Three things share the name "vagus nerve stimulation," and they are not the same product
"Vagus nerve stimulation" covers a surgical implant, a handheld or wearable neck device, and an ear clip. Reading a single study or a single ad as if it covers all three is where most of the confusion starts.
Implanted VNS means a pulse generator in the chest wired to a stimulating lead coil on the left cervical vagus nerve, placed surgically, which is how the epilepsy trials on this page describe it. Non-invasive cervical tVNS holds an electrode against the skin on the side of the neck, with no incision; that is how the FDA describes gammaCore. Auricular tVNS puts an electrode on the outer ear to reach the auricular branch of the vagus nerve: at the cymba conchae in the acute HRV trial below, at the tragus in the jaw-pain trial, at the ear conch in the epilepsy trial.
| Type | Site & invasiveness | US regulatory status | Typical use |
|---|---|---|---|
| Implanted VNS | Surgical lead coil on the left cervical vagus nerve, generator in the chest | FDA-approved prescription device (PMA P970003), for named conditions | Partial-onset seizures refractory to medication; treatment-resistant depression |
| Cervical tVNS | Handheld or wearable electrode on the side of the neck, no incision | Split: gammaCore is FDA-cleared for headache; consumer neck wearables are sold as general wellness products | Cluster headache and migraine pain (prescription); relaxation, stress, sleep wind-down (consumer) |
| Auricular tVNS | Electrode at the cymba conchae, tragus or ear conch, no incision | Research devices in the trials on this page; consumer ear clips are sold as general wellness products | Studied for heart-rate variability, inflammation, pain and seizures; sold for stress and sleep |
Regulatory framing per FDA PMA P970003 (approved July 16, 1997; depression supplement S050 approved July 15, 2005) and FDA De Novo DEN150048 plus 510(k) K173442 and K182369 for gammaCore, read September 7, 2026. Device-site detail per Morris GL, Mueller WM, Neurology, 1999, PMID 10563620; Atanackov P, et al., Biomedicines, 2025, PMID 40149675; Prott LS, et al., J Oral Rehabil, 2025, PMID 40814998; Bauer S, et al., Brain Stimul, 2016, PMID 27033012. "General wellness product" wording per the vendor's own page, read September 2, 2026.
What implanted VNS is actually approved for, and the responder rates behind that approval
This is the medical-grade end: surgery, a prescription, and years of trial data behind each approved use. None of it describes a wearable.
Notice what gammaCore is: non-invasive, but still a prescription device that went through FDA review and published a randomised trial. That is a bar the consumer device we reviewed has not cleared, for any condition, as of this reading.
Epilepsy approval date per FDA PMA P970003 (decision July 16, 1997); indication wording per supplement S207 (June 23, 2017). Depression approval and indication per PMA supplement S050 (decision July 15, 2005). Pooled responder data per Morris GL, Mueller WM, Neurology, 1999;53(8):1731–1735, PMID 10563620. Twelve-month depression outcomes per Rush AJ, et al., Biol Psychiatry, 2005;58(5):355–363, PMID 16139581. gammaCore decisions per FDA De Novo DEN150048 (April 14, 2017), 510(k) K173442 (January 23, 2018) and 510(k) K182369 (November 27, 2018); PREVA response rates per Gaul C, et al., Cephalalgia, 2016;36(6):534–546, PMID 26391457. All read September 7, 2026.
What tVNS studies actually measured: HRV, inflammation, pain, and one pooled null result
Consumer tVNS marketing leans hard on heart-rate variability (HRV) as the proof point. The research on that exact claim is more mixed than the marketing.
A living Bayesian random-effects meta-analysis of 16 single-blind studies comparing ear-clip taVNS against sham in healthy participants found strong evidence for the null hypothesis on vagally mediated HRV (g = 0.014, BF01 = 24.678), concluding that acute taVNS does not alter vagally mediated HRV compared to sham and that there is no support for vagally mediated HRV as a robust biomarker of acute taVNS (Wolf V, Kühnel A, Teckentrup V, Koenig J, Kroemer NB, Psychophysiology, 2021;58(11):e13933, PMID 34473846).
