occur relatively frequently in prolonged DoC and have been linked to long-term outcomes, likely mediated by effects on cortical excitability, arousal regulation, and network integrity. 42 VNS engages broadly, projecting neuromodulatory systems, such as the locus coeruleus–norepinephrine pathway, and modulates thalamo-cortical and large-scale cortical dynamics.43 –45 Consequently, reduced seizure burden and improved behavioral responsiveness may reflect a common network-level modulation—or a preserved network capacity to respond to stimulation—rather than a direct causal pathway from seizure reduction to consciousness recovery. 6 Future prospective studies employing standardized EEG monitoring and time-varying modeling will be essential to disentangle these relationships. Our results extend the limited clinical literature on VNS for DoC by examining patients with comorbid epilepsy and by incorporating a propensity score-matched conservative-control cohort. Current evidence for implanted VNS in DoC stems primarily from a landmark single-case report and small open-label series, whereas non-invasive auricular or transcutaneous VNS has advanced from feasibility studies to sham-controlled randomized trials.19,20 Across these heterogeneous studies, a consistent signal is that clinical responsiveness appears more frequently in MCS than in VS/UWS, though effect estimates vary substantially with etiology, chronicity, stimulation protocols, and outcome definitions.23,38,46,47 The present study extends prior work in two clinically meaningful directions. First, we focused on DoC patients with comorbid epilepsy, an important subgroup seldom addressed specifically in prior DoC-VNS research, and provided comparative real-world evidence against a conservatively managed cohort. Second, beyond CRS-R trajectories, we systematically quantified seizure outcomes to enable an exploratory analysis of the association between seizure control and consciousness recovery. This association is clinically intriguing but requires cautious interpretation due to the retrospective design and the possibility that reduced seizure burden and improved arousal reflect shared network-level modifications rather than a direct causal pathway. Current pharmacotherapy for DoC demonstrates limited efficacy. Although amantadine exhibits modest benefits in TBI recovery,7,48 evidence quality is inconsistent.49,50 Zolpidem produces transient neurological effects in rare cases,51,52 but most neural rehabilitation drugs show unsatisfactory outcomes in DoC. Conservative treatments consequently remain the standard approach for most patients. In addition to pharmacological and supportive care, non-invasive neuromodulation approaches are increasingly being explored in DoC, particularly transcutaneous auricular vagus nerve stimulation (taVNS), which has progressed from feasibility studies to sham-controlled randomized trials.20,23 Other non-invasive techniques, including transcranial direct current stimulation and repetitive transcranial magnetic stimulation, have also shown preliminary potential in selected patients, although therapeutic effects remain heterogeneous and require further validation.53,54 Our findings indicate that pharmacological interventions for consciousness improvement prove less effective than neurostimulation methods, regardless of whether patients present with comorbid epilepsy or isolated DoC. Surgical neurostimulation techniques may therefore offer more viable options for consciousness restoration. Among available neuromodulation approaches, SCS currently represents the most established method. Originally developed for chronic pain management, SCS may enhance consciousness by activating ascending pathways such as the reticular activating system. 55 Clinical observations demonstrate SCS-induced changes in cortical activity and metabolism, suggesting therapeutic potential for DoC. 56 The proposed mechanism involves both central and peripheral pathways, with studies reporting increased cerebral blood flow, elevated cerebrospinal fluid catecholamines, and reduced oxidative stress in comatose patients following SCS. 57 Functional MRI studies further reveal enhanced connectivity between the dorsal attention network, visual networks, and components of the default mode network. 58 A 2023 systematic review by Piedade et al. 10 analyzed 19 SCS studies involving 532 DoC patients, reporting an overall response rate of 47.9% (170 UWS and 78 MCS cases). However, significant heterogeneity existed in response criteria across studies, most of which lacked control groups. In the PSM cohort, VNS was associated with a higher 1-year responder rate than conservative management (34.78% vs 4.35%; p = 0.022), and longitudinal mixed-effects analyses further demonstrated greater CRS-R improvement over time in the VNS group. VNS presents additional practical advantages over alternative neuromodulation methods through its simplified surgical protocol and reduced invasiveness. These collective findings position VNS as a potentially valuable therapeutic option for DoC. Other invasive approaches, such as DBS, 59 have been investigated in selected cases, but the overall evidence remains limited by small sample sizes, heterogeneous protocols, and a lack of well-controlled comparative studies. Consequently, further comparative and mechanistic investigations are needed to define optimal patient selection and to validate clinical efficacy. High-quality research specifically addressing DoC with comorbid epilepsy remains