Abstract
Abstract
Background
Paroxysmal hemicrania and hemicrania continua are indometacin-sensitive trigeminal autonomic cephalalgias, a terminology which reflects the predominant distribution of the pain, observable cranial autonomic features and shared pathophysiology. Understanding the latter is limited, both by low prevalence and the intricacies of studying brain function, requiring multimodal techniques to glean insights into such disorders. Similarly obscure is the curious response to indometacin. This review will address what is currently known about pathophysiology, the rationale for the current classification and, features which may confound the diagnosis, such as lack of cranial autonomic symptoms and those which are typically associated with migraine such as nausea, photophobia, phonophobia and aura. Despite these characteristics, a dramatic response to indometacin, which is not seen in migraine nor the other trigeminal autonomic cephalalgias , provides the hallmark of the diagnosis. The main clinical differential for paroxysmal hemicrania is based on temporal pattern and lies between cluster headache and short-lasting-neuralgiform headache attacks with tearing or additional cranial autonomic symptoms. For hemicrania continua it is more challenging as the main differential for which the disorder is often treated is migraine. A prior episodic pattern, often days at a time, and the tendency to exacerbation with analgesics will further deflect from the diagnosis. The relevance of this is that there is little overlap in therapeutics between paroxysmal hemicrania and hemicrania continua and other headache disorders and there are limited effective alternatives to indometacin. The most effective are other non-steroidal anti-inflammatory drugs including the newer COX-II inhibitors. Even though early reports suggest that a higher indometacin dose-requirement may herald a secondary precipitating pathology, this does not seem to be the case, with syndrome and response to treatment being similar with the primary disorder. In this context imaging of new onset paroxysmal hemicrania or hemicrania continua and implication of the results will be discussed as will alternative treatment options.
Introduction
Paroxysmal hemicrania (PH) and hemicrania continua (HC) fall under the terminology of the trigeminal autonomic cephalalgias (TACs), together with cluster headache and short-lasting unilateral headache (SUNHA) (1). The core features comprise unilateral, primarily ophthalmic trigeminal distribution pain and a variable combination of autonomic features, reflecting an increase in parasympathetic outflow and a sympathetic deficit (Table 1). Cluster headache is the most common with a lifetime prevalence of 124 per 100,000 from a meta-analysis of population-based epidemiological studies (2). The disorders are linked phenotypically and show shared pathophysiological characteristics. However, treatment responses remain distinct. PH and HC are peculiar in having a dramatic response to indometacin. The limitation of interpreting published cohorts is that while the syndromes may be consistent with PH and HC, not all were trialled with indometacin (3–6). This review will address what is known about pathophysiology, diagnostic and potentially confounding features, differentiation from other primary headaches and secondary cases and, treatment.
International classification for paroxysmal hemicrania and hemicrania continua (1).
Pathophysiology
Pathophysiology of PH and HC will be addressed together based on shared traits observed in the TACs and in PH and HC, treatment response.
There is likely to be a predisposition to a phenotype of primary headache. There are 181 independent single nucleotide polymorphisms (SNP) robustly associated with migraine (7). In a genome-wide association study (GWAS) meta-analysis of cluster headache the estimated SNP-based heritability of cluster headache was reported to be 14.5% (8). Nine independent signals in seven genomewide significant loci were identified in the primary meta-analysis and one additional locus in a trans-ethnic meta-analysis. There are as yet a handful of reports of familial PH and HC; increasing numbers of cases reported may allow further insights into heritability.
The terminology of a TAC serves to highlight the prominent autonomic symptoms observed where there is a bilateral representation despite the unilateral predominance of symptoms, ictally and interictally (9,10). Early observations in PH showed an increase in ipsilateral forehead sweating, tearing and nasal secretions in neck-movement precipitated attacks. The sweating occurred prior to the onset of the pain, while atropine blocked the autonomic symptoms without affecting the pain. This dissociation is seen in spontaneous attacks of PH where only autonomic features, and not pain, occurred with the same temporal pattern of usual attacks (11). Aniscoria, supporting a sympathetic deficit, is observed in both HC and PH, ictally, interictally and in response to pharmacological challenges ipsilateral to the side of the attacks; reversal is seen following treatment with indometacin (12–14). The role of the central nervous system over a more peripheral and or vascular-dominant pathway is increasingly supported. The observations of periodicity and alteration of hypothalamic-determined functions, more prominent in cluster headache than PH or HC (15), is consistent with the preferential involvement of the hypothalamus. Initially observed in functional imaging studies in cluster headache (15), activation of the posterior hypothalamus was subsequently observed during attacks of PH (16) and HC (17). The clinical overlap in HC with migraine, is reflected in activation of both the hypothalamus and the pons, (17–19) the latter as seen in migraine (18,20). Although initially attributed to being the drivers of these disorders, more recent imaging work in migraine shows a more complex interplay between the posterior hypothalamus and the dorsal rostral pons (21). What the precise role of the hypothalamus is in disrupting the anti-nociceptive-nociceptive equilibrium is yet to be determined.
