tVNS for drug-resistant epilepsy

This page is intended for UK healthcare professionals. Clinical information, including dosing parameters and effect sizes, is included.

---

The clinical context

Epilepsy affects approximately one in a hundred people in the UK. Two-thirds achieve satisfactory seizure control on anti-seizure medications. The remaining third, defined by the International League Against Epilepsy as having failed two or more adequately trialled ASMs, are classified as having drug-resistant epilepsy (DRE). DRE carries a substantially elevated burden: reduced quality of life, increased SUDEP risk, and significant psychological and social consequences.

For patients in this group, resective surgery offers the highest chance of seizure freedom, but only where an epileptogenic focus can be identified and safely removed. Where surgery is not viable, neuromodulation becomes the principal alternative. Of the neuromodulation options (DBS, RNS, VNS), implanted vagus nerve stimulation has the longest track record, with FDA approval since 1997.

Transcutaneous vagus nerve stimulation removes the surgical barrier entirely. It is an adjunctive, non-invasive, home-based intervention. The evidence is low certainty and mixed: randomised trials show reductions in seizure frequency, the largest European trial missed its primary endpoint, and pooled responder differences have not reached significance. It has never been compared directly with implanted VNS, and it does not replace referral for surgical assessment (NICE NG217 recommendation 8.2.2); NICE places implanted VNS after that assessment, where resective surgery is not suitable (8.3.1).

The tVNS device referred to throughout this page is the tVNS® E system (tVNS Technologies GmbH, Germany), CE-marked since 2010 and holding Class IIa certification under the EU MDR 2017/745. Anatomical Concepts UK is the distributor in the UK.

---

Mechanism

The mechanistic picture is increasingly well characterised, though not yet complete. Four contributing pathways are worth understanding.

1. The noradrenergic pathway (the best-supported mechanism)

The auricular branch of the vagus nerve (ABVN) provides cutaneous afferent access at the cymba conchae. Vagal afferents project to the nucleus tractus solitarius (NTS) in the medulla, which sends dense projections to the locus coeruleus (LC). LC is the primary source of cortical norepinephrine. Vagal stimulation increases LC firing and elevates extracellular NE in the cortex and hippocampus. LC lesion abolishes the anti-seizure effect in animal models. In humans, LC activation during tVNS has been confirmed through pupil dilation and attenuation of occipital alpha oscillations, both established signatures of LC-NE activity.

Pharmacological studies suggest the antiepileptic effect is mediated specifically by α2-adrenoreceptor activation downstream of NE release.

2. Serotonergic and GABAergic modulation

Vagal afferents also drive activity in the raphe nuclei and elevate cortical and hippocampal GABA levels. GABAergic effects of tVNS have been demonstrated in human EEG studies after single sessions. The combined noradrenergic, serotonergic, and GABAergic modulation plausibly accounts for both the anti-seizure and mood-stabilising effects of vagal stimulation.

3. Anti-inflammatory effects

The cholinergic anti-inflammatory pathway (efferent vagal activation of α7-nicotinic acetylcholine receptors on macrophages, suppressing TNF-α and other pro-inflammatory cytokines) is better established for implanted VNS but is thought to be engaged by tVNS to some degree. Potentially relevant in epilepsies with a neuroinflammatory component.

4. Network-level stabilisation

EEG-derived functional connectivity analyses show that short-term tVNS induces measurable modifications in large-scale epileptic brain networks, enhancing resilience, stability, and robustness (properties inversely correlated with seizure propensity). The modifications differ systematically between focal and generalised epilepsy, suggesting type-specific network engagement.

Key point. The mechanisms of tVNS are multi-layered. The noradrenergic pathway through the locus coeruleus is the best-supported mechanism; serotonergic, GABAergic, anti-inflammatory, and network-level effects all appear to contribute. This mechanistic breadth may also explain the therapeutic relevance to comorbid depression.

