Migraine and cluster headache apart, the International Classification of Headache Disorders, 3rd edition (ICHD-3), contains other disabling primary headaches; among these are trigeminal-autonomic cephalalgias, including hemicrania continua and chronic paroxysmal hemicrania, or the short-lasting unilateral neuralgiform headache attacks, and more than a dozen primary headaches classified as “other,” such as cough headache, hypnic headache, or exertional headache (Table 1) (1).

A majority of these headaches have a low prevalence rate, and the therapeutic armamentarium is usually limited to pharmacological options with a wide range of side effects, such as indomethacin or antiepileptic medications (2).

Vagal nerve stimulation (VNS) was first investigated in 1990 for the treatment of intractable epilepsy (3, 4). In this review, we will cover findings related to non-invasive, transcutaneous cervical (nVNS). The United States Food and Drugs Administration approved gammaCore (Figure 1) for the acute and preventive treatment of migraine headache in adults and adolescents, adjunctive preventive in adults with cluster headache, and acute treatment in episodic cluster headache, and recently, also for adults with hemicrania continua and paroxysmal hemicrania (5).

We performed a narrative review of the literature using PubMed, EMBASE, and Web of Science on September 2022. Keywords included primary headaches 3.2–4.10 from the ICHD-3 (1) AND “gammacore” OR “vagus nerve stimulation” and “pathophysiology” AND “gammacore” OR “vagus nerve stimulation.” Articles addressing the mechanism of action and translational research that help comprehending the pathophysiology of vagus nerve stimulation were included. The search included publications in English and Spanish. Individual case reports were included. Additional information was obtained from abstracts presented in international conferences. There were no exclusion criteria, given the low number of studies in this specific topic.

The first time that electrical VNS was used in research was in an animal model in 1952, to demonstrate that the effects observed following carotid compression, spontaneous movements, were due to the neurogenic component and not to blood pressure change (6). Consistent with the neurovascular pathophysiology of primary headache disorders (7), the efficacy of VNS does not seem to involve vascular changes, as mentioned in a 2015 abstract of a pilot study (8).

The mechanism of action of VNS in primary headache disorders has not been fully elucidated. Its effects are multiple and have been collated into four areas, namely, nociceptive modulation, neurotransmitter regulation, cortical spreading depression, and autonomic effects (9). The anatomical basis of the therapeutic effect of nVNS is likely to be the connection between the vagus nerve with relevant central nervous system structures (Figure 2). Among them, the key intersection in the medulla with the nucleus tractus solitarius (NTS) is important. The NTS is activated by nociceptive afferents from intracranial structures innervated by the trigeminal nerve (10) and projects in a topographic fashion to several higher structures, including medullary structures, such as the reticular formation or the periaqueductal grey. NTS also projects ipsilaterally to the parvicellular part of the ventral posteromedial thalamic nucleus (11) and connects with the caudal parts of the spinal trigeminal nucleus and thus the trigeminocervical complex (12).

The spinal part of the trigeminal nucleus and the other three vagal nuclei, the NTS, nucleus ambiguous, and dorsal motor nucleus, are highly interconnected (13, 14). The antinocifensive effect of VNS in rats can be inhibited by the administration of gamma aminobutyric acid A and serotonin receptor antagonists in the upper spinal cord, causing an effect similar to that of sumatriptan (15). The mechanism of action of VNS also involves other neurotransmitters, and among these is the opioidergic system (16), especially through the activation of delta opioid receptors (17).

Electrical stimulation of vagal afferents is capable of directly altering the response of trigeminal and trigeminothalamic neurons that have been stimulated with noxious inputs in the orofacial region of cats. These were wide-dynamic-range neurons that responded both to tactile stimuli and to painful heat stimuli. Continuous stimulation at 5 Hz, 3 ms, 2 mA inhibited a majority of the heat-induced neuronal responses, whereas one-third either remained unaffected or were facilitated. Intermittent stimulation before the test stimulus was assessed in Aδ and C-fibres, delivering seven pulses at 333 Hz, 5 mA, in 200 ms, with similar results of inhibition or inactivity (18). In rats, the application of VNS for 24 h at 2 mA intensity, 20 Hz frequency, 0.5 ms pulse width, and a duty cycle of 20/18 s in ON/OFF, respectively, reduced the responses to formalin injected in trigeminal areas, as compared with sham stimulation (19). In another study, continuous VNS at 10 Hz suppressed trigeminal responses evoked by dural electrical stimulation in 48% of neurons; however, there was a 29.5% of neuron activity facilitated by VNS, and 22.5% was unresponsive to stimulation. This difference was first hypothesised to be due to the recruitment threshold of the different types of neurons, with rapidly conductive pathways needing a lower intensity of stimulation, and high intensity stimuli would activate slowly conducting pathways and promote an antinociceptive effect; however, it may be actually affected by descending pain control mechanisms (20). Indeed, the beneficial effect might not even be dependent on C-fibres (21, 22). Another study evaluating single pulses of VNS in dural and superior salivatory nucleus–evoked trigeminocervical firing found the beneficial effect to be dose-dependent and side-independent (23).