The putative neuroanatomical pathway in the TACs, incorporates the trigeminocervical nucleus, its peripheral and central neural connections, the latter at the levels of the brainstem (including the hypothalamus), thalamus and cortex (22). Pre-ganglionic parasympathetic efferent fibres originate from the superior salivatory nucleus in the pontine tegmentum. The fibres which travel with the facial nerve with pre-ganglionic sympathetic fibres, then synapse in the sphenopalatine ganglion. Postganglionic fibres innervate the lacrimal and salivary glands and the cerebral vessels (23). A trigeminal-autonomic reflex has been proposed to explain the co-activation of the trigeminocervical-mediated pain pathway and the accompanying autonomic phenomena. In a pre-clinical model superior salivatory nucleus stimulation resulted in both neuronal trigeminocervical firing and an alteration of facial blood flow at the level of the lacrimal gland/duct as a measure of cranial parasympathetic activation (24). Treatment with indomethacin inhibited both the trigeminocervical and autonomic activation, which was also greater than that observed with naproxen.
The mechanism of therapeutic responsiveness of indometacin remains elusive. Regularly taken indometacin behaves in a preventative manner and not as an acute analgesic, as is the case with other non-steroidal anti-inflammatory drugs (NSAIDs) and indomethacin used to treat other disorders such as arthritis. There have been a number of neurochemical targets which have been investigated (25). Indometacin is a cyclo-oxygenase (COX)-I inhibitor (26). Although some individuals respond to other NSAIDs with COX-I inhibitory properties, the response to indometacin remains the most consistent. The selective COX-II inhibitors can also be effective (27–29), thus COX-inhibition is unlikely to answer the selective indometacin-responsiveness. Calcitonin-gene-related peptide (CGRP), which has been shown to be elevated during an attack in PH, normalised following treatment with indometacin (30). Consistent with this is a recent report of a case of HC responding best to NSAIDs, including indometacin but complicated by intolerance. The individual responded completely to galcanezumab, the humanised monoclonal antibody that binds to the calcitonin gene-related peptide ligand. There was a consequent reduction in CGRP levels as compared to pre-treatment level (31). Nitric oxide (NO) donors can precipitate attacks of migraine and cluster headache in known sufferers (32,33). In another preclinical model, treatment with indomethacin was able to inhibit NO-induced dural vasodilatation which was not observed with naproxen nor ibuprofen. However, none were able to inhibit CGRP-induced vasodilatation (34). For the time being indometacin remains the mainstay of drug management but its mechanism of action unclear.
Paroxysmal hemicrania
Paroxysmal hemicrania was first reported in 1974 by Sjaastad and Dale (35). Two female patients presented with a 10–15-year history of daily headache with characteristics similar to cluster headache, but refractory to medical treatments (35). A trial of non-steroidal anti-inflammatory drugs revealed a dramatic response to indometacin within 24 hours; this was subsequently confirmed in a placebo-controlled manner. In one of the patients, indometacin rendered her pain-free, but with the same regularity of previous exacerbations, she experienced an irregular heartbeat, sweating, a veil over the eye on the affected side and a peculiar feeling in the left arm. The disorder entered the International Classification for Headache Disorders (ICHD) in 1988, defined by phenotype and response to indometacin. The initial reports were of a chronic pattern of pain without remission. Recognition of an episodic form with recurrent bouts and remissions led to two subdivisions in the 2004 classification (36).
Epidemiology
A response to indometacin precludes obtaining reliable population-based data on prevalence. Therefore, the best estimates are from clinic-based populations. Data from consecutive patients with strictly unilateral headache were collected from a total of 528 patients seen in a headache clinic in Spain; 38 were diagnosed with cluster headache (thus 7% of the total cohort) and four with PH (0.75% of the total cohort) (37).
Table 2 summarises the published cohorts which provide the most complete information on phenotyping. The median age of onset is in the fourth decade, but the disorder can manifest in the first decade (38) and the ninth (39). Although pooling of earlier reports suggested a female predominance more recent cohorts suggest both sexes are equally affected.
Paroxysmal hemicrania cohorts.
V2, second division of the trigeminal nerve; CAS, cranial autonomic symptoms.
a13% < 4 attacks/day for more than half the time.
bRemission between bouts < one month.
cAll previously indometacin responders (stopped due to AEs).
dBased on available data from a retrospective collection. In 25% there was not a consistent response to indometacin.
ePredictable time of onset of attacks for unspecified periods of time.
