This literature review was developed according to the guidance for narrative syntheses [11, 12] and the Enhancing Transparency in Reporting the Synthesis of Qualitative Research (ENTREQ) reporting guidelines [13]. Our research focused on papers on TMS and migraine, using the research keywords “migraine” AND “transcranial magnetic stimulation”. Papers were searched from PubMed, MEDLINE, PeDro, Google Scholar, and Cochrane Database. We included peer-reviewed journal and English-written papers, regardless of the date of publication, if they described the clinical impact of TMS on migraine in adult humans. We included only pilot and/or randomized clinical trials to consider different aspects of migraine management. Papers were excluded if they did not fit into the study’s conceptual framework, focused on other headache types, or belonged to gray literature. The search strategies were drafted by A.N. and L.B. and further refined by R.S.C., who also solved any disagreements on article inclusion.

The first search returned 1005 papers (pruned from duplicates by A.N.). Two-hundred seventy-one articles were screened for eligibility using the abovementioned research keywords, main judgment criteria, and publication types. To this end, the papers were assessed, focusing on titles and abstracts. Therefore, 16 articles were included in the present review by a full-text assessment of the papers. All the reviewed studies are summarized in Table 1 (Ref. [14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29]). The flow diagram depicting the flow of information through the different phases of the review is reported in Fig. 1.

Fig. 1.

The flow diagram depicts the flow of information through the different phases of the review. It maps out the number of records identified, included and excluded, and the reasons for exclusions.

sTMS was initially applied as an alternating real and sham pulsed magnetic field to different scalp sites, including the occipital region, in 40 patients with headache (thus, not only migraine), with positive results in terms of headache frequency and intensity reduction [30]. The study by Pelka et al. [31] adopted a global magnetic field with good results in terms of attack interruption.

Other studies applied sTMS in 42 [14] and 164 patients with migraine (82 real sTMS and 82 sham sTMS) [15] over the painful area of the skull (in migraine without an aura) or over the occipital cortex (in migraine with aura) within one hour after the aura onset, achieving a reduction in headache intensity up to the complete headache abortion for 24 h in 32% of the patients reported. Only minor side effects (including headache, migraine, sinusitis, and paresthesia) were appreciable, without significant between-group differences.

Finally, 190 patients (59 individuals with episodic migraine with or without, and 131 with chronic migraine) who were provided with sTMS (according to [14] and [15]), participated in a 3-month telephone survey. The subjects reported a significant clinical improvement in 62% of patients, including the reduction in monthly headache days and attack duration in both episodic and chronic migraine, in addition to a reduction in disability [32].

rTMS has been implemented only for a prophylactic action on migraine. A first double-blind study with real and sham rTMS on two patients with migraine and major depression reported the disappearance of migraine [16]. High-frequency (20 Hz) rTMS over the left dorsolateral prefrontal cortex (DLPFC) in six chronic migraine patients compared to five sham-rTMS patients was shown to be effective for prophylaxis of migraine, including attack frequency, headache index, and the number of acute medications. On the contrary, lower frequency rTMS (10 Hz over the left DLPFC [33], or 1 Hz rTMS over the vertex [17]) were ineffective. On the contrary, an open-label observational study employing 1 Hz rTMS over the vertex obtained a significant reduction in pain intensity, frequency and, and the duration of migraine attacks [18].

Two other studies applying 5 Hz [19] and intermittent theta-burst [20] stimulation of the left DLPFC showed partially positive data in the primary outcome (i.e., reduction of headache-days per month).

Instead, a single session of high-frequency rTMS (600 pulses at 10 Hz) over the left primary motor cortex (M1) reduced the number of headache days per month (of about 3 days), the migraine attack intensity, the quality of life impairment, and the painkiller intake, particularly within the first two weeks after the paradigm application in a chronic migraine sample [21]. Notably, providing the patients with three of these sessions was not significantly superior to the single session [22]. An 8-week, high-frequency rTMS protocol over left DLPFC in chronic migraine individuals was feasible and safe, but inferior to sham rTMS to decrease the number of headache days. Thus, M1 stimulation with rTMS seems a more promising target than the DLPFC does [23]. In a double-blind, parallel-group, randomized clinical trial employing neuronavigated TMS, the researchers delivered 600 pulses at 70% of resting motor threshold patterned in 10 trains at 10 Hz with an inter-train interval of 60 s in a daily session to the left M1. This paradigm was repeated 5 days per week, over two consecutive weeks. The authors found a significant, one-month lasting reduction in the mean pain rating scale, headache frequency, and Migraine Disability Assessment questionnaire in the real rTMS group [24].

