Evidence library
This page is intended for UK healthcare professionals. It summarises the trial and real-world evidence for transcutaneous vagus nerve stimulation across its four EU MDR approved indications, with attention to study quality, blinding, and the limits of the active-control sham conditions used in much of this literature.
How to read this page
Three caveats apply to almost every randomised controlled trial of tVNS, and they are worth stating once at the top.
The active-control sham problem. Most tVNS RCTs use either a low-frequency (1 Hz) cymba conchae stimulation as the active control, or stimulation of the earlobe (which has essentially no vagal innervation). Neither is a clean sham. 1 Hz stimulation may itself be biologically active in ways that suppress between-group separation. Earlobe sham reduces blinding fidelity, because patients can sometimes distinguish active from sham by the location of the stimulation. The implication is that pooled effect estimates may underrepresent the true effect.
Heterogeneous parameters. Trials have used a range of frequencies (1 Hz to 25 Hz), pulse widths (200 to 500 µs), session durations (30 minutes to 4 hours daily), and trial durations (4 weeks to 20 weeks). Heterogeneity attenuates pooled estimates and complicates dose-response analysis.
Outcome variability. Indication-specific endpoints are not always uniform across trials. Responder definitions in epilepsy use ≥50% seizure reduction, but the time horizon varies. Migraine endpoints differ between episodic and chronic populations. Depression endpoints sometimes use HAMD-17, sometimes MADRS, sometimes PHQ-9. Pooling across instruments requires care.
These are not reasons to dismiss the literature. They are reasons to read it carefully.
Drug-resistant epilepsy
The strongest indication. Two substantial randomised controlled trials, multiple pilot studies, real-world prospective cohort data over 1 to 2 years, and two meta-analyses published in 2023 and 2026.
Bauer et al., 2016 — cMPsE02
A randomised, double-blind, parallel-group trial of 76 adults with drug-resistant focal epilepsy. Patients were randomised 1:1 to active stimulation (25 Hz, n=39) or low-frequency active control (1 Hz, n=37) at the cymba conchae for 4 hours daily over 20 weeks.
Primary endpoint. Reduction in seizure frequency at 20 weeks. ITT analysis: 23.4% reduction in active vs −2.9% in control, p=0.146 (not statistically significant). Per-protocol analysis (n=26 completers in the 25 Hz group): 34.2% reduction, p=0.034.
Interpretation. The first randomised double-blind evidence that high-frequency tVNS produces a measurable anti-seizure effect in DRE. The non-significant ITT result is most plausibly explained by two factors: 1 Hz stimulation acting as a partial active rather than a true sham, and the trial being underpowered. Both factors were addressed in subsequent trials.
Yang et al., 2023
The strongest single piece of evidence in the field. A 150-patient multi-centre, double-blind RCT across four Chinese epilepsy centres, randomising adults with DRE 2:1 to active tVNS (n=100; 25 Hz, 250 µs, 30s on/30s off, 2 hours daily, maximum tolerated intensity, cymba conchae) or sham (n=50; 1 Hz at minimum perceptible intensity) for 20 weeks.
Primary endpoint. ≥50% responder rate at 20 weeks: 44.74% active vs 16.67% control, p<0.05. Statistically significant on the strongest endpoint in the literature.
Secondary endpoints. Mean seizure reduction at 20 weeks: 30.75% active vs 15.66% control, p=0.038. The anti-seizure effect was cumulative, becoming statistically significant only at 20 weeks; earlier time points showed numerical separation but not significance.
Safety. No severe adverse events in either group. The active group reported one episode of self-resolving sinus bradycardia.
Quality. Pre-registered. Independent monitoring. Standardised seizure diary. The strongest sham design in the published literature, although the active-control problem still applies.
Interpretation. A clinically meaningful responder rate (≥45%) for a non-invasive adjunctive intervention in a population by definition refractory to first-line care. The cumulative time course has direct practical implications: 20 weeks is the minimum useful trial duration in this indication.
Beijing Tiantan Hospital real-world study, 2024
A single-centre prospective real-world study of 99 adult DRE patients receiving tVNS, with 1 to 2 years of follow-up. 65 patients had successful follow-up.