A family history of PH has been reported but the low prevalence limits further understanding of heritability (40).
Clinical characteristics
Laterality
The pain is strictly unilateral but not side locked. In up to 15% either side can be affected. There are reports of cases with bilateral pain (39,41,42).
Location and character of pain
The pain primarily affects the orbital, frontal and temporal regions, thus, 1st division trigeminal distribution, but can affect the back of the head, parietal and occipital regions, thus C2, the cheek (2nd division trigeminal distribution) jaw (3rd division trigeminal distribution), ear (9th and 10th cranial distribution), neck (C2), shoulder (C3) and in one cohort (43) radiation to the upper limb, as observed in Sjaastad and Dale’s original report (35). The involvement of the arm is consistent with the subclinical supratentorial representation seen in migraine (44). In Ziegeler and May’s cohort of patients with head and facial pain, the facial pain being areas supplied by the 2nd and 3rd branches of trigeminal innervation, there were nine of 20 (45%) individuals with PH or probable PH with facial pain involvement. In five of the nine (56%) the pain was confined only to the second and third division trigeminally innervated regions (6).
The character of the pain can vary (Table 2) to an extent which is not diagnostically distinct. Thus, the pain can be throbbing, sharp, claw-like, dull and pressing amongst other descriptions.
Interictal pain
Ipsilateral interictal pain can exist between the discrete attacks, as is seen in cluster headache. A small cohort reported those with interictal pain to have a longer disease duration and higher dose requirement of indometacin (45). The interictal pain can be both intermittent and persistent between the discrete attacks. A similar observation has been made in cluster headache (46). In the cohort of Paliwal et al. three individuals had a preceding history of headache more consistent with HC with duration of constant pain for three to six months, in two with remissions of 18 months and three years between the two headache types (5). In the one case there was spontaneous remission, and another responded to indometacin. In the third case there was no interval between the two syndromes and the HC-like headache responded to NSAIDs (5).
Autonomic features
There are observable cranial autonomic symptoms (CAS), most commonly lacrimation (62–87%), conjunctival injection (25–81.5%), rhinorrhoea (7.4–58%) and nasal congestion (41–55%). Less frequently observed are ptosis (12.5–55%), eye-lid swelling (18–42%), facial sweating (55%), flushing (53–55%) and miosis (6–18%) (see Table 2 for references). As in the index case more generalised autonomic manifestations are also reported such as generalised sweating (39). Although the CAS are usually ipsilateral to the attacks, they can be bilateral (39), dissociated with the pain (in up to a third of cases) or absent in about 10% of cases.
Other features
In 39–52% there is motion sensitivity and restless and or agitation in up to 80% (range 14.8–80%). Only in one report are both characteristics reported; 21% had motion sensitivity and restlessness/agitation (39), whether in different or the same attacks is not given. Aural fullness was reported in 26% in one cohort (43).
Features typically associated with migraine are reported in at least a quarter of cases, thus not the same frequencies seen in migraine (47). Most frequently observed are photophobia (21–74%) and phonophobia (7–80%). In one cohort laterality of photophobia was reported in 40% of cases and in the same report 39% experienced osmophobia (43).
Aura has been reported with PH (48–50), visual, sensory and motor, tinnitus (39) and in one cohort 12% reported vertigo and or dizziness (4).
Duration and frequency
The duration of attacks ranges between 10 seconds up to four hours. The classification has defined duration 2–30 minutes which includes the means across the cohorts (8–20 minutes). The attack frequency in the previous classification was ‘>5 day for more than half the time’ and would encompass the range of observed attack frequencies. The current revision defines frequency of >5 a day, which is likely to represent the mean of the observed frequencies.
Temporal pattern
The majority suffer chronically from onset, whereby there are no remissions of at least three months in one year. Although the majority will have a new daily persistent onset, 10–27% evolve from an initially relapsing remitting pattern. In 9–25% there is a relapsing and remitting course.
A circadian and or a circannual rhythmicity have been observed. More consistently reported is nocturnal preponderance (typically waking the individual from sleep) in 6–65%.
Precipitating factors
In the cohort of Cittadini et al. (43) triggers perceived included stress or relaxation from stress (26%), exercise (23%), alcohol and neck movement (19%) sneezing (13%), a warm environment, cold weather, strong odours, bending down and coughing (16% each), tiredness (10%) straining (10%), in relation to menstruation (6% also reported in the cohort of Antonaci et al. (39)) and lifting (6%). Also observed are cutaneous triggers, weather change, alteration of sleep pattern, perimenstrual or menstrual period. Glyceryl trinitrate triggered a typical attack in one of four patients challenged. In the cohort of Boes and Dodick in eight of 21 patients, alcohol triggered an attack and in five of 10, neck movements (4). In Antonaci et al.’s report of 84 cases 10% reported triggered attacks by neck movement and alcohol in 7% (39).