In 78 healthy adults tested across seven randomised sessions (six active taVNS protocols plus an inactive sham at the cymba conchae of the left ear), three settings, 10 Hz at 250 µs, 10 Hz at 500 µs, and 25 Hz at 100 µs, significantly increased one HRV metric (SDNN) versus sham during a single 15-minute stimulation, with no significant change in RMSSD, the vagally mediated metric, for any protocol (Atanackov P, Peterlin J, Derlink M, Kovačič U, Kejžar N, Bajrović FF, Biomedicines, 2025;13(3):700, PMID 40149675).
A randomised, blinded pilot in 21 patients hospitalised with COVID-19 measured IL-6, IL-10, cortisol and C-reactive protein across 14 sessions of active versus sham taVNS; between the groups, only CRP was significantly lower with active stimulation (p = 0.04), and overall clinical outcomes did not improve (Uehara L, Corrêa JCF, Ritti R, et al., Expert Rev Med Devices, 2022;19(11):915–920, PMID 36540947). A pilot in 20 adults with chronic jaw (TMD) pain, 10 on active taVNS and 10 on an inactive sham electrode, assessed pain and related measures at 4 and 8 weeks; none of the differences were statistically significant (Prott LS, Kaldenhoven V, Hugger A, et al., J Oral Rehabil, 2025;52(12):2351–2360, PMID 40814998). Both are pilot-scale, not confirmatory.
Put together: single small trials, in the 20-to-80-person range, sometimes find a positive direction on one measure, in one sitting, and no significant difference on the rest. The living meta-analysis of the field's most-studied biomarker, vagally mediated HRV, came back with strong evidence for the null.
HRV meta-analysis per Wolf V, et al., Psychophysiology, 2021;58(11):e13933, PMID 34473846. Acute frequency/pulse-width trial per Atanackov P, et al., Biomedicines, 2025;13(3):700, PMID 40149675 (abstract and PMC full text). COVID-19 inflammation pilot per Uehara L, et al., Expert Rev Med Devices, 2022;19(11):915–920, PMID 36540947. Jaw-pain pilot per Prott LS, et al., J Oral Rehabil, 2025;52(12):2351–2360, PMID 40814998. All read September 7, 2026.
How a tVNS session actually runs: frequency, pulse width, and how many weeks before anyone measured anything
The dial settings behind "vagus nerve stimulation" are narrower and more specific than a wellness label would suggest.
25 Hz, pulse width 250 µs, 28 s on / 32 s off, 4 hours daily at the left tragus
Active-arm protocol of the jaw-pain pilot, Prott LS, et al., J Oral Rehabil, 2025, PMID 40814998The active protocols in the trials on this page run at 10 Hz or 25 Hz with pulse widths of 100, 250 or 500 microseconds (the acute crossover trial tested all six combinations), 25 Hz at 250 microseconds in the jaw-pain pilot, and 25 Hz in the epilepsy trial, whose control arm was stimulated at 1 Hz instead of receiving nothing.
Session length is where the studies split hard. The acute HRV crossover trial measured its effect inside a single 15-minute stimulation phase, with a 15-minute baseline before and 10 minutes of recovery after, same day. The taVNS epilepsy trial (76 patients: 39 at 25 Hz, 37 at 1 Hz), aiming at a therapeutic outcome rather than a same-day HRV blip, judged seizure frequency over 20 weeks: mean seizure reduction per 28 days at end of treatment was 23.4% in the 25 Hz group versus -2.9% at 1 Hz, a difference that did not reach significance (p = 0.146), so superiority could not be proven (Bauer S, et al., cMPsE02, Brain Stimul, 2016;9(3):356–363, PMID 27033012). A wellness claim measured in one sitting and a clinical claim measured after 20 weeks of daily use are not the same kind of evidence, even when both use the phrase "tVNS."