scarce. Systematic reviews estimate seizure prevalence at approximately 27% in DoC populations, 60 with antiepileptic drugs constituting the primary management strategy. The pathophysiological relationship between these conditions remains unclear, warranting investigation into shared mechanisms and treatment interactions. Notably, our VNS cohort demonstrated a significant association between consciousness improvement and seizure reduction (p = 0.002), suggesting a potential therapeutic link between these conditions. This study has several limitations. First, as a single-center retrospective study, treatment allocation was not randomized, which limits causal inference and prevents the full exclusion of residual confounding despite propensity score matching and regression-based adjustment. The propensity model did not incorporate some potentially relevant time-varying or incompletely captured factors, such as rehabilitation intensity, changes in antiepileptic or sedative medications during follow-up, and intercurrent medical complications, which may have contributed to outcome variability. Nevertheless, we attempted to mitigate baseline imbalance using PSM with balance diagnostics and applied doubly robust and longitudinal mixed-effects models to obtain consistent estimates across analytic approaches. Second, CRS-R scores were obtained from routine clinical assessments; evaluators were typically aware of treatment assignments, and inter-rater reliability was not formally assessed, potentially introducing measurement variability. Seizure outcomes relied on clinically documented seizure frequency and may therefore be susceptible to reporting bias. Third, while our definition of effectiveness—a CRS-R improvement of at least three points—aligns with common clinical practice, it lacks universal standardization across DoC studies, potentially limiting comparability. Finally, the modest sample size, especially for subgroup and interaction analyses, may have reduced statistical power and increased uncertainty in the effect estimates. Larger prospective multicenter studies employing standardized assessments are needed to confirm these findings and refine patient selection. Conclusion In this retrospective propensity score-matched cohort, implantable VNS was associated with a higher 1-year clinically meaningful improvement rate and greater longitudinal CRS-R gains compared with conservative management in DoC patients with comorbid epilepsy. Within the VNS-treated cohort, responder status was associated with baseline MCS and better seizure control, while etiology-specific differences in improvement require further validation. Prospective controlled studies are warranted to confirm these findings and refine patient selection. Supplemental Material sj-docx-1-tan-10.1177_17562864261452312 – Supplemental material for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Supplemental material, sj-docx-1-tan-10.1177_17562864261452312 for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study by Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan and Yuguang Guan in Therapeutic Advances in Neurological Disorders sj-pdf-1-tan-10.1177_17562864261452312 – Supplemental material for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Supplemental material, sj-pdf-1-tan-10.1177_17562864261452312 for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study by Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan and Yuguang Guan in Therapeutic Advances in Neurological Disorders sj-pdf-2-tan-10.1177_17562864261452312 – Supplemental material for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Supplemental material, sj-pdf-2-tan-10.1177_17562864261452312 for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study by Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan and Yuguang Guan in Therapeutic Advances in Neurological Disorders sj-pdf-3-tan-10.1177_17562864261452312 – Supplemental material for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Supplemental material, sj-pdf-3-tan-10.1177_17562864261452312 for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study by Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan and Yuguang Guan in Therapeutic Advances in Neurological Disorders sj-pdf-4-tan-10.1177_17562864261452312 – Supplemental material for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Supplemental material, sj-pdf-4-tan-10.1177_17562864261452312 for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study by Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan and Yuguang Guan in Therapeutic Advances in Neurological Disorders sj-pdf-5-tan-10.1177_17562864261452312 – Supplemental material for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Supplemental material, sj-pdf-5-tan-10.1177_17562864261452312 for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study by Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan and Yuguang Guan in Therapeutic Advances in Neurological Disorders sj-pdf-6-tan-10.1177_17562864261452312 – Supplemental material for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study Supplemental material, sj-pdf-6-tan-10.1177_17562864261452312 for Implantable vagus nerve stimulation for the treatment of disorders of consciousness with epilepsy: a retrospective propensity score-matched clinical study by Jinyi Zuo, Kaiqiang Ma, Zizhang Cheng, Shu Wang, Xiongfei Wang, Weijin Sun, Junhong Pan, Xintao Peng, Wenlong Su, Haoran Zhang, Jian Zhou, Guoming Luan and Yuguang Guan in Therapeutic Advances in Neurological Disorders