The central activity is further supported by studies assessing VNS in processes with higher structures implicated in pain processing. Activation of the trigeminal-autonomic reflex is significantly inhibited by nVNS, which involves several central structures, including the pontine nucleus, hypothalamus, or parahippocampal gyrus (24, 25).

Associated symptoms involving central sensitisation processes such as allodynia also respond to VNS stimulation, with a reduction in extracellular glutamate (26), and a modulation of the nociceptive withdrawal reflex is also possible following VNS, with an increase in the threshold to stimuli in animals and humans’ withdrawal effect on healthy subjects (27). Electrophysiological studies support a modulation of the pain matrix at cortical levels (28), and a direct or indirect cortical effect could also be inferred from animal studies showing a modulation of cortical spreading depression (29, 30).

Neuroimaging research supports the role of VNS in modulating the activity of brainstem and supratentorial structures. From studies in epileptic patients with cortical frontal and bitemporal in H₂¹⁵O PET scan 2 milliamps 30 Hz 60 s, regional cerebral blood flow (rCBF) was increased in the ipsilateral structures of the left posterior cerebellum and left putamen and in the contralateral areas of the right middle temporal gyrus and right thalamus. Larger increases were seen in the cerebellum and thalamus in the patient with the best clinical response (ipsilateral cerebellum and contralateral thalamus) (31). Dorsal medulla and somatosensory cortex contralateral to the stimulation, hippocampus, and amygdala can also present alterations in rCBF in participants with different VNS stimulation patterns (32). More recent studies assessing transcutaneous VNS on healthy subjects using functional MRI found significantly activated primary vagal projections, remarkably, an activation of the NTS and deactivation of the spinal trigeminal nucleus, among others (33). At 15 V, cervical VNS is capable of eliciting a vagal response in more than 80% of participants (34).

An inflammatory component involving cytokines has been suggested to be part of the pathophysiology, outside the central nervous system, of other primary headaches (35, 36). Assuming a possible regulation of inflammatory mediators through a cholinergic pathway (37), one study found changes in cytokine levels in migraineurs following treatment with nVNS (38) that were, however, not significantly different in comparison with those of controls in previous studies (35, 36).

Several studies have demonstrated an excellent safety profile (39–42) and preliminary efficacy (39–41) in episodic and chronic migraine, with a reduction in headache intensity, severity, and duration of migraine attacks in three open label trials, two of them using the stimulation acutely. Participants with menstrual and menstrually related migraine (43) and migraine in adolescents (44) showed a reduction in the number of migraine days and analgesic consumption and successfully aborted attack, respectively, when treated with nVNS. Three randomised controlled studies included participants with cluster headache, who also presented a good tolerability to the device, with especially good efficacy in patients with the episodic form. The Prevention and Acute Treatment of Chronic Cluster Headache (PREVA), against standard of care, showed a significantly greater reduction in the number of attacks per week vs. controls (−5.9 vs. −2.1, respectively), ACT1 demonstrated response rates of 34.2% vs. sham 10.6% (P = 0.008) and ACT2 showed similar results, both in patients with the episodic form, with an efficacy of 48% vs. 6% in the sham group (P < 0.01) (45–49). nVNS has proved cost-effective in other trigeminal-autonomic cephalalgias, with a lower cost in combination with the standard of care acute treatment approach, higher benefits in terms of quality-adjusted life years, and potentially reduced visits to clinics and emergency departments (50).

