A clinical synthesis based on the European Academy of Neurology guideline (Bendtsen, Eur J Neurol 2019), the ICHD-3 international classification of headache disorders currently in force, the reference review in Lancet Neurology (Bendtsen 2020), the worldwide incidence and prevalence meta-regression 1945-2024 (Jeong, J Clin Neurol 2026) and on real published cases indexed in PubMed: 35 references checked one by one against PubMed and CrossRef metadata.

How is trigeminal neuralgia recognised from the patient's story alone?

The four features, and nothing else

The 3^rd edition of the international classification of headache disorders (ICHD-3) remains the version in force in August 2026. It has to be said, because an ICHD-4 is circulating in conversation: an “alpha” version was indeed announced in 2020 by Goadsby and Evers, but the work is explicitly under way, with two progress reports published in Cephalalgia in 2024 and then 2025 (PMID 38388354). Citing “the ICHD-4” as an authoritative reference would be a mistake today: it is the 2018 ICHD-3 that stands (PMID 29368949).

The diagnosis, under code 13.1.1, rests on recurrent paroxysms of unilateral facial pain, in the territory of one or more branches of the trigeminal nerve, with no radiation beyond it, and meeting all of the following characteristics:

• a duration running from a fraction of a second to two minutes;
• a pain of severe intensity;
• a quality that is electric shock-like, shooting, stabbing or sharp;
• precipitation by innocuous stimuli within the affected trigeminal territory.

The fourth criterion is the pivot. The classification takes care to state that some attacks may be, or may seem, spontaneous: there must then be a history or a finding of pain provoked by an innocuous stimulus for the criterion to hold. It adds that the clinician should ideally try to reproduce the triggering phenomenon, while acknowledging that this is not always possible, whether the patient refuses or the trigger zone is awkward to locate. In other words: the trigger is something you look for, not something you infer.

The trigger: the most useful sign in the presentation

What strikes you in the story is how ordinary the triggers are. In a retrospective series of 36 patients from a tertiary orofacial pain clinic, pain was set off by light touch to the face, washing, tooth brushing, speech and chewing (Limrachtamorn 2026,PMID 42220298). None of these actions hurts a healthy person: that is the very definition of an innocuous stimulus. A patient who describes pain brought on by a sustained chewing effort, or by firm pressure on the muscle, is describing something else.

Two details of the story are worth an examination manoeuvre. First, avoidance: the patient no longer shaves one side, no longer wears make-up, eats on the other side, speaks less. Then, the refractory period: ICHD-3 notes that after a paroxysm there is usually a period during which the pain cannot be triggered. That detail, reported spontaneously, is hard to invent and has no equivalent in myofascial pain.

Purely paroxysmal, or with concomitant continuous pain

ICHD-3 distinguishes, for the classical form as for the idiopathic form, a purely paroxysmal variant and a variant with concomitant continuous pain (a background pain of moderate intensity, burning in quality, between the shocks, in the same territory). The old terminology spoke of “typical” and “atypical” neuralgia; it has been abandoned.

The distinction is not cosmetic. In a surgical series of 330 patients, 31 (9.4 %) had a form with concomitant continuous pain (Segura-Lozano, Cureus 2024,PMID 38855490). These patients respond less well: after Gamma Knife radiosurgery, the rate of complete long-term pain relief was 15.2 % in patients with continuous background pain against 35.7 % without it (Dong, Medicine 2024, 46 vs 112 patients,PMID 39705465). For the physiotherapist the point is twofold: continuous background pain does not disqualify the diagnosis, a frequent error, and it signals a harder road ahead.

Sources: ICHD-3 (PMID 29368949); classification by Cruccu et al., Neurology 2016 (PMID 27306631); EAN 2019 guideline (PMID 30860637); Gamma Knife figures from Dong 2024 (PMID 39705465).

What do you need to know about the trigeminal nerve to understand where the pain sits?

Three branches, and one precise limit

The trigeminal is the fifth cranial nerve and the main sensory nerve of the face. It divides into three branches: V1 ophthalmic (forehead, upper eyelid, bridge of the nose, cornea), V2 maxillary (lower eyelid, cheekbone, ala of the nose, upper lip, upper teeth and gum, palate) and V3 mandibular (lower lip, chin, lower teeth and gum, anterior two thirds of the tongue, pre-auricular region), V3 being the only branch that also carries motor innervation, for the muscles of mastication.