Why the cymba conchae specifically

An fMRI study comparing four auricular sites (inner tragus, posterior canal, cymba conchae, and earlobe sham) established that cymba conchae stimulation produces the greatest and most specific activation of the NTS and locus coeruleus. Anatomical studies describe the cymba conchae as supplied by the auricular branch of the vagus; Frangos and colleagues (2015, funded by the original manufacturer, Cerbomed) describe it as "exclusively innervated by the ABVN" (doi:10.1016/j.brs.2014.11.018). Other sites, including the tragus, have mixed innervation. Stimulation at the wrong site does not reliably activate the brainstem pathways that drive the therapeutic effect. This anatomical precision is clinically critical, and is one of the reasons the tVNS® device is designed specifically to sit in the cymba conchae.

---

The evidence

The cMPsE02 trial (Bauer et al., 2016)

A randomised, double-blind controlled study (doi:10.1016/j.brs.2015.11.003) of 76 DRE patients randomised to 25 Hz (n=39) or 1 Hz active control (n=37) at the cymba conchae for 20 weeks, using the Cerbomed NEMOS, the predecessor of the tVNS E. The primary endpoint was not met (23.4% vs −2.9%, p=0.146). The often-quoted 34.2% (p=0.034) is a within-group change from baseline among the 26 of 39 patients in the 25 Hz group who completed treatment, not a between-group effect. Responder rates were similar in both groups, and the authors concluded that superiority of 25 Hz over 1 Hz could not be proven.

Two possible contributors to the negative result: 1 Hz stimulation may itself be biologically active, making the "control" an imperfect sham (the 1 Hz group also received about double the current), and the trial was probably underpowered. An active control makes a between-group difference harder to show, but it cannot turn a non-significant result into a positive one.

Yang et al. 2023 (the largest randomised trial, different device)

A 150 patient multi-centre double-blind RCT (doi:10.1007/s13311-023-01353-9). Adults with DRE were randomised 2:1 to active tVNS (n=100; 25 Hz, 250 µs, 30s on/30s off, 2 hours daily at maximum tolerated intensity, ear conch) or a low-dose active control (n=50; 1 Hz at minimum perceptible intensity, not a sham) for 20 weeks, across four Chinese epilepsy centres. The device was the TVNS-100 (Xinzhile, Jiangxi, China), not the tVNS Technologies device, so this trial is evidence for auricular tVNS in general rather than for the tVNS E.

Results:

• ≥50% responder rate at 20 weeks: 44.74% active vs 16.67% control (p < 0.05), among the 112 of 150 analysed (76 vs 36)
• Mean seizure frequency at 20 weeks: 3.08 vs 6.05, not significantly different between groups (p = 0.130); the relative reduction in active completers (30.75%) was reported as significantly greater than control
• Anti-seizure effect increased with duration, becoming significant only at 20 weeks
• No severe adverse events in either group

Two observations stand out. First, a ≥50% responder rate just under 45% among completers is a meaningful signal for a non-invasive intervention in DRE, though about a quarter of those randomised were not analysed. Second, the benefit is cumulative and emerges clearly only at 20 weeks. This has direct practical implications for setting patient expectations and for trial duration.

The trial did not show significant improvements in QOLIE-31, anxiety, depression, or cognition at 20 weeks, though the duration may have been insufficient for these secondary endpoints.

Beijing Tiantan real-world study (Liao et al., 2025)

An uncontrolled prospective single-centre study of 99 DRE patients undergoing tVNS, with 1 to 2 years of follow-up (doi:10.1177/17562864251396022). The device is not named. 34 of the 99 were lost or refused follow-up. Of the 65 patients with successful follow-up:

• Overall efficacy rate (≥50% seizure reduction): 61.54% of the 65 (about 40% of the 99 enrolled if those not followed up are counted as non-responders)
• 15 patients (23.1%) achieved more than 90% seizure reduction
• 8 patients (12.3%) achieved 75 to 90% reduction
• 17 patients (26.2%) achieved 50 to 75% reduction
• Mild adverse events (ear tingling, tinnitus) in 10 patients; no severe events

Efficacy was independent of measured clinical variables, suggesting broad applicability. The higher efficacy rate compared with the RCTs may reflect the completers-only denominator, the absence of a control group, longer follow-up and possibly greater adherence in a motivated clinical population.