Parameter values per Atanackov P, et al., Biomedicines, 2025;13(3):700, PMID 40149675; Prott LS, et al., J Oral Rehabil, 2025;52(12):2351–2360, PMID 40814998; Bauer S, et al., Brain Stimul, 2016;9(3):356–363, PMID 27033012. Acute-trial session structure per the Atanackov PMC full text. Read September 7, 2026.
The sham problem: why a tingle is hard to fake, and the one question to ask of any device study
tVNS has a specific blinding problem that a sugar-pill trial does not: the "treatment" itself is a sensation, and a fake version has to be convincing or the trial is not really blinded.
The trials on this page handled it in different ways. The acute HRV trial used an inactive sham: the same electrode at the cymba conchae, the same calibration routine, no current delivered, with participants told beforehand that any tingling "could vary or subside throughout the session." Its authors contrast that with other groups that used active shams at the helix or earlobe, noting that although the earlobe lacks vagal innervation, stimulation there may still influence autonomic regulation (Atanackov P, et al., Biomedicines, 2025, PMID 40149675). The jaw-pain pilot used an inactive, non-functional sham electrode (Prott LS, et al., 2025, PMID 40814998). The epilepsy trial went double-blind with an active control stimulated at 1 Hz rather than 25 Hz (Bauer S, et al., 2016, PMID 27033012). The depression cohort began with 10 weeks of active or sham VNS before its open naturalistic year (Rush AJ, et al., 2005, PMID 16139581). PREVA, the gammaCore cluster-headache trial, was open-label with no sham arm at all (Gaul C, et al., 2016, PMID 26391457).
An open-label study tells participants they are receiving real stimulation, with no attempt at a matched fake condition. That is why the meta-analysis in the section above, which pooled only studies comparing taVNS against sham, is a more trustworthy number than any single open-label report.
Before trusting any tVNS result, find the control group
- Sham-controlled (an inactive electrode, or an active control at a different setting) with participants blinded: the stronger design.
- Open-label, no control group, or self-reported percentages from a company survey: read the number as a testimonial, not a trial result.
- A single small positive study (n=20 to n=80) is a lead, not a finding, until it survives being pooled with the rest of the field, the way the vagally mediated HRV claim did not.
Where the evidence stops for the wearable in the ad: no device-specific trial, a survey instead of a study, and what a certification mark does not mean
None of this page's trial data was run on a consumer product. That gap is worth naming plainly before any purchase.
What "FCC and CE certified" tells you
- The device's radio and electrical emissions have been assessed against interference and safety limits.
- The maker declares conformity with the applicable product-safety rules, in the US and EU respectively.
- It can legally be sold as an electronic product in those markets.
What it does not tell you
- That the device did anything measurable in a controlled trial.
- That any study, sham-controlled or otherwise, tested this exact model.
- Anything about effect size, responder rate, or how long a benefit lasts.
Innerbody, which reviewed the Pulsetto neck device we also cover, put it directly: "the only studies that support Pulsetto's exact operating parameters are those whose funding came from Pulsetto itself. What's more, none of them appear to be published in peer-reviewed journals, and two of them are explicitly master's theses." The supporting research is about tVNS as a technique, run in university labs on other equipment.
The same vendor's page reports what "1000+ users saw" after 14 days of daily use. A self-reported survey of paying customers, with no control group, is a different kind of number than a sham-controlled trial, even printed in the same font on the same page. One tells you what buyers who already wanted it to work said afterward. The other tells you what happened when the participant did not know which condition they were in.
Innerbody quotation from innerbody.com/pulsetto-reviews, read September 7, 2026. "Fully certified by the FCC and CE," "general wellness product" and the 14-day survey line read on the vendor's own pages, September 2, 2026.
How to run your own 4-to-5-week test, since no one else's trial covers your exact device
If the published research will not tell you what a specific wearable does for you, a short, disciplined self-test at least tells you something real about your own nervous system.