Non-ICHD paroxysmal hemicrania
These individuals fall into four groups: 1) those who respond to indometacin but do not fulfil all the ICHD criteria for PH 2) those who are phenotypically consistent but do not have an ‘absolute’ response to indometacin but, nevertheless, indometacin provides the best response compared to other treatments 3) those who are non-responsive despite being phenotypically ICHD consistent and 4) those who have not been able to tolerate indometacin to an adequate dose to confirm the diagnosis.
In the first group, those who pose diagnostic difficulties are those whose frequency and duration of attacks may fall outside the criteria and where there is an absence of observable CAS. In this group the absolute response to indometacin will define the diagnosis. This is more pertinent in those without CAS; although the latter may be absent, there is little syndromic overlap with a non-TAC, thus an absence of CAS should not steer the clinician away from a trial of indometacin, as is seen in the cohorts presented.
In those with a partial response to indometacin, an alternative treatment reported to have an effect in PH could be tried instead or in addition (see section on treatment) but to date there are no reports where one of these alternative treatments has ultimately been reported to be more effective than indometacin at adequate dose.
The phenotypically consistent but not indometacin–responsive is a more frustrating group. The clinical goal of a classification for primary headaches is a syndrome which is tightly coupled to a reliable treatment response. The question is whether this group should be classified independently in the ICHD. At present there is no alternative group other than 14.1 ‘Headache not classified elsewhere’ which albeit inadvertent, will relegate the group into a neglected hotchpotch of syndromes. For the purpose of research, a separate subgroup within the TACs would help address whether the disorder is independent or whether there are more complex factors which render these individuals non-responsive to indometacin. It would also allow much cleaner phenotyping.
Last is the individual who has not been able to tolerate an adequate dose of indometacin for therapeutic effect and hence receive confirmation of the diagnosis. If they respond well to treatment for a flanking TAC (see Table 1) it would seem reasonable to re-allocate the disorder diagnostically.
Secondary paroxysmal hemicrania
There are very few cases where one can make a presumption that the evolution of PH is associated with a secondary pathology. There is a report of post-traumatic PH with aura, with onset within three days of a mild head injury, with subsequent response to indometacin (49). Another is a case of attacks precipitated by use of the phosphodiesterase inhibitor tadalafil and sildenafil, ceasing on stopping either drug, but responding to indometacin while symptomatic (51). A number of cases have relied on surgical intervention for a lesion found after imaging for the headache and, for which surgical intervention resulted in resolution of the headache (52). The problem with these cases is the issue of improvement seen in relation to surgical intervention per se unrelated to the nature of the lesions (53). Moreover, there is limited follow-up of surgical cases. Where patients are reported as part of a cohort there are surgical outcomes where intervention resulted in transient or no improvement (43,54). In the cohorts presented here, situations where the imaging abnormalities have not required intervention these have been termed incidental (4,43), for example small vessel disease. Other than cases as per the examples given above, interpretation of causality seems reliant on interpretation by individual clinicians and/or authors.
The numbers of secondary PH reported are small, but there is no clear differentiation between those with primary versus secondary PH, in terms of clinical syndrome or responses to treatment. Moreover, there is no obvious predominance of anatomical site nor pathology of the precipitating disorder or event (55).
Comorbid primary headaches
From the reported cohorts a past history of migraine is reported in 5%–58%. In the 84 cases of Antonaci and Staastad, where 5% had migraine, there was one patient with cluster headache (39). Other primary headaches have also been reported with PH, either occurring simultaneously or independently and include cough headache (43,56,57), trigeminal neuralgia (so-called PH-Tic) (58,59) and SUNHA. In these cases the PH responded to indometacin, the trigeminal neuralgia responded to carbamazepine and the SUNHA responded to lamotrigine (27). The responses for each syndrome were independent, in that treatment of one did not have an impact on any comorbid headache syndrome. Given the rarity of these disorders the comorbidity is greater than would be expected.
Differential diagnoses
The main differential diagnoses will lie between cluster headache and short-lasting unilateral neuralgiform headache SUNHA. The key features being the shorter duration and greater frequency of attacks in SUNHA, in which attacks are more likely to be triggered than in PH.
Treatment
The ‘Indotest’ was reported in 1998 in both PH and HC to rapidly ascertain response to indometacin using 50–100 mg intramuscular indometacin (60). The mean interval between attacks untreated is shorter than with treatment; there is also a refractory period after treatment when the next expected attack did not occur.