Meta-analyses

Two meta-analyses of the randomised trials were published in 2025 (Moro et al., four RCTs, 368 patients; Makkawi et al., four RCTs, 417 subjects; references in the evidence library). Both found significant reductions in seizure frequency favouring tVNS. Neither found a statistically significant difference in pooled responder rate, to which the active-control problem (1 Hz is not a true sham), heterogeneity and a high risk of bias may all contribute. The Yang 2023 trial, which also used a 1 Hz active control, did show a significant responder rate difference among completers. The pooled results should be read as they stand: low certainty evidence.

Trial in progress: TRAVAST

An independent, triple-blind trial of this device is under way in Germany under the G-BA testing directive (DRKS00039592): 164 adults with DRE, active tVNS against a low-intensity sham, primary endpoint 50% responder rate at 52 weeks, results expected in 2028. The manufacturer supplies the device but does not fund the trial. It is the first trial in this indication with a one-year primary endpoint and a sham designed to sit below biological activity, and it will bear directly on the evidence summarised above. Full details are in the evidence library.

Key point. The evidence base for tVNS in DRE now includes two substantial RCTs, multiple pilot studies, real-world prospective data, and two meta-analyses. The pooled finding is a statistically significant reduction in seizure frequency with a favourable safety profile, but no significant pooled responder difference, and the largest European trial (this device's predecessor) missed its primary endpoint. Responder rates range from just under 45% among completers in the largest RCT active arm (versus about 17% on a low-dose control; a different manufacturer's device) to around 62% among the 65 of 99 followed up in an uncontrolled real-world cohort. The evidence is low certainty. Implanted VNS has the longer and stronger evidence record.

---

Comparison with implanted VNS

No head-to-head RCT exists. With that caveat:

AutoStim mode in implanted VNS (responding to ictal tachycardia, present in approximately 80% of seizures) is a meaningful advantage; one series reports 69.3% ≥50% responder rate with AutoStim active. Closed-loop tVNS prototypes are under development but not yet clinically available.

Key point. tVNS is not a replacement for implanted VNS. It is a non-invasive option for patients who have been assessed for resective surgery and are not candidates or decline it, and who decline or are not suitable for an implant. Because there is no head-to-head trial, the two cannot be ranked on efficacy from this evidence; implanted VNS has the stronger evidence record. Whether response to tVNS predicts response to an implant has not been tested, so a trial of tVNS should not be used as a screen for implantation. The absence of surgical risk, full reversibility, and substantially lower cost are its practical advantages.

---

Practical guidance

Candidate selection

• Adults aged 18 and over with confirmed DRE having failed two or more ASMs (the device's intended population is adults 18+)
• Referred for surgical assessment (NICE NG217 recommends this for everyone with DRE) and not suitable for, or unwilling to undergo, resective surgery
• No active implant (pacemaker, implanted defibrillator, implanted VNS, or cochlear implant) and no cerebral shunt
• Not pregnant; intact skin at the electrode site; cardiac arrhythmia requires medical clearance first
• Ability to maintain a consistent daily stimulation regimen

Dosing parameters

Converging on the following based on the strongest trials:

Left-sided stimulation follows the convention established for implanted VNS, reflecting the asymmetric cardiac innervation of the right and left vagus. The cardiac risk from auricular stimulation is anatomically much lower than for cervical stimulation.

Initiation protocol

1. Baseline seizure diary, a minimum of 8 weeks.
2. Introduce tVNS at the parameters above. Titrate intensity in the first sessions to maximum tolerated sub-pain threshold.
3. Maintain stable ASM regimen during the 20 week assessment window.
4. Brief baseline cardiac monitoring is prudent, particularly in patients with pre-existing cardiac disease.
5. Review at weeks 8, 12, and 20 using the seizure diary.

What to tell the patient

• Anti-seizure benefit typically emerges from around 12 weeks, with the clearest effect at 20 weeks or beyond. This is not a fast intervention.
• In the largest controlled trial (a different manufacturer's device), just under 45% of those who completed it achieved a ≥50% reduction in seizures, versus about 17% on a low-dose control. An uncontrolled real-world cohort reports around 62% among the 65 of 99 followed up (about 40% counting those lost), but without a control group this is not directly comparable, and both 2025 meta-analyses found the pooled responder difference did not reach significance.
• Expect mild local effects (tingling, redness at the electrode site). These are generally well tolerated.
• The device is adjunctive. It does not replace anti-seizure medication.
• Adherence matters. The benefit depends on consistent daily use.