The limit that matters in practice lies elsewhere: the angle of the mandible and the region behind the ear are not trigeminal. That territory belongs to the upper cervical roots, C2 and C3, of which the greater occipital nerve is the posterior branch. A flash of pain sitting there is not trigeminal neuralgia, and the question to ask becomes the one in the chapter of our article on occipital neuralgia.

Which branch, and what that does not tell you

The two lower branches dominate. In the Badran series, joint involvement of V2 and V3 concerned 40.5 % of patients, a single branch 31.7 % and all three branches 12.5 %; the right side of the face was affected in 67 % of cases (PMID 41663983). It is this V2-V3 predominance that explains the dental detour: the pain projects exactly where the patient and their practitioner are looking for a tooth.

One belief deserves correcting here, because it circulates and because it has been measured. You often read that involvement of the first branch should raise the fear of a secondary cause. The meta-analysis in the European guideline calculated the value of that sign: sensitivity 27 % (17-39), specificity 72 % (64-79), and a positive likelihood ratio of 1.0 (PMID 30860637). A likelihood ratio of 1.0 literally means that the sign does not change the probability. V1 involvement is not a red flag; sensory deficit and bilaterality are, and that is the subject of chapter 5.

The mechanism, in two steps

The reference review in Lancet Neurology describes a two-stage mechanism (Bendtsen et al. 2020,PMID 32822636). Contact between an artery and the trigeminal root at its entry into the brainstem produces, when it is marked enough to deform the nerve, a focal demyelination. Demyelinated fibres become the seat of ectopic discharges and of cross-talk from one fibre to another: an innocuous tactile input then recruits nociceptive fibres. This is the most widely accepted explanation of the clinical paradox, that a light touch produces an unbearable shock.

Which artery? A recent series of 412 patients whose conflict was confirmed in theatre during a microvascular decompression gives a solid answer: an offending vessel was found intraoperatively in 407 of them (98.8 %), and the superior cerebellar artery was the commonest cause (Huang et al., J Neurol 2026,PMID 42342895).

Two practical consequences follow. The first: in the form secondary to multiple sclerosis, the lesion is not vascular but a demyelinating plaque in the root entry zone or in the pontine trigeminal complex, the same mechanism with a different cause (Di Stefano, J Headache Pain 2019,PMID 30782116). The second, more important still for a physiotherapist: the lesion sits at the root of the nerve, in the posterior fossa, not in the tissues of the face. No technique applied to the masseters, the pterygoids or the peri-orbital tissues reaches the site of the problem. That is a mechanistic reason, not an opinion, for not expecting facial manual therapy to treat the paroxysms.

One nuance deserves stating: the mere existence of a neurovascular contact on MRI is not enough. The European guideline is explicit: demonstrating a neurovascular contact must not be used to confirm the diagnosis; it serves to decide whether and when to refer a patient for microvascular decompression. Contacts exist in people who have no symptoms; only those that deform the nerve count.

Who is affected, and how often?

What the widest available measurement says

The most complete synthesis to date is a systematic review with meta-regression covering 1945 to 2024: 17 eligible studies, a cumulative population of 174,485,958 people including 109,070 affected patients, across twelve countries and four continents (Jeong et al., J Clin Neurol 2026,PMID 41517817). It gives a pooled worldwide incidence of 25.33 per 100,000 person-years, an annual prevalence of 45.38 per 100,000 and a lifetime prevalence of 108.43 per 100,000.

These three numbers must never be quoted on their own. Their 95 % confidence intervals are respectively 11.87-54.02, 15.41-133.61 and 30.54-384.18: the upper bound is worth, depending on the parameter, four to twelve times the lower one. That spread is not a flaw in the review, it is its result: the primary studies differ too much, in method as in populations, for a single value to mean anything. The seminal Rochester study, still the historical reference, found a crude incidence of 4.3 per 100,000 over forty years of follow-up (Katusic, Ann Neurol 1990,PMID 2301931), six times less than the modern pooled estimate. A recent Swedish population study confirms that the order of magnitude is still debated (Svedung Wettervik, Eur J Pain 2023,PMID 36680398).

Age, and why it is a signal

Trigeminal neuralgia is a disease of the second half of life. The mean age was 52.5 years in Badran's neurosurgical series (range 17-85 years,PMID 41663983) and 66.8 ± 10.7 years in a tertiary orofacial pain clinic (PMID 42220298). That distribution is what gives weight to the first red flag: the meta-analysis in the European guideline established that patients with a secondary form are significantly younger (p < 0.0001) than those with a primary form (PMID 30860637). A typical presentation in a thirty-year-old is not good news; it is an indication for imaging.