Overall efficacy rate (≥50% seizure reduction): 61.54%.
- 23.1% achieved >90% seizure reduction
- 12.3% achieved 75 to 90% reduction
- 26.2% achieved 50 to 75% reduction
Adverse events. Mild, transient, in 10 patients (15%); ear tingling, transient tinnitus. No severe events.
Interpretation. The higher efficacy rate compared with the RCTs likely reflects longer follow-up, motivated patient population, and possibly higher real-world adherence. Efficacy was independent of measured clinical variables, suggesting broad applicability across DRE phenotypes. As an uncontrolled study, it cannot speak to placebo effect; but the durability of response over 12 to 24 months is informative.
Meta-analyses
Two pooled analyses of RCT data have been published.
Wang et al., 2023. Pooled analysis of available RCT data. Significant reduction in seizure frequency favouring active tVNS over control. Pooled responder rate did not reach statistical significance, an effect partly explained by the active-control problem (1 Hz is not a true sham).
He et al., 2026. Updated pooled analysis incorporating Yang 2023 and subsequent trials. Significant reduction in seizure frequency (p=0.008). Heterogeneity remained substantial across studies, attributable to varying parameters, sham conditions, and trial durations. The Yang 2023 trial, which used the strongest sham design and the longest trial duration, reported the largest responder-rate effect, suggesting pooled estimates may underrepresent the true effect when better-designed studies are weighted appropriately.
Key point. The epilepsy evidence base now includes two substantial RCTs, multiple pilot studies, real-world prospective data, and two meta-analyses. Responder rates range from just under 45% in the controlled RCT active arm (versus about 17% on control) to around 62% in uncontrolled real-world cohorts, though both 2025 meta-analyses found the pooled responder difference did not reach significance. The safety profile is favourable. The evidence is not yet at the level of implanted VNS in epilepsy, but it is substantial and increasingly mature.
Read the clinician page on tVNS for epilepsy
Depression
The depression evidence base is younger than the epilepsy literature and substantially more variable. The mechanistic rationale is shared (noradrenergic and serotonergic modulation through brainstem relays), but the controlled trial base is smaller.
Implanted VNS for context
Implanted vagus nerve stimulation has FDA approval for treatment-resistant depression (TRD) since 2005. Long-term cohort studies report response rates around 67% at 24 months in TRD populations, materially higher than continued pharmacotherapy alone in the same population. The implanted VNS evidence is the strongest single piece of evidence that vagal modulation has a clinical antidepressant effect; the question for tVNS is whether transcutaneous delivery engages enough of the same circuitry to produce a comparable, even if smaller, clinical effect.
Hein et al., 2013
An early sham-controlled randomised trial of tVNS in 37 patients with major depression. Active stimulation produced a statistically significant reduction in BDI scores compared with sham. Small sample, short duration (2 weeks), but the first controlled signal that auricular tVNS has an antidepressant effect.
Rong et al., 2016
A 12-week nonrandomized, single-blind, sham-controlled study of 160 patients with mild to moderate major depression: a first cohort of 91 received active taVNS only, and a second cohort of 69 received four weeks of sham before crossing over to active. Active taVNS produced significantly greater reductions in HAM-D-24 scores than sham; at four weeks there were 24 responders (a 50% or greater reduction) in the active group versus none in the sham group (p < 0.00001), with improvement continuing through 12 weeks.
Liu et al., 2020 and follow-up reports
A series of trials and open-label extensions in patients with TRD and partially responsive depression. Aggregate response rates in the 40 to 50% range with treatment durations of 8 to 12 weeks. Effect sizes consistent with adjunctive antidepressant treatment, smaller than implanted VNS but with no surgical risk.
What the depression literature does not yet show
- Head-to-head comparison with implanted VNS in TRD.
- Trials with the rigorous sham fidelity of the Yang 2023 epilepsy trial.
- Adequately powered trials specifically in TRD as defined by failure of two or more antidepressant trials at adequate dose and duration.
- Long-term outcome data beyond 6 to 12 months.
Key point. The depression evidence base supports tVNS as an adjunctive intervention with a credible antidepressant signal across multiple sham-controlled trials. It does not yet support tVNS as a first-line intervention or as a replacement for pharmacological or psychological care. The mechanistic rationale is strong; the trial evidence is moderate and improving.