Comorbidity considerations

Depression and anxiety are highly prevalent in DRE. The tVNS device carries an approved indication for depression in addition to epilepsy (anxiety is not among its approved indications), and the shared mechanistic basis (noradrenergic and serotonergic modulation) makes the depression overlap a clinically attractive feature. Cognitive and mood benefits may emerge more slowly than seizure benefits, so patience with the longer endpoints is warranted.

---

Safety

No severe device-related adverse events were reported in the pivotal controlled trials cited here. The Instructions for Use nonetheless list a range of possible effects (including, uncommonly, arrhythmia, dyspnoea, and a preliminary increase in seizures), so counsel patients accordingly.

Common, mild, transient (reported in up to 20% of patients):

• Local erythema, tingling, itching at the electrode site
• Headache, sleep disturbance
• Ear tingling, tinnitus

Uncommon:

• Transient sinus bradycardia (1 case in the Yang 150 patient RCT, self-resolving)
• Nausea, dizziness, hoarseness (much less frequent than implanted VNS)

By comparison, implanted VNS carries surgical risks including infection (3 to 7%), vocal cord paresis, bradycardia during implantation, and device fracture. Cumulative stimulation side effects (hoarseness, dyspnoea) affect up to 40% of implanted VNS patients at therapeutic intensities. The avoidance of these is a significant practical advantage.

---

Regulatory and access context

• tVNS® E device: CE marked since 2010. Class IIa under EU MDR 2017/745, verified by TÜV SÜD.
• EU MDR approved indications are epilepsy, depression, chronic migraine, and Prader-Willi syndrome (per the Instructions for Use).
• UK status: the device can be used in UK clinical practice under its current regulatory status. NICE has not issued guidance on tVNS in epilepsy; the NG217 evidence review noted that "tVNS is not used within the NHS". NG217 recommends referral for surgical assessment for everyone with DRE, and implanted VNS only where resective surgery is not suitable.
• Distribution: Anatomical Concepts UK holds UK distribution. Warranty, returns, and support are handled in the UK.

---

What we do not yet know

Honest acknowledgement of the gaps matters for patient discussions.

• Long-term outcomes beyond 2 years: limited controlled data.
• Head-to-head comparison with implanted VNS: no RCT exists.
• Closed-loop tVNS: prototypes under development, not yet clinically available.
• Optimal parameter space: 25 Hz / cymba conchae / 2 to 4 h daily is empirically supported but not formally dose-optimised.
• Status epilepticus: implanted VNS has shown benefit in NORSE case series; whether tVNS has a role is unanswered.

---

Summary

In summary: transcutaneous vagus nerve stimulation is a CE marked, non-invasive, adjunctive option for drug-resistant epilepsy with a plausible neurophysiological rationale and low certainty, mixed clinical evidence. Meta-analyses of the randomised trials show significant reductions in seizure frequency but no significant pooled responder difference, and the largest European trial, with this device's predecessor, missed its primary endpoint. In the largest controlled trial (a different manufacturer's device), just under 45% of completers achieved a ≥50% reduction versus about 17% on a low-dose control; an uncontrolled real-world cohort reports around 62% of the 65 of 99 followed up. It has not been compared directly with implanted VNS. Its safety profile is favourable and it carries no surgical risk.

For patients who have had a surgical assessment and decline surgery or are not surgical candidates, including those who have not yet tried neuromodulation, tVNS is a reasonable option to discuss. Its accessibility, reversibility, substantially lower cost, and dual indication for comorbid mood disorders make it especially appealing in practice.

---

Next steps

• Book a clinician demo or call with Anatomical Concepts UK
• Read the practice guide
• See the full evidence library with references
• How to access the tVNS device

---

Key references are available in the Evidence Library. This page summarises clinical evidence current as of April 2026.