Multiple sclerosis is the commonest identifiable secondary cause. Patients with multiple sclerosis carry a risk of trigeminal neuralgia twenty times higher than the general population, and 1.9-4.9 % of them have it; conversely, multiple sclerosis is found in 2-14 % of patients with trigeminal neuralgia. In these forms, age at onset lies between 40 and 50 years, and bilateral involvement is reported in about 18 % of patients (Di Stefano, J Headache Pain 2019,PMID 30782116).

A rare disease inside a common complaint

The two scales have to be put side by side. An annual prevalence of 45 per 100,000 means, in simple arithmetic from the pooled figure, about nine patients a year in a town of 20,000 inhabitants (an order of magnitude, not a local measurement). Meanwhile orofacial pain is an everyday reason for consulting: a recent review in Practical Neurology recalls that it can arise from more than 160 different conditions, and that trigeminal neuralgia is the rarest form of all, far behind temporomandibular disorders and post-traumatic neuropathic pain linked to dental care (Renton, Pract Neurol 2026,PMID 42320903).

This asymmetry dictates the clinical stance. Faced with facial pain, the prior probability leans massively towards something other than the trigeminal nerve; it is the presentation, not the frequency, that should make you raise a hand. And the reverse is true: it is because the disease is rare that those who have it pass through so many clinics before being recognised.

How do you avoid confusing it with a TMJ disorder, dental pain, occipital neuralgia or cluster headache?

The differential table for facial pain

Each row carries the code of the corresponding international classification, so that the reader can go back to the criteria themselves rather than rely on a summary.

Codes and criteria: ICHD-3, Cephalalgia 2018 (PMID 29368949). Diagnostic criteria for temporomandibular disorder: DC/TMD, Schiffman & Ohrbach, JADA 2016 (PMID 26922248). International Classification of Orofacial Pain (ICOP), Cephalalgia 2020 (PMID 32103673). On the boundary between genuine neuralgia and post-dental trigeminal pain, a diagnostic pathway study of 672 patients from a tertiary centre distinguishes five phenotypes, among them confirmed alveolar neuropathy and simple odontogenic misclassification (Shemesh et al., Diagnostics 2026,PMID 42279542).

Trigeminal and occipital: two nerves, two presentations

The two conditions share a word and an attack duration, and nothing else. Occipital neuralgia affects the greater occipital nerve and hurts at the back of the head; trigeminal neuralgia affects the fifth cranial nerve and hurts in the face. That is the sentence to remember.

Three requirements in the ICHD-3 13.4 criteria complete the separation of the two presentations, and they are absent in trigeminal neuralgia: occipital pain comes with dysaesthesia or allodynia of the scalp, it is associated with tenderness on palpation of the nerve branches concerned or with trigger points at the emergence of the greater occipital nerve, and it is temporarily relieved by an anaesthetic block of the nerve. A patient whose pain yields to thumb pressure below the nuchal line does not belong in this article; they belong in our synthesis on occipital neuralgia.

The confusion nevertheless runs both ways, and it is documented. Anatomical variations in the course of the greater occipital nerve make some occipital pain radiate towards the maxilla or the external ear, which can wrongly point to the trigeminal nerve. Symmetrically, a pure V3 trigeminal presentation projects close to the angle of the mandible, at the border of the cervical territory. The practical rule: look for the trigger. A trigger on the skin or mucosa of the face, set off by a light touch, points to the trigeminal nerve; a pain on palpation of the nerve trunk at the neck points to occipital neuralgia.

The TMJ: the commonest neighbour, and the costliest in sessions

This is the confusion that costs the patient most, because temporomandibular disorder is itself a legitimate indication for physiotherapy: the patient enters a coherent pathway, but for the wrong disease. The reference review by Maarbjerg, Di Stefano, Bendtsen and Cruccu stresses that confusion between the two persists in practice despite distinct criteria (Cephalalgia 2017,PMID 28076964).

Three discriminators are usually enough. The rhythm: temporomandibular disorder gives a dull pain, present most days; the neuralgia gives shocks of a few seconds separated by strictly pain-free periods. The trigger: the first responds to a sustained mandibular effort and to firm palpation of the muscles, the second to a light touch that ought to produce nothing. Reproducibility at examination: TMJ pain is reproduced by palpating the masseter and moving the mandible, which is part of the standardised assessment; a neuralgia is reproduced by brushing a trigger zone a few millimetres across, often perioral or nasolabial.