Read the clinician page on tVNS for depression
Migraine and headache disorders
The migraine evidence base sits across two related modalities: cervical transcutaneous vagus nerve stimulation (tcVNS, the gammaCore device) and auricular transcutaneous vagus nerve stimulation (taVNS, the tVNS® device). These engage different branches of the vagus and use different stimulation parameters. The tcVNS literature is larger; the taVNS literature is smaller but consistent in direction.
Cervical tVNS
Goadsby et al., 2014 (acute migraine). An open-label, single-arm, multi-centre pilot study using the cervical gammaCore device: of 30 patients enrolled, 27 treated 80 attacks. The pain-free rate at two hours was about 21 to 22% (12 of 54 moderate or severe attacks; 4 of 19 for the first treated attack).
EVENT trial (Silberstein et al., 2016) (preventative chronic migraine). A randomised sham-controlled trial of 59 patients. Active cervical tVNS produced a non-significant reduction in headache days at 8 weeks but a significant reduction at the open-label extension at 6 months.
ACT1 and ACT2 trials (cluster headache). Two pivotal sham-controlled trials of cervical nVNS (gammaCore) for the acute treatment of cluster headache attacks. Prophylaxis was tested separately in the PREVA trial (Gaul et al., 2016): adding cervical nVNS to standard care reduced attacks by about 3.9 per week more than standard care alone (a between-group difference; 5.9 active vs 2.1 control; 95% CI 0.5-7.2, p=0.02). These trials used the cervical (neck) device, not the auricular tVNS device distributed here.
These results led to FDA clearance for episodic cluster headache (2017) and chronic migraine (2018) for the cervical device.
Auricular tVNS
Straube et al., 2015. A sham-controlled randomised trial of auricular tVNS in 46 patients with chronic migraine. 12 weeks of treatment. The 1 Hz auricular stimulation arm produced significantly greater reductions in headache days than the 25 Hz arm, an unexpected result that highlighted the dose-response complexity in this indication.
Kraus et al., 2017. An open-label study of auricular tVNS in chronic migraine, with significant reductions in headache days and analgesic use over 6 to 12 weeks.
Smaller open-label and pilot studies. Several studies of auricular tVNS in migraine and chronic tension-type headache have reported reductions in headache days, attack frequency, and analgesic consumption. Effect sizes vary; methodological quality is moderate at best.
What the migraine literature shows
Auricular tVNS has a smaller and less rigorous evidence base than cervical tVNS for headache disorders. The mechanistic case is consistent: vagal afferents converge with trigeminal afferents in the spinal trigeminal nucleus, and vagal stimulation modulates descending pain inhibition pathways. The clinical signal is real but more variable than in epilepsy.
For chronic migraine, taVNS is a reasonable adjunctive option in patients who have failed at least one preventative agent and who prefer a non-invasive approach. The expected effect size is moderate. The evidence is weaker than in epilepsy and the time course may be shorter (12 weeks rather than 20 in some protocols).
Key point. The migraine evidence base is mixed but moving in a constructive direction. Cervical tVNS has the stronger data, particularly in cluster headache. Auricular tVNS has smaller trials with consistent direction. The mechanism is plausible across both modalities. We do not oversell.
Read the clinician page on tVNS for migraine
Prader-Willi syndrome
The smallest published evidence base of the four indications. EU MDR approval is specifically for temper outbursts, which are among the most distressing behavioural features of the syndrome and a major source of carer and family burden.
Manning et al., 2019 (a foundational case series)
A non-blind single-case study (modified ABA design) of 5 adults with PWS, aged 22 to 41, using tVNS. Participants received four hours of stimulation daily for 12 months. Four of the five who completed the study showed a statistically significant reduction in the number and severity of temper outbursts after approximately nine months of daily use. The study was non-blind, uncontrolled beyond within-participant baselines, and small, but it is the principal published study behind clinical interest in this indication.
Subsequent case reports and small series
Several case reports and small case series have been published describing reductions in temper outburst frequency, improvements in carer burden, and qualitative improvements in mood and engagement. None reaches the methodological standard of an RCT.