Oral treatment doses up to 300 mg have been used, with most patients showing a response in the first week (Table 2). In some it can take up to two weeks to see a response. Recurrence on withdrawal is seen within 2–10 days.
In a critical appraisal of treatments (of at least N = 5), a mean of 89% had a complete response to indometacin (N = 168; mean 97 mg ± SD 39) with a dropout rate of 27%, related to adverse events (61). Acutely only 4% responded to subcutaneous sumatriptan and none to high flow oxygen. Other responses were piroxicam 40% (N = 5; 36 mg ± 9) topiramate 42% (N = 12; 172 mg ± 75), and verapamil 17% (N = 30; 248 mg ± 87). There was no response seen with carbamazepine, amitriptyline, supra-orbital and greater-occipital nerve blocks. Also reported for at least five cases per drug group are responses to the COX-II inhibitors, refocoxib, celecoxib and etoricoxib (27,28,62–64).
In a paediatric series of eight individuals, at follow-up (3.1 to 10.7 years) six were headache-free and off treatment for the last 5.4 ± 3.4 years. The mean treatment time was 2.2 ± 1.9 years. Two continued on treatment at the time of reporting, 4.5 and 4.9 years, respectively. Both unsuccessfully tried to reduce the indometacin treatment, two and six times respectively, but remained pain free on treatment (65). In 10 adults with PH, four were able to reduce the indometacin dose by 52% and remained on treatment for 2.2 ± 1.9 years. Six remained on treatment for 3.1 ± 1.4 years (dose range 25–75 mg). In some the maintenance dose can be reduced to 12.5 mg alternate days (39).
Hemicrania continua
Hemicrania continua (HC) was first recognised in 1981 by Medina and Diamond (66), described as a ‘cluster headache variant’. It was more clearly defined in 1983 as a strictly unilateral headache with constant background pain and superimposed exacerbations with autonomic features, responding ‘absolutely’ to indometacin (67).
Table 3 gives details of cohorts with the most comprehensive information about the patients with HC and from which subsequent figures have been obtained; any information additional to the tables is referenced accordingly within the body of the manuscript. The original reports have been of a new daily persistent onset. Cases of an episodic pattern were subsequently reported (68–70). Both episodic evolving to chronic HC and chronic evolving to an episodic pattern (71,72) have been reported, with typical symptoms lasting from days to months at a time, interspersed with clear remissions. Of note, in those experiencing symptomatic days, these have not necessarily been consecutive days, thus some individuals have experienced three days a week with attacks. At least 60% of cases are unremitting from onset while about 20% evolve from a remitting pattern.
Hemicrania continua cohorts (1).
HC-A, hemicrania continua without autonomic features; HC-I, hemicrania continua phenotype but incomplete response to indometacin (at least 50%); Mo, months; yrs, years; CAS, cranial autonomic symptoms; AE, adverse effects; NS, non specified; CHC, chronic hemicrania continua.
Retrospective
Prospective
Epidemiology
As with PH, there are no population-based prevalence data as the diagnosis is reliant upon a therapeutic response to indometacin. From a meta-analysis of adults attending headache clinics, an estimate of relative frequencies was found to be 1.8% (73).
From a series of published cohorts, the male to female ratio varies between 1:1 and 1:2.4 (3), veering more towards the former (Table 3), thus equal in men and women. The mean age of onset lies between the 4th and 5th decades, but the disorder has also been reported in the 1st (74) and 8th decades (75).
There are two reports of familial HC, both in a mother and a daughter (76,77).
Clinical characteristics
Laterality
The pain is strictly unilateral in at least 90% of cases, right or left equally. Less than 10% can experience attacks on one side or the other at different times (78,79) and even less commonly attacks can be bilateral (80).
Location and character of pain
As seen in cluster headache and PH the pain most often affects the orbital, frontal and temporal regions but can affect regions outside the distribution of the 1st trigeminal division. In a third of cases the pain can affect the back of the head (parietal and occipital regions, thus C2) and upper and lower teeth cheek (V2), less frequently the jaw (V3), ear (IX/X), neck (C2), shoulder (C3) and upper limb.
The character of the pain most often experienced is a non-throbbing constant, dull, pressure pain or ache, and or throbbing, the latter often associated with exacerbations (3,79). About 40% can experience a stabbing component.
Autonomic features
The most common autonomic features are lacrimation (63–77.1%) followed by conjunctival injection (11.7–77%), nasal blockage (20.6–63%) and rhinorrhoea (11.7–63%); of note the lowest frequencies were all from the same cohort (3). Between 32–43% experience a sandy itching sensation in the eye and aural fullness (a sensation of fullness in the ear and muffled sound) was reported in three (81), 13 (79) and 19% (78) of cases in three cohorts. The autonomic features are most consistently reported ipsilateral to the pain, but have also been reported to be bilateral (78) and occurring with the same pattern of attacks but without pain (11). Peres et al. reported at least one autonomic symptom in 26% in a summary of 41 cases reported in the literature (3).