One point deserves saying bluntly: the two can coexist, and a genuine temporomandibular disorder does not rule out an associated trigeminal neuralgia. If a patient treated for the TMJ keeps having paroxysmal shocks set off by a light touch, the question must be reopened. Our synthesis on TMJ syndrome covers the temporomandibular side, and the feature on bruxism and temporomandibular disorders the parafunctional side.

When should you suspect a secondary cause?

About one form in seven is not primary

Secondary trigeminal neuralgia, that is to say attributable to an identifiable neurological disease other than a simple vascular conflict, accounts for up to 15 % of cases. Multiple sclerosis plaques are the abnormality most often identified; then come space-occupying lesions of the cerebellopontine angle, arteriovenous malformations and, more rarely, inflammatory or ischaemic lesions of the brainstem (Di Stefano, J Headache Pain 2019,PMID 30782116).

These forms do not merely have another cause: they respond less well. A retrospective observational study matching 36 patients with multiple sclerosis and trigeminal neuralgia to 36 patients without multiple sclerosis, matched for age and sex, found in the multiple sclerosis group higher pain scores, significantly more frequent recourse to blocks and to radiofrequency, and a markedly higher prevalence of refractory forms, correlated with the EDSS score (Yön & Ulusoy, Noro Psikiyatr Ars 2026,PMID 41613036).

Highly specific, barely sensitive: what that means at the bedside

The chart below sets side by side, for each sign, its sensitivity and its specificity as the meta-analysis in the European guideline calculated them. The shape of the chart is the message: every specificity bar is high, every sensitivity bar is low.

The consequence for the physiotherapist

It is simple to state and sometimes hard to hold to: it is not for the physiotherapist to decide between a primary and a secondary form. That requires an MRI, and imaging is not within their scope. What is within their scope is to do three things.

First, look for a sensory deficit, because it is the most informative sign of the lot and because it is tested with a wisp of cotton wool and a blunt point, comparing the two sides in the three territories. Then, record what you find and what you do not find: a sensory examination recorded as normal has value for the doctor who will see the patient. Finally, write. A letter that says “paroxysms of less than two minutes, right V2 territory, reproducible nasolabial trigger, normal comparative sensory examination, no improvement after six sessions” allows an immediate referral; a letter that says “persistent facial pain” allows nothing.

What can medical treatment and surgery do?

Carbamazepine, a solid reference on fragile evidence

The European guideline is clear: for maintenance treatment, carbamazepine (200 to 1200 mg a day) or oxcarbazepine (300 to 1800 mg a day) remain the most effective drugs, particularly at early stages; higher doses are sometimes needed (PMID 30860637). In an acute exacerbation, admission to hospital may be needed in order to rehydrate and to give intravenous fosphenytoin or lidocaine; a recent case documents exactly this course during a crisis in a neuralgia secondary to a petroclival meningioma, with almost complete relief after the loading dose (Mochizuki et al., Front Pain Res 2026,PMID 42100134). These choices belong to the doctor; they appear here so that the physiotherapist understands their patient's pathway and recognises the adverse effects that will explain the cancelled sessions.

The evidence does have to be faced, however. The Cochrane review devoted to carbamazepine in chronic neuropathic pain included ten studies and 480 participants, among them patients with trigeminal neuralgia. Its authors write that no study provided first- or second-tier evidence for any efficacy outcome, and conclude that carbamazepine is “probably effective in some people”, with this explicit caveat: no trial lasted longer than four weeks, none had satisfactory reporting quality, and none used an outcome equivalent to substantial clinical benefit (Wiffen et al., Cochrane Database Syst Rev 2014,PMID 24719027).

This does not mean the drug does not work: clinical experience and the strong stance of the European guideline say the opposite. It means that the reference treatment for a disease described since the 18^th century rests on old, short and small trials. A physiotherapist wondering why the literature of their own discipline is so thin on this subject will find part of the answer here: it is thin everywhere.

The adverse effects, for their part, are well documented and frequent. In four studies of that review, 65 % of patients on carbamazepine (113 of 173) had at least one adverse effect, against 27 % on placebo: for every five patients treated, two suffer an adverse effect they would not have had on placebo. Drowsiness, dizziness, ataxia and impaired concentration explain many a missed session, and sometimes a complaint of “imbalance” wrongly brought to the physiotherapist.