Baseline week
Before turning the device on, track one number for 7 days: a plain 1-to-10 evening-stress rating in a note, or resting HRV from a wearable you have already worn long enough to have a settled baseline. Garmin's own documentation puts that at around three weeks of overnight wear, the window our what is a good HRV page sets out. Seven days of logging is your before-window, not the moment you start owning the wearable.
One outcome, one timing
Pick a single metric to track and a fixed session time, using your device's own recommended session length. For scale: the acute trial above stimulated for 15 minutes per session, and the device we reviewed advertises 4-minute sessions. Do not change either mid-test; that is what erases the comparison.
Run it daily, 3 to 4 weeks
Keep the session length constant and log the one metric daily, same time, same method, so week 4 is comparable to week 0. For scale, the tVNS trials on this page measured at one sitting, at 4 and 8 weeks, and at 20 weeks.
Stop rules
Stop and see a doctor for skin irritation that does not resolve, lightheadedness, or palpitations. Stop using it, unhurried, if 3 to 4 weeks of consistent logging show no change in your one tracked number.
Session-length figures per Atanackov P, et al., 2025, PMID 40149675 (15-minute stimulation phase) and the vendor's own "4 minutes" claim, read September 2, 2026; 4-week timepoint per Prott LS, et al., 2025, PMID 40814998. Baseline-window figure per Garmin's own HRV Status documentation ("you need to routinely wear your Garmin device overnight for around three weeks before HRV status becomes fully active"), read Sep 7, 2026. The test design and stop rules are our own general guidance, not a trial protocol and not a diagnosis.
Questions people also ask
Is vagus nerve stimulation FDA approved?
Implanted vagus nerve stimulation is FDA-approved for specific conditions: partial-onset seizures refractory to medication (PMA P970003, July 16, 1997) and chronic or recurrent treatment-resistant depression in adults (supplement S050, July 15, 2005). One non-invasive neck device, gammaCore, is separately FDA-cleared for acute pain of episodic cluster headache (De Novo, April 2017), acute migraine pain (January 2018) and adjunctive prevention of cluster headache (November 2018), all in adults. Consumer neck and ear-clip tVNS wearables sold for relaxation or stress are marketed as general wellness products that are "not intended to diagnose, treat, cure, or prevent" any disease, in the vendor's own words, not under an FDA clearance.
How long does it take for vagus nerve stimulation to work?
It depends which kind. Implanted VNS for epilepsy is judged over years: one open-label study of 454 implanted patients found a 50%-or-greater seizure reduction in 36.8% at one year, 43.2% at two years and 42.7% at three years (Morris GL, Mueller WM, Neurology, 1999;53(8):1731-1735, PMID 10563620). Acute tVNS lab studies measure heart-rate-variability changes inside a single 15-minute session (Atanackov P, et al., Biomedicines, 2025, PMID 40149675). The multi-week tVNS trials on this page assessed outcomes at 4 and 8 weeks (jaw pain) and 20 weeks (epilepsy). No trial on this page tested a consumer wearable for relaxation, so there is no published timeline for that use.
What are the side effects of vagus nerve stimulation?
Implanted VNS carries surgical risks, and in the 454-patient epilepsy cohort the most common adverse events at one year were hoarseness (28%) and paraesthesias (12%) (Morris GL, Mueller WM, 1999, PMID 10563620); in the depression cohort, voice alteration, dyspnea and neck pain were reported most often (Rush AJ, et al., 2005, PMID 16139581). In the 20-week ear tVNS epilepsy trial, adverse events were usually mild or moderate and comprised headache, ear pain, application-site redness, vertigo, fatigue and nausea (Bauer S, et al., 2016, PMID 27033012); the acute trial in healthy adults recorded one report of pain and one of itching at the stimulation site (Atanackov P, et al., 2025, PMID 40149675). Anyone with a pacemaker or other implanted device, and anyone pregnant, should ask a doctor before trying either kind.
Is tVNS the same as a TENS unit?