Other features
Typically, migrainous features can be experienced with nausea in up to 58%, with vomiting less common in 15–19%, except one cohort of 34 patients where vomiting occurred in 34.8%. The range of photophobia (26–79%) and phonophobia (16–74%) varies across the cohorts. The cohort of Cittadini and Goadsby noted that over half of the cases experiencing phonophobia, the perception was unilateral and similarly so for photophobia. Between 53–81% reported restlessness and/or agitation with two of the cohorts citing aggressive behaviour in 19 and 28% (78). However, only Cittadini and Goadsby asked about motion sensitivity which was experienced in 69%; 69% also experienced restlessness, thus the assumptions could be made that although restless, individuals can also have motion sensitivity, unlike migraine (82) and cluster headache (83). Whether within or across attacks was not mentioned.
Typical aura can occur in HC (50,84,85) and one of the original cases reported vertigo associated with the attacks (66).
Duration and temporal pattern
Duration of exacerbations can be less than five minutes (19–32%), in one cohort 5–30 minutes in 42% but most commonly an hour to a day in 77–90% and more than a day in 19–31%. The number of exacerbations a day was between 1–5 in 42–46% individuals across three of the cohorts and one where 91% had exacerbations with this frequency. A circadian and or circannual pattern of symptoms has been reported in a handful of patients (70,78) but is not a prominent characteristic, while a nocturnal preponderance is reported in 26–63%. Although up to 63% can have nocturnal exacerbations there is no indication that these occur at predictable times. The duration of symptoms is reported to be between three months and 49 years; the mean is three to four years.
Precipitating factors
Reported triggers (>9%) include stress, alcohol, irregular sleep, bright lights, exercise, a warm environment, neck movement and skipping meal (78).
Non-ICHD hemicrania continua
In Cortijo et al.’s cohort 23% of individuals did not have cranial autonomic symptoms (75). The 62 patients of Prakash and Golwala (81) were divided into three groups, those fulfilling all criteria for HC (N = 30), those without cranial autonomic symptoms but indometacin-sensitive (N = 13), and those who fulfilled criteria for HC but had only a partial response to HC (N = 19) (Table 3). Marmura et al. reported on 43 patients with ICHD consistent HC and 122 who were phenotypically identical but did not respond to indometacin and seven who did respond but not completely (87). CAS did not predict indometacin-responsiveness. Given that part of the goal of a classification is to align phenotype with best treatment response, it would seem more prudent to subclassify HC with or without CAS for research purposes but for the purpose of clinical practice to align those who are indometacin responsive in the same group.
In those with a partial response to indometacin, many will have tried other preventative treatments for migraine or a TAC, yet the best response is with indometacin. It would be reasonable to label the diagnosis within a third subgroup for HC where there is not an ‘absolute’ response, but indometacin still provides the best response. This allows further research into other factors which may dictate the partial responsiveness in a more clearly defined group.
A fourth subgroup is for those clinically consistent but not indometacin-responsive. There is more relevance to subtype this group scientifically on par with non-indometacin-responsive PH. The diagnosis is likely to need additional characteristics to differentiate it as a TAC rather than migraine with features such as circadian and circannual symptom propensity and restlessness.
There are then those who have not tried indometacin, or those who have not tried an adequate dose because of adverse effects. This group should not be included in publications of HC, as the pertinent information needed to be relevant to the disorder, is not available. In this group some may respond to another NSAID (86) better than any other preventative treatment which supports the diagnosis. There is a clear difference in the role of indometacin as a preventative rather than an acute treatment. Taken daily, regardless of whether the individual is in pain at the time, it prevents the development of the pain, which a drug behaving as a typical analgesic has not been shown to do. This group therefore must share some fundamental pathophysiological characteristics with ICHD-HC.
Where addressed, at least half of the patients reported on also fulfil diagnostic criteria for migraine (81,87). If not sensitive to indometacin or another NSAID, these individuals are usually diagnosed and treated as chronic migraine by default. The question is whether these are indeed migraine or still have an underlying TAC pathophysiology.
Secondary hemicrania continua
As discussed for PH and relevant to all secondary headaches, often an imaging abnormality is found to be present when imaging for the headache alone (88). Treatment may result in resolution of the headache, but this does not automatically confirm a correlation. Those who improve are often heralded in case reports but those who do not are less likely to be reported (89).