High-frequency rTMS (10 Hz, 2000 pulses per session) over the left M1 was not superior to botulinum toxin-A injection up to 8 weeks post-treatment, but it was less effective at 12 weeks [25]. A 20 Hz-rTMS study over left M1 was shown to reduce attack frequency, headache index, and the number of abortive pills used for up to 2 months [26]. Lastly, high-frequency rTMS has been reported effective in converting chronic migraine to episodic migraine, particularly when combined with amitriptyline, compared to rTMS alone [27].

On the sidelines, we report on a deep TMS study aimed at relieving pain in chronic migraine patients. The authors reported that high-frequency deep TMS of both DLPFCs was more effective than the standard pharmacological treatment alone in reducing the frequency and intensity of migraine attacks, drug overuse, and depressive symptoms, thus resulting as a potentially helpful add-on strategy in chronic migraine [34].

An innovative approach described in a randomized, double-blinded, sham-controlled study employing swiping and, soon then, multifocal-spot high-frequency rTMS in episodic migraine patients significantly reduced headache days and severity [28].

sTMS has also been studied as a preventative treatment. In a multi-center, prospective, open-label, observational study, 217 patients were provided with daily sTMS treatment for 3 months (3 consecutive pulses as needed for acute therapy, with the possibility to add other two pulse-sequences if pain relief was not achieved within 15 min; 4 pulses twice daily were instead delivered as a preventative therapy). It was found a mean reduction of about three headache days from baseline versus placebo and a reduction in acute medication use and disability [29].

Forty-two migraine patients with both migraine with aura and migraine without an aura were provided with paired-pulse sTMS (at either 50% or 30% of the maximum stimulator output) soon after attack onset. The percentage of patients complaining of migraine relief, mainly those among migraine with aura, increased with increasing stimulation rounds, suggesting a cumulative effect of sTMS, which was used to treat episodic migraine [14].

Finally, we mention a prospective, open-label, observational study pilot study carried out on twelve adolescents (12–17 years old, mean age 15 $\pm$ 1.5 years, 14 females) with episodic or chronic migraine [35]. Patients were provided with four single pulses over the inion twice a day (two pairs of two pulses separated by 15 min intervals) for 12 weeks, with further pulses available for acute treatment. The authors found that the stimulation strategy was feasible, well-tolerated, and acceptable as a non-pharmacologic preventive treatment. Even though the efficacy remains to be assessed in larger trials, the authors found a significant reduction in headache days per month (about 5 days) and in Migraine Disability Assessment score (about –36).

The data from our literature review indicate that TMS can consistently reduce the number of migraine days per month, mitigate the intensity and duration of migraine attacks for several days after TMS application, up to one month, interrupt migraine attack quickly, and improve the quality of life. In addition, TMS is safe and well-tolerated, also considering that the majority of trials employed subthreshold stimulation intensities, as only mild light-headedness, tinnitus, dizziness, paresthesia, tinnitus, scalp discomfort, and worsening of migraine have been reported [38]. Furthermore, the more specific theoretical mechanisms of action of TMS, the reversibility of any potential side effects, and the minimal-risk options for drug interaction (i.e., a prophylactic pharmacologic therapy seems not to influence the efficacy of rTMS) are other strength points [39].

Although such non-invasive neuromodulation devices are relatively expensive, TMS seems cost-effective in some cases (e.g., cheaper than botulinum toxin-A for chronic migraine treatment, cheaper than the whole, complex pharmacological management of chronic migraine) [40, 41, 42, 43, 44, 45, 46].

The effectiveness of TMS in managing migraine critically depends on the stimulation paradigm employed. Migraine reduction improves while increasing the number and intensity of stimuli, suggesting a cumulative effect of TMS stimuli consistently with neuroplasticity principles [47]; however, this is not an absolute rule as sham TMS outperformed real TMS in some clinical trials. There is also a consistent variability in the number of sessions (ranging from one to 12), the overall number of pulses, and the number of pulses per session (from 600 to more than 20000). Another critical factor is the stimulation target. TMS seems more effective when applied to the M1 than to the DLPFC or the vertex [48], although the research of TMS over the DLPFC is growing [49].

Furthermore, TMS efficacy is significantly influenced by the stimulation frequency selection. The number of migraine attacks and days did not differ significantly between real and sham low-frequency rTMS [17]. Conversely, high-frequency rTMS often promoted more evident migraine improvement. The greater efficacy of high-frequency rTMS may lie in its capability to engage several functionally connected brain networks through a spatial summation process [50]. Furthermore, the high-frequency stimulation may have a temporal resolution sufficient to interfere more with the network connectivity (i.e., the migrainous brain is hyperexcitable between attacks), which is related to migraine attack generation.