The second lines, and botulinum toxin

When first-line treatments fail for lack of efficacy or of tolerance, the European guideline gives weak recommendations, based on low to very low quality evidence, to lamotrigine, gabapentin, botulinum toxin type A, pregabalin, baclofen and phenytoin, alone or in combination (PMID 30860637).

Botulinum toxin type A has been the most studied since. An updated meta-analysis of 23 studies, four of them randomised trials, found a significant reduction in the visual analogue scale score from baseline in the randomised trials (effect size −4.05; 95 % CI −6.13 to −1.97; p = 0.002), concluding that it is an effective and safe option in refractory patients while calling for better quality studies (Hu et al., Clin J Pain 2024,PMID 38385501). Four randomised trials out of twenty-three studies: the caution of the conclusion is justified.

Surgery: when, and which one

The guideline is explicit about timing: it is recommended to offer surgery if the pain is not sufficiently controlled by drugs or if they are poorly tolerated, and there is no need to have tried every drug before asking for a neurosurgical opinion. This is information many patients do not have, and which a physiotherapist may legitimately remind them of while sending them back to their doctor. A prospective study cited by the guideline shows, moreover, that 65 % of patients referred to a specialist centre were still adequately controlled on drugs two years later, while 35 % were sent on to surgery.

In classical forms (those in which MRI shows a neurovascular conflict deforming the nerve), microvascular decompression is recommended as the first surgical line, with a strong recommendation despite low quality evidence. The historical series by Barker and Jannetta, covering 1,185 patients operated on over twenty years, reported 70 % excellent results at ten years, that is to say patients pain-free and off medication, plus 4 % with occasional pain not requiring long-term treatment (N Engl J Med 1996,PMID 8598865). The same series puts a number on the price to be paid, and it must be given alongside the benefit: two deaths shortly after the operation (0.2 %), one brainstem infarct (0.1 %) and sixteen ipsilateral deafnesses (1 %). Thirty per cent of patients relapsed during the study period, most of them within the first two years, and 11 % were reoperated on; beyond ten years the annual recurrence rate falls below 1 %.

Single source: EAN 2019 guideline, a synthesis of 45 non-randomised cohorts (PMID 30860637). No randomised trial has compared these techniques with each other: these figures are not direct comparisons.

In idiopathic forms, on the other hand, no recommendation can be made for choosing between the neuro-ablative techniques, nor between those and decompression: neuro-ablative treatments are to be preferred when MRI shows no neurovascular contact. Finally, in secondary forms, treatment generally follows the same principles as in primary forms, and the guideline adds that, on top of medical and surgical management, one should offer these patients psychological and nursing support. That is the only non-drug, non-surgical opening in the document, and the next chapter draws the consequences.

What can physiotherapy actually do, and at what level of evidence?

Measurement, rather than impression

Rather than assert that “the evidence is limited”, here is the measurement, made on 15 August 2026 on PubMed by crossing MeSH descriptors. The descriptor Trigeminal Neuralgia holds 8,451 references. Crossed with Physical Therapy Modalities, 267 remain, of which 22 randomised trials. Crossed with Exercise Therapy: zero. With Musculoskeletal Manipulations: nine references, no randomised trial.

The 22 trials deserve reading one by one, because the MeSH descriptor Physical Therapy Modalities does not cover what a physiotherapist calls their practice. Going through the 22 title by title gives: fifteen trials of percutaneous radiofrequency or related interventional neuromodulation (thermocoagulation or pulsed radiofrequency of the Gasserian ganglion, that is to say pain-specialist procedures), five trials of acupuncture or electroacupuncture, one trial of motor cortex stimulation, and a single trial of TENS. Fifteen plus five plus one plus one: all 22 are accounted for, none is left aside. No trial of exercise, none of manual therapy, none of patient education. A scoping review published in 2026, which mapped the interventions used in this condition between 2020 and 2025 from five databases, arrives at the same landscape: pharmacological, surgical and non-invasive interventions, with rehabilitation forming no documented category at all (Mateos-Romero et al., Pain Manag Nurs 2026,PMID 42315398).

The cross-check confirms this void on the guideline side. In the full text of the European Academy of Neurology guideline, the terms “physiotherapy”, “physical therapy”, “exercise”, “rehabilitation”, “manual therapy” and “massage” do not appear a single time. By contrast, “psychological” appears three times, “nursing” three times and “self-management” twice, and the document explicitly recommends offering patients psychological and nursing support (PMID 30860637). So an indication does exist: it simply is not the one we would have wanted.