Why the evidence base is small
PWS is a rare condition (UK prevalence approximately 1 in 25,000). Recruiting an adequately powered RCT is genuinely difficult. Beyond rarity, the heterogeneity of the syndrome's behavioural phenotype and the practical difficulty of objective endpoint measurement in this population have limited the size and quality of the published trials.
What we tell families and clinicians
The honest framing is that the published evidence is limited but not absent, the mechanistic rationale (modulation of arousal and emotion-regulation circuits) is plausible, the safety record across all tVNS indications is favourable, and the alternative pharmacological approaches for outbursts in PWS have their own limitations. tVNS is a regulated option in this population. We do not present it as a guaranteed solution, and we are particularly attentive to expectation-setting in families who have already lived with multiple disappointing interventions.
Key point. The PWS evidence base is the smallest of the four indications. The regulatory status is real, the mechanistic case is plausible, the safety record is good. The clinical case rests more on the absence of better alternatives in some patients and on the willingness of families to try a non-pharmacological adjunctive intervention than on a robust trial literature.
Read the clinician page on tVNS for Prader-Willi syndrome
Off-label and emerging applications
A non-exhaustive summary of the literature in conditions outside the four approved indications, included here for completeness. We do not market the device for any of these.
Stroke rehabilitation
A meta-analysis of 18 randomised trials involving 954 patients (multiple Chinese centres, several Western trials) found meaningful improvements in upper-limb motor function and in swallowing function after stroke. The evidence quality was rated moderate to high for some outcomes. The mechanism is plausible: vagal afferent activation produces cortical state changes that may enhance motor learning during paired rehabilitation. The TRICEPS trial, due to report in 2026, is expected to add a substantial randomised dataset.
Functional gastrointestinal disorders
Systematic reviews of functional dyspepsia and irritable bowel syndrome report improvements in symptoms, quality of life, and mood after taVNS protocols of 4 to 8 weeks. The mechanism is consistent with the vagus nerve's role in the brain-gut axis and in inflammation modulation.
Spinal cord injury
Small pilot studies have explored taVNS in autonomic dysreflexia, blood pressure regulation, and inflammation modulation in SCI. The literature is early-stage but mechanistically motivated.
Chronic pain
The convergence of vagal afferents with descending pain-inhibitory pathways provides a rationale for taVNS in fibromyalgia, chronic primary pain, and post-surgical pain. Published trials are small and heterogeneous. We are unwilling to overclaim on the basis of the current literature.
Cardiac rehabilitation, Parkinson's disease, multiple sclerosis, traumatic brain injury
All have small, mechanistically motivated published literatures. None has yet reached the maturity required for regulatory approval as a tVNS indication. We are happy to discuss the available studies on request.
Two specific reading recommendations
For clinicians who want to read the original literature rather than summaries:
- Yang et al., 2023. Transcutaneous Auricular Vagus Nerve Stimulation (ta-VNS) for Treatment of Drug-Resistant Epilepsy: A Randomized, Double-Blind Clinical Trial. The strongest single trial in the field.
- Kraus, Hösl, et al., 2007. BOLD fMRI Deactivation of Limbic and Temporal Brain Structures and Mood Enhancing Effect by Transcutaneous Vagus Nerve Stimulation. The foundational human imaging paper.
Both are open-access and worth reading in original form.
What the literature does not yet say
For honest framing, the gaps that matter most:
- Long-term outcomes beyond 2 years in any indication.
- Head-to-head comparison with implanted VNS in epilepsy or depression.
- Adequately powered RCTs in TRD-defined depression.
- Closed-loop tVNS is research-grade only.
- Optimal parameter space is empirically supported but not formally dose-optimised.
- Predictive biomarkers for response at the individual patient level are not yet established.
We mention these because the absence of evidence is not the same as the evidence of absence, and because clinicians making real decisions in real clinics deserve to know where the certainty ends.
Where to next
Book a demo or call to discuss the evidence
tVNS is a Class IIa medical device manufactured by tVNS Technologies GmbH, Germany. Distributed in the UK by Anatomical Concepts UK Ltd. Its approved indications are epilepsy, depression, migraine, and Prader-Willi syndrome.