Patients’ selection is another factor limiting judgement on TMS usefulness in migraine management. Indeed, most of the available clinical trials enrolled female-dominated samples of patients who were refractory to pharmacological therapy, thus limiting the generalizability of the present findings to the entire migraine population [51]. Furthermore, patients with episodic migraine complained of at least four episodes a month. The number of patients with migraine with aura was exiguous. Finally, most studies enrolled episodic and chronic sufferers without differences in improvement, although chronic migraine seems to respond better to rTMS than episodic migraine does [21].

This issue, together with the lack of reliable, large, sham-controlled clinical trials (actually, the reviewed clinical trials were all randomized but lacked random sequence generation or concealed allocation) and the relatively high costs of the devices, limit the widespread use and clinical applications of noninvasive neuromodulation therapies.

A non-negligible problem related to the large-scale clinical use of TMS in migraine is the not yet complete knowledge of TMS mechanisms of action in migraine concerning both the prophylactic treatment (mainly rTMS) and the acute phase treatment (sTMS). However, a solid rationale for adopting TMS to manage migraine remains.

Migraine patients’ cortical excitability and connectivity are largely abnormal and intimately correlated with migraine attack pathogenesis [52]. Furthermore, migraineurs’ brain shows an abnormal responsivity to TMS pulses as suggested by high-amplitude motor evoked potentials [53], lower phosphene threshold [54, 55, 56], high sensitivity of M1 to light exposure during both the pre-ictal and ictal phases of migraine [57, 58]. This baseline hyperexcitability concerns the M1 and DLPFC, which play a significant role as pain modulators, and the secondary somatosensory cortex. Furthermore, the baseline activity of such brain areas, which seems largely abnormal in chronic pain conditions [59, 60], has been proposed to tune other pain-associated neural areas [61]. Similarly, thermal pain measures and cold and heat pain threshold are largely abnormal in migraineurs compared to healthy controls, suggesting a reduced intracortical inhibition and a hypofunction of inhibitory pain modulation mechanisms [62]. All these issues justify the potentially positive response of patients with migraine with or without aura to rTMS, which has been estimated putatively capable of reverting the mentioned above (mal) adaptive plasticity in chronic pain conditions [63, 64]. Contemporarily, these baseline alterations may explain some of the paradoxical responses of the M1 in migraine to both inhibitory (low-frequency) or facilitatory (high-frequency) rTMS paradigms [65, 66, 67, 68, 69], explaining the reason why sham-rTMS was often superior to real-rTMS. However, over and above that, high-frequency rTMS carries a high potential of inducing a placebo effect, which can be usefully leveraged to enhance patients’ coping strategies [70]. Furthermore, rTMS over sensorimotor and premotor regions seems to entrain various biochemical mechanisms related to pain control, including plasma $\beta$-endorphins, vasoactive neuropeptides, predominantly calcitonin gene-related peptide, substance P, and neurokinin A [71]. However, it remains uncertain how long the effects of TMS last, although applying more stimuli in structured patterns may increase the duration of the aftereffects [47].

Regarding migraine with aura, TMS is proposed to challenge the spread of cortical spreading depolarization, particularly when applied over the occipital cortex [72]. High-intensity TMS pulses may inhibit the cortical spreading depolarization by activating inhibitory GABA${}_B$ fibers in the upper cortical layers [39]. Another contribution to TMS efficacy in migraine with aura may come from a blocking action onto the activity of third-order neurons in the VPM thalamic nucleus, potentially through reverberant cortico-thalamocortical loops [9]. Lastly, cortical TMS may modulate thalamus and hypothalamus activity by entraining motor cortex projections to the medial thalamus and the anterior cingulate/orbitofrontal cortices, with relevant consequences on endogenous painkiller mechanisms [71].

Our analysis has some limitations to acknowledge. First, we included only published clinical trials. It is thus necessary to be cautious when making conclusions that come from synthesizing evidence. Even though randomized, double-blind, controlled clinical trials provide reliable results, we had to include only 16 trials in this analysis, thus likely limiting us from achieving any different outcome for TMS employment in migraine management. Furthermore, all studies in the analysis presented lots of heterogeneity, including lack of grouping by migraine severity, sexuality, age, paradigm setup, and migraine pharmacological treatment.

Moreover, the patients enrolled in the studies came from general hospitals or major institutions, which somehow limits data generalizability. Lastly, we did not assess of quality bias or blinding validity in the included trials. However, this type of assessment may be challenging and unreliable, as a quantitative synthesis could be hard to perform consistently with the significant heterogeneity in headache type studied, protocol stimulation parameters, location and duration of treatment, and outcome measures.