What the only TENS trial contains, and why it is not enough

This single trial deserves describing precisely, because it is regularly cited as proof of the efficacy of TENS in this indication. It covers 31 patients who were refractory or partial responders to drug treatment, randomly allocated between two modes of TENS, continuous and burst, with pain assessed before treatment and fifteen days later. Twenty-six patients out of 31 (83.7 %) improved, the continuous mode doing slightly better than the burst mode (Yameen et al., J Pak Med Assoc 2011,PMID 22204173).

The figure of 83.7 % is impressive, and it says nothing about the efficacy of TENS. There was no group without TENS. Both arms received the treatment; the comparison was about the mode, not about the principle. An improvement rate without a comparator, in a disease that has spontaneous remissions (the European guideline does note that periods of partial or complete remission occur), measures the natural course as much as the treatment. Thirty-one patients, fifteen days of follow-up: this is a pilot study, and it should be cited as such.

Electroacupuncture: the only methodologically solid trial

A recent trial changes the landscape, and honesty requires reporting it as carefully as the previous one, in the other direction. It is a multicentre randomised trial with a 2×2 factorial design, in 120 participants, comparing active electroacupuncture against sham electroacupuncture, and low-dose carbamazepine (300 mg/day) against placebo, over four weeks, with follow-up at 28 weeks (Li et al., J Neurol 2024,PMID 38816482).

The main effects of electroacupuncture and of carbamazepine were significant (p < 0.001), with a significant interaction between the two (p = 0.041). At four weeks, the reduction in the visual analogue scale score was −1.6 for electroacupuncture alone, −0.9 for carbamazepine alone and −3.7 for the combination, each compared with its own control group. Adverse effects related to electroacupuncture were less frequent than those of carbamazepine (10.2 % against 25.4 %). One point deserves noting: at 28 weeks, the effect of carbamazepine alone was no longer distinct from its placebo, whereas that of electroacupuncture persisted.

Three reservations, though. It is a single trial, run in a single country, on a practice, electroacupuncture, that is neither taught nor authorised within the usual scope of physiotherapy practice in France. The result is interesting for understanding the disease; it does not amount to a transferable indication for rehabilitation.

What is left, and what it is worth

The physiotherapist's real place, in five points

Recognise, and say so. This is the main contribution, and it is considerable given the documented delay. A physiotherapist who identifies the presentation and writes a precise letter saves their patient months, and sometimes a tooth.

Do not treat what is not yours to treat. Refusing six weeks of sessions on masseters that have nothing to do with the pain is a professional act, not a surrender.

Explain. Many patients arrive convinced they have “something wrong with their teeth” or a psychiatric illness. Naming the disease, describing its mechanism (an irritated nerve root, fibres that short-circuit, a light touch that recruits pain fibres) and warning that spontaneous remissions occur is fully part of patient education.

Treat what coexists, provided you name it separately. A patient can have trigeminal neuralgia and neck pain, and a genuine temporomandibular disorder. These last two belong to rehabilitation. The mistake is not to treat them: it is to let the patient believe that this treatment will act on the shocks.

Support the impact. That is the subject of the next chapter, and it is the part of the European guideline where the physiotherapist has most to do, even though the document does not name them.

Why is this pain still called the “suicide disease”?

What a national registry says over seventeen years

The phrase had long circulated without solid data. A Swedish national retrospective cohort study has just documented it: 3,531 adults diagnosed between 2001 and 2005, followed for a median of more than seventeen years from the national health and death registers (Ehinger, Kronvall & Uvelius, J Headache Pain 2026,PMID 42001017).

The standardised incidence ratios are significantly raised for cerebrovascular disease (1.4 to 1.8 depending on the analysis), myocardial infarction (1.3), external causes of injury (1.5) and intentional self-harm (2.1). Men had an increased risk of infarction and of ischaemic stroke, women an increased risk of self-harm and of unintentional poisoning. The most striking result comes in two statements that must be quoted together: overall mortality was not increased, but the risk of suicide was more than doubled in women (standardised mortality ratio of 2.5). The authors conclude that these data support calling this disease a genuine “suicide disease”.

What that looks like in a clinic

These figures translate into behaviours that are observed without being named: a patient who speaks little and quietly, who refuses to let a hand come near their face, who arrives masked or with a scarf pulled up for fear of a draught, who cancels when the weather turns cold. These are not personality traits, they are strategies for avoiding a pain that the slightest stimulation sets off.