No. A TENS unit is sold for pain relief at a muscle or joint. tVNS targets one specific nerve branch, at the ear or neck, with parameters chosen to engage vagal afferents: the trials on this page used 10 Hz or 25 Hz with pulse widths of 100 to 500 microseconds, and the jaw-pain pilot ran 25 Hz at 250 microseconds. The two use similar hardware but are designed around different nerves and different outcomes.
How we built this page
This is a literature-and-record page, not a lab test, and no product was tested to write it. Every number on the page was checked against the abstract, the PMC full text where one exists, or the FDA record it is attributed to, on September 7, 2026.
- Implanted VNS approval dates and indication wording come from FDA PMA P970003 (July 16, 1997), supplement S207 (June 23, 2017) and supplement S050 (July 15, 2005). The epilepsy responder rates (n=454, open-label) are from Morris GL, Mueller WM, Neurology, 1999;53(8):1731–1735, PMID 10563620. The 12-month depression outcomes (n=205, naturalistic) are from Rush AJ, et al., Biol Psychiatry, 2005;58(5):355–363, PMID 16139581.
- gammaCore's decisions come from FDA De Novo DEN150048 (April 14, 2017) and 510(k)s K173442 (January 23, 2018) and K182369 (November 27, 2018). The PREVA trial's response rates are from Gaul C, et al., Cephalalgia, 2016;36(6):534–546, PMID 26391457.
- The vagally mediated HRV null result is from Wolf V, et al., Psychophysiology, 2021;58(11):e13933, PMID 34473846. The acute frequency/pulse-width trial (n=78) is Atanackov P, et al., Biomedicines, 2025;13(3):700, PMID 40149675, read in full text. The COVID-19 inflammation pilot (n=21) is Uehara L, et al., Expert Rev Med Devices, 2022;19(11):915–920, PMID 36540947. The jaw-pain pilot (n=20) is Prott LS, et al., J Oral Rehabil, 2025;52(12):2351–2360, PMID 40814998. The 20-week taVNS epilepsy trial (n=76) is Bauer S, et al., Brain Stimul, 2016;9(3):356–363, PMID 27033012.
- The sham descriptions come from those same papers, not from a separate methods review. Stimulation parameters quoted are the ones those trials report; we did not print a field-wide "default."
- The Innerbody quotation is from innerbody.com/pulsetto-reviews. The vendor's "FCC and CE," "general wellness product," "4 minutes" and 14-day survey wording was read on its own pages on September 2, 2026.
- The three-week wearable-baseline window in the self-test is not a study finding: it comes from Garmin's own HRV Status documentation, read September 7, 2026, and is printed so the 7-day before-window here is not mistaken for the time it takes a wearable to learn your baseline.
- We did not test any consumer device, we have not worn the device our review covers, and we did not extrapolate a device-specific result from a study that tested tVNS in general.
- Nothing here is medical advice. Starting any nerve-stimulation device while pregnant, with a pacemaker or other implant, or on medication is a conversation with your doctor first.
Where we asked this of one specific device
Everything above is the field. The harder, narrower question, does the research on this page actually cover the wearable in front of you, is one we answered for exactly one device.
For the consumer end of that evidence, we wrote a dedicated Pulsetto Fit review at /pulsetto-fit-review/, with the year-one cost added up and a straight answer on whether any study tested the device itself.
Before buying any tVNS wearable, ask the seller directly whether an independent, peer-reviewed study has tested that exact model, not tVNS in general. If nobody can name one, that silence is itself the answer.
Separate a company's own multi-week survey from a sham-controlled trial before you weigh either one. A self-reported percentage from people who already wanted the device to work is a different kind of evidence than a result from participants who did not know which condition they were in.
Affiliate disclosure: the review linked above contains affiliate links, and a purchase made through them may earn us a commission at no extra cost to you. This page carries no affiliate link of its own.
Read our Pulsetto Fit reviewThe year-one cost, the independent rating, and a direct answer on the device-specific research.