The association of a secondary headache is more probable when the onset of the headache is in close temporal association with the pathology. For post-traumatic headache the ICHD defines onset within seven days, but this is not consistent for other secondary pathologies where ‘a close temporal relation’ remains undefined. Prakash et al. (90) reported on two cases of ophthalmic zoster reactivation and simultaneous onset of headache which persisted once the zoster settled. Both cases subsequently responded to indometacin (90). In a group of event-related cases of HC, 12 followed head injury, post-stroke, post-partum or a surgical procedure. Two did not trial indometacin. Nine responded to indometacin and one had a partial response. There were no clear differences with primary HC phenotypically nor in relation to indometacin-sensitivity and dose requirements (88).
Co-morbid primary headaches
In the original 1981 report by Medina and Diamond (66), 20 of 54 patients described additional multiple jabs lasting seconds at a time and consistent with primary stabbing headache. One also reported exacerbation with valsalva, all the headaches responded to indometacin. Of Cittadini and Goadsby’s cohort 36% described stabbing pains (78). The population-based and headache-clinic-based prevalence of primary stabbing headache has varied, the former ranging from 2–35% while the latter is more consistent between 11–13% across three headache clinics (Turkey, Taiwan and Korea) and 5% in one Spanish headache clinic. The prevalence is considered to be higher at 40% with other primary headaches such as migraine (91).
The prevalence of migraine that was reported in the cohort of Cittadini, 61%, is much higher than population prevalence, albeit from a small cohort. There are also reports of multiple TACs in the same individual (92,93). Cosentino et al. described a 45-year-old man who sequentially developed cluster headache, SUNHA and HC, occurring at different times and each responding to the most appropriate treatments for each disorder (94).
Paliwal et al. described five patients with preceding ipsilateral paroxysmal headache attacks with autonomic symptoms lasting for three to six months. One transformed from paroxysmal attacks into continuous pain consistent with HC. Four had spontaneous resolution of the paroxysmal attacks for six to 18 months prior to the new onset of HC consistent headache. None trialled indometacin, thus confirmation of the two comorbid disorders could not be made (5).
Differential diagnoses
The main differential diagnoses for HC will be migraine and at the other end of the spectrum PH and cluster headache with interparoxysmal background pain. The differentiation from chronic migraine is the response to indometacin. If co-existent with PH, both respond to indometacin. With cluster headache if the interparoxysmal pain is in fact HC, then this background pain should be indometacin-responsive, while the typical attacks will respond to treatments for cluster headache.
Treatment
Indometacin dose ranges are reported up to 500 mg, with the mean not greater than 200 mg in most of the cohorts, as shown in Table 3. Although a response is seen in about two thirds of cases within 14 days, in the remainder it can take up to eight weeks (95). On cessation symptoms tends to recur within 48 hours. Thus, missing doses will be informative about whether the disorder remains active. The most common side effects are gastrointestinal and dizziness.
Acute-relief medication overuse is reported in a number of cohorts. Although in those where the acute-relief medication was withdrawn, there appeared to be no change in symptoms, there are reports of exacerbation of the disorder in association with acute-relief medication overuse (96). In the cohort of Cittadini and Goadsby, 72% fulfilled the criteria for medication overuse at some point in time; of this group 86% in whom medication withdrawal was attempted there was no change in phenotype (78). Of 39 individuals 32 were indometacin responders; it is not detailed what proportion responded to indometacin while overusing acute-relief medication. It is also not apparent from the reports of non-responders, whether there was concomitant overuse and if so, whether there was retrial of indometacin in the absence of medication overuse.
As reported for PH the Indotest can be used to rapidly ascertain whether an individual is indometacin-sensitive using IM indometacin 50–100 mg (60). Indometacin itself can precipitate headache (78); Cittadini and Goadsby reported a bilateral throbbing headache induced by indometacin which settled on cessation of the drug; it was noted that this patient had a personal history for migraine. Jurgens et al. reported a similar case but without a prior or family history of migraine (97).
In individuals observed in the longer term, seven of 16 were able to reduce their indometacin dose by 60% with treatment time 4.8 ± 3.7 years. Eight remained on treatment over 4.3–3.5 years (dose range 50–150 mg) (98).
In those not tolerant of indometacin, most frequently reported benefit is with other non-steroidal anti-inflammatory drugs (NSAIDs) including the COX-II inhibitors refecoxib, celecoxib and etoricoxib (27,64). Refecoxib was withdrawn because of an increased risk of cardiovascular disease. A meta-analysis suggested that COX-II selectivity may not play a role in the cardiovascular risk of NSAIDs. Refecoxib was the only drug to demonstrate harm and skewed the data of the COX-2 selective group (99).
