Therapeutical approaches to paroxysmal hemicrania, hemicrania continua and short lasting unilateral neuralgiform headache attacks: a critical appraisal

A MEDLINE search using the electronic data-base pubmed has been performed to check all articles dealing with the treatment of primary HC, PH, SUNCT and SUNA form the 1st of January 1989 (the first complete year in which the first International Headache Society classification was available) onwards. All articles types were considered and non-English written ones were excluded. The research was performed using the following terms: “((paroxysmal hemicrania) AND (“1989/01/01”[Date - Publication]: “3000”[Date - Publication])) AND English[Language]” for PH, “((hemicrania continua) AND (“1989/01/01”[Date - Publication]: “3000”[Date - Publication])) AND English[Language]” for HC, “((short lasting neuralgiform headache attacks) AND (“1989/01/01”[Date - Publication]: “3000”[Date - Publication])) AND English[Language]” for SUNCT and SUNA. Short lasting unilateral neuralgiform headache attack was treated as one entity, not differentiating between short lasting neuralgiform headache attacks with conjunctival injection and tearing (SUNCT) and short lasting neuralgiform headache attacks with autonomic signs (SUNA). A few articles cited in the references of the above-mentioned ones were cited even though they were not present in pubmed, but were found in SCOPUS and EMBASE.

Altogether, 691 articles were found of which 290 articles for HC, 250 for PH and 151 for short lasting unilateral neuralgiform headache attacks. Cited articles should fulfill the ICHD-III beta guide-lines for TACs diagnosis, not deal with a symptomatic case and correctly state treatment. Reviews were considered only if new cases were included. For HC, 230 articles were excluded: 67 summarized results from other studies without adding any new case, 138 didn’t deal with HC therapy and 24 referred to symptomatic cases. For PH, 195 articles were excluded: 67 reported and summarized only the results of different works, 90 didn’t consider PH therapy or described it unsatisfactorily, 29 referred to symptomatic PH and 9 didn’t fulfill all ICHD-III diagnostic criteria, making a diagnosis of “probable PH”. For SUNCT and SUNA 95 articles were excluded: 60 were reviews, 20 of them didn’t deal with SUNCT or SUNA therapy or reported it unsatisfactorily, 11 reported symptomatic cases and 4 didn’t full-filled all diagnostic criteria. Steps followed for article selection are summarized in Fig. 1. For every article, each patient was analyzed and only those treatments correctly stated in terms of regimen and response were considered. If a patient took a drug in different dosages or underwent a non-pharmacological procedure following different regimens, only the one giving the maximum effect was considered. Every patient was classified as a responder if he/she was accredited with, at least, a partial relief. Moreover, as to grade the different therapies better, pain-free patients were sub-classified as complete responders. Finally, the signaled AEs were collected. Since all these diseases are characterized by exacerbations periods in which pain attacks develops and inter-critic periods in which pain is absent (PH, SUNCT and SUNA) or slight-moderate (HC), treatments were divided in two categories: treatments used to cease attacks during exacerbations and treatments taken regularly to control pain (especially in HC), trying to prevent the incoming of new active phases. The first treatments were indicated as “acute