The specialist clinic series already cited gives the measure of the intensity: 8.17 ± 1.08 out of 10 at the first assessment, with moderate to moderately severe depressive symptoms, moderate to severe anxiety and moderate perceived stress frequently found on the PHQ-9, GAD-7 and PSS-10 scales (PMID 42220298). An American study of patients referred for pre-surgical assessment of facial pain has also examined the link between the dimensions of facial pain and suicidal ideation (Campbell et al., Pain Rep 2026,PMID 42147755).

Two practical consequences. The first: adapt the session by announcing every contact before touching, avoiding the painful territory, not persisting with a palpation that has already set off an attack, accepting that a patient speaks little. The second, more serious: know that the distress is real and do not leave it unanswered. A patient who says they “cannot go on like this” is not exaggerating; the excess risk of self-harm is measured at 2.1. The physiotherapist is not the mental health professional, but they are often the one who sees the patient most often, and the European guideline does explicitly call for psychological support.

What do published case reports teach us?

Published case no. 1: the sensory deficit that changes everything

A man of 59, with a history of hyperlipidaemia, hepatic steatosis and hyperuricaemia, presents with progressive numbness and a sensation of swelling of the left gum and the left side of the face, present for twenty days. He is first seen at a dental practice, where a diagnosis of trigeminal neuritis is made; he is given prednisone, with no effect whatsoever.

Neurological examination finds reduced sensitivity to pinprick and to temperature in the left trigeminal territory, with normal masseter power. MRI shows a diffusion-restricted lesion along the trigeminal sensory pathway, in the left pontine tegmentum: this is a subacute pontine infarct selectively involving the trigeminal sensory fibres, sparing the motor fibres, a presentation the authors call extremely rare and prone to diagnostic error (Wang, Niu & Wang, Front Pain Res 2026,PMID 42317569 ·PMC13272495).

What this case teaches. The objective sensory deficit was there from the start, and it was the most informative sign of the presentation, a likelihood ratio of 20.6 for a secondary cause. A comparative sensory examination, with a wisp of cotton wool and a blunt point, in the three territories and on both sides, takes two minutes and needs no equipment. The patient was treated with corticosteroids at a dental practice while he was having a brainstem stroke.

Published case no. 2: a V3 numbness that was not a neuralgia

A man of 45 has had, for several months, discomfort in the right side of the face and numbness in the mandibular territory (V3), a presentation initially put down to trigeminal neuralgia. Brain MRI reveals a strongly enhancing lesion in the right Meckel's cave, along the nerve. Surgical exploration of the skull base and histological analysis conclude that this is a neurosarcoidosis of the trigeminal nerve, with non-caseating granulomas, no neoplastic or infectious lesion, an isolated cranial nerve lesion that closely mimics a trigeminal schwannoma (Msherghi et al., J Med Case Rep 2026,PMID 41814347 ·PMC13088403).

What this case teaches. Two red flags coexisted: a sensory deficit and an unusually young age. It also teaches a point of vocabulary that has consequences: facial pain with permanent numbness is not trigeminal neuralgia in the ICHD-3 sense, but a painful trigeminal neuropathy (code 13.1.2), a neighbouring and distinct entity whose discovery calls for imaging.

Published case no. 3: the perfectly typical presentation that hid a malformation

A man of 54 presents with severe paroxysmal facial pain, compatible with trigeminal neuralgia. Magnetic resonance imaging reveals an arteriovenous malformation of the trigeminal nerve, an exceptional cause of which this was the first manifestation (Kadiri et al., Radiol Case Rep 2025,PMID 41035745 ·PMC12481911).

What this case teaches. This is the most unsettling of the four, because it carried no obvious red flag. It illustrates exactly the limit measured in chapter 5: the clinical signs are highly specific and barely sensitive, and it is for that reason that the European guideline asks for an MRI in all patients, not only in suspicious cases. A typical presentation is no guarantee: it is an indication for imaging like any other.

Published case no. 4: two years of wandering for a pain in the tongue

A woman of 63 has had, for two years, isolated pain in the left half of the tongue. Glossopharyngeal neuralgia is suspected first; oxcarbazepine brings insufficient relief. Referred to a specialist centre, she has a high-resolution MRI which shows vascular compression of the trigeminal nerve; a glossopharyngeal nerve block rules out the other hypothesis, and microvascular decompression of the trigeminal nerve confirms the diagnosis (Xu et al., Front Pain Res 2026,PMID 42181615 ·PMC13190604).