In a critical appraisal of treatments (each of at least N = 5), a complete response was seen in a mean of 95% with indometacin (N = 159; dose range [DR] 25–225 mg), celecoxib 73% (N = 11; DR 200–800 mg), piroxicam 71% (N = 7; DR 20–40 mg), topiramate 62% (N = 13; DR 50–300 mg), gabapentin 46% (N = 13; DR 600–3600 mg), OnabotulinumtoxinA 33% (N = 12; DR 100–185 u), melatonin 29% (N = 17; DR 3–30 mg)), supra-orbital nerve blocks 29% (N = 17; local anaesthetic alone or with steroid), occipital nerve stimulation 21%, (N = 14) and greater occipital nerve block 7% (N = 15; local anaesthetic alone or with steroid). There was no response seen to sumatriptan, high flow oxygen and verapamil. In the indometacin group 55% stopped treatment from adverse effects (61). In a more recent report there were four responders to Etoricoxib, DR 30–120 mg (27).
Summary
PH and HC have shared features observed in the TACs, of strictly unilateral pain primarily in the first division of the trigeminal nerve, which can swap sides and rarely can be bilateral. There are accompanying CAS, primarily of a parasympathetic outflow, and sympathetic deficit. However, the absence of observable CAS does not preclude the diagnosis and response to indometacin. The presence of more typical symptoms seen in migraine, such as nausea, photophobia, phonophobia and aura, should not deter the clinician away from the diagnoses. Although experienced across the TACs, such features are seen less consistently than in migraine.
In PH the diagnosis will be given by the shorter duration and higher frequency of attacks which is not seen in adults with untreated migraine attacks. The alternative differential of cluster headache will be addressed by a trial of indometacin; moreover, the attacks of PH do not respond to acute treatment such as subcutaneous sumatriptan nor high flow oxygen, as seen in cluster headache. There can be a more persistent background interictal pain in PH which may be more resistant to treatment.
HC has more overlap features with migraine, including the tendency for exacerbation with overuse of acute-relief medication. The distinction lies in the response to indometacin. Thus, any individual with strictly unilateral unrelenting hemicranial pain, even if there are migrainous features or a lack of CAS, should have a trial of indometacin. Additional characteristics such as restlessness and or agitation will favour HC.
The response to indometacin can be assessed within a few weeks, at most a few months. The risk of not trialling indometacin is that of repeating the failures reported in the historical literature (39,100); individuals spend years, sometimes decades, unsuccessfully trialling treatments used in migraine and cluster headache until an enlightened individual or serendipity reveal an incomparable response to indometacin (or another effective NSAID). For both operational and research purposes, subtypes of PH and HC where there is partial or no response to indometacin would benefit from being defined within the ICHD. Pathophysiological differences can then be studied in cleaner phenotypic cohorts to understand the limitation of indometacin responsiveness and the development of more effective therapeutic interventions.
Precipitated cases of PH and HC are clinically and therapeutically similar to the primary disorder with no differentiating characteristics within the headache syndrome. The most reliable indicators will be the additional features which accompany the secondary pathology. The more difficult question is whether the headache syndrome alone is more likely to indicate a secondary pathology requiring intervention, in comparison to an asymptomatic population. The low prevalence, clinical and publication bias continue to incumber a more pragmatic and medicolegally sound solution.
PH and HC can occur comorbidly with other primary headaches, more than would be expected based upon population prevalence suggesting a possible predisposition to primary headaches. Each syndrome seems to respond to the treatment most effective for that disorder.
Indometacin behaves in a preventative manner in the treatment of PH and HC. The main limitations are the gastrointestinal adverse effects. Alternatives include other NSAIDs including the COX-II inhibitors, which have less adverse effects; the cardiovascular effects of the COX-II inhibitors were mainly seen with Refecoxib which has now been withdrawn.
PH and HC can be precipitated by alternative pathologies. This is clearer when event-related rather than when found incidentally on imaging for the headache. The low prevalence does not yet allow a judgement to be made as to whether each syndrome has a greater probably of heralding an alternative pathology or, like migraine and tension-type headache, whether in fact the probability of picking an alternative pathology when presenting in isolation (thus pain syndrome only) is the same as the population prevalence (101). What is evident is that the phenotype and treatment responses are the same as the primary syndromes and remain independent to any presumptive precipitating pathology. Thus, each should be treated independently, and one cannot make the assumption that the course of each disorder is intertwined with the other.
Clinical implications
The response to indometacin does not have to be ‘absolute’.
Phenotypical PH and HC not responsive to indometacin should be accommodated in a subgroup within the ICHD and published separately for research purposes.
Absence of cranial autonomic symptoms and presence of migrainous symptoms do not preclude a diagnosis paroxysmal hemicrania nor hemicrania continua.
Primary and secondary paroxysmal hemicrania and hemicrania continua are similar phenotypically and in response to indometacin.
Hemicrania continua can be exacerbated by acute-relief medication overuse.