Most papers mentioned here have reported some kind of benefit in patients with long-lasting, debilitating primary headache disorders that were unresponsive to several lines of preventive treatments. For therapeutic studies, this corresponds with the level of evidence 4, which can be difficult to improve in disabling conditions with low prevalence. The reason why a cohort of patients responded better to stimulation than others with the same diagnosis is unknown. Device parameters, dose, or duration could play a key role in treatment efficacy, and given the clinical heterogeneity of these primary headaches, a careful tailoring of the treatment, personalised for every individual clinical presentation, could be considered.

In general, there has been incongruence regarding the different methodological designs of studies, as seen in the diverse range of electrical parameters reported in the pre-clinical models discussed herein. Likewise, human studies assessing VNS for heart failure presented a contrasting frequency and dose of stimulation (56). Similarly, these methods are heterogeneous in headache studies (Table 1). It has been proposed previously that individually adjusted doses could have influenced the treatment outcome (48).

One of the benefits of cervically applied nVNS is the possibility to regulate the intensity of the stimulation by the patient. Unfortunately, pain threshold has an inter-individual variability. Even though pain threshold can increase in a majority of patients following nVNS stimulation (57, 58), in more than one-third of them, it can diminish (58). This variability could be especially relevant in patients with a migrainous background (59), meaning that what a subject considers a tolerable intensity of stimulation could be unbearable for another. In view of the fact that a majority of people experience a headache throughout the course of their lives (60) and the high prevalence of migraine (61), a migrainous background, especially allodynic symptoms, should be considered when assessing tolerability to nVNS.

Among the proposed checklist of stimulation parameters to be reported in VNS, current (intensity) should be specified. When the option of selecting a dose is available, and there is titration to sensation or muscle response (such as lip pulling), the actual dose selected should be reported (62). Most studies described in this article included stimulation amplitude adjustments available in their methods, but the data regarding mA of output current have not been presented. Differences in efficacy among neuromodulation techniques in other primary headaches such as migraine (63) could be due to unstandardised settings and a lack of knowledge about the total doses of devices other than gammaCore. The therapeutic dose needed could also vary inter-individually and be age-dependent: younger patients may need higher output currents (64).

Another explanation for the disparity of results could be the treatment duration, which was also an important factor in epilepsy studies (65–67). Descending pain inhibition is not an immediate effect of nVNS (68). Studies on cluster headache assessed efficacy after 3 months of treatment (69). The results presented above represent an example that the latency of response to preventive treatment in primary headaches (70), also with neurostimulation, could be demonstrated after 3 months or above.

Also, patients with hemicrania who responded to treatment may need a significantly longer treatment schedule, as compared with those who did not respond in a retrospective cohort, and a higher number of total daily seconds (Villar-Martinez and Goadsby, unpublished results).

Adherence could also represent a main issue when assessing nVNS efficacy, as seen in migraine studies (71). Adherence should be considered in other primary headaches when assessing responsiveness.

A smaller and discrete device that uses transcutaneous auricular vagus nerve stimulation has been studied in migraine. The cutaneous distribution of the auricular branch of the vagal nerve is heterogeneous in cadaveric studies (72). However, the mechanism of action of auricular nerve stimulation also seems to target central pathways and structures involved in pain modulation. A significant modulation of activity in several brainstem areas was seen in functional MRI, including a decreased activity of the NTS, but also the noradrenergic locus coeruleus, serotonergic nuclei raphe, or parabrachial nuclei, and an increased connectivity with higher structures involved in the pain matrix, including cortical areas such as the temporoparietal junction or the somatosensory cortex (73). A randomised, double-blind monocentric clinical trial comparing low and high frequency of stimulation, showed promising results (74). At the time of the writing of this article, we could not find any literature referring to transcutaneous auricular vagus nerve stimulation for other primary headaches.

Treatment with nVNS is well tolerated and efficacious in patients with other primary headache disorders, especially indomethacin-responsive headaches such as cough headache, hemicrania continua, and paroxysmal hemicrania. As with any preventive treatment, nVNS should be trialled for at least 3 months to consider any lack of efficacy, and assessment of adherence and correctness of the technique of stimulation should be granted before withdrawal. Training should be offered and the technique assessed in unresponsive patients. Tolerability could be inversely related to sensitisation in patients with a concomitant migrainous background. A more standardised way of reporting stimulation parameters is desirable for future studies.