What this case teaches. The anterior two thirds of the tongue belong to the lingual nerve, a branch of V3: tongue pain can therefore be trigeminal, even if it first suggests the glossopharyngeal nerve. Above all it is a reminder of how long these delays run, two years, and of the fact that they happen even within sound medical pathways.

The fifth case, the one the physiotherapist will meet most often

This is not an isolated case but a profile, documented in a case review from a dental practice: a man of 72 with chronic stabbing pain, intermittent, electric shock-like in quality, sitting in a maxillary tooth and radiating along the trigeminal distribution; carbamazepine alone had not been enough, and the pain did not recur after management combining a psychological intervention with antidepressant and anxiolytic treatment (Tizzoni, Tizzoni & Clerici, F1000Research 2021,PMID 35966965 ·PMC9345266).

What this case teaches. Dental pain that yields to no dental care should make you reconsider the organ. And the psychological dimension is not a fallback diagnosis: it is part of the picture, as the previous chapter put into figures.

How do you apply this in the clinic?

The five questions that do the sorting

They take three minutes, in this order, and most of the time they are enough to point the way.

1. “How long does an attack last? A few seconds, a few minutes, a few hours?” An answer in seconds points to a neuralgia, trigeminal or occipital. An answer in hours rules both out.
2. “Between two attacks, do you have pain?” “Nothing at all” is highly suggestive. Background pain does not rule out the diagnosis, but it points to the form with concomitant continuous pain.
3. “What sets it off? Is a simple light touch enough?” This is the central question. Light touch, speech, tooth brushing, a draught: innocuous stimuli. Prolonged chewing and firm pressure: you are somewhere else.
4. “Show me exactly where”, asking the patient to point with a finger. Forehead, cheekbone, lip, jaw: trigeminal. Back of the head, neck: occipital. A single tooth: a dental cause until proven otherwise.
5. “Have you changed anything in your daily life because of this pain?” No longer shaving, no longer wearing make-up, eating on one side only, avoiding speech: these avoidances are almost pathognomonic and are not met in myofascial pain.

The four examination manoeuvres

1. The comparative sensory examination. A wisp of cotton wool and a blunt point, the three territories, both sides, every time. It is the highest-yield manoeuvre in the whole examination: the objective sensory deficit carries a likelihood ratio of 20.6 for a secondary cause. A normal examination is recorded too: it has value for the doctor who comes next.

2. The search for the trigger zone, once only, with the patient's explicit agreement and after warning them. You brush very lightly, with a fingertip or a wisp of cotton wool, the perioral, nasolabial and chin areas. If the light touch sets off a shock, the ICHD-3 criterion is met. You do not do it again: the refractory period is no reason to test the limit.

3. Palpation of the masticatory muscles and mandibular movement, for the temporomandibular hypothesis. What you are looking for here is to reproduce the patient's usual pain, not simply to find a tender muscle: a masseter that hurts on palpation is commonplace.

4. Upper cervical examination and palpation of the greater occipital nerve, for the occipital hypothesis. Pain reproduced by pressure over the course of the nerve, scalp dysaesthesia: the picture tips over.

The letter, and why its precision changes everything

A generic letter produces a generic referral. Here is what, in a letter, lets a doctor decide immediately on an MRI or a neurological opinion:

Knowing when to stop

It is the hardest decision, because it goes against professional reflex. Three situations require suspending rehabilitation and writing:

The presentation is typical and has never been investigated. No session should come before the medical opinion. This is not a condition where you try six sessions “to see”.

A red flag appears during treatment. A sensory deficit that was not there at the first session, an extension of the territory, the appearance of signs from another cranial nerve: you stop and you refer the same day.

Rehabilitation of a coexisting problem changes nothing about the shocks. If you are treating a genuine temporomandibular disorder and the paroxysms persist unchanged, that confirms they belong to something else, not that the treatment should be intensified.

Running the session, when there is one

When rehabilitation is justified for a coexisting reason, a few adaptations change the patient's experience: announce every contact before touching the face; work on the pain-free side when possible; avoid draughts, blowing air conditioning and manoeuvres that require long stretches of speech; offer to postpone a session during a flare rather than impose it; and never persist with an area that has set off a shock, even “to desensitise it”. There is no data supporting progressive desensitisation of the trigger zone in this condition, and the mechanism, demyelinated fibres that short-circuit, gives no grounds to hope for it.

On the same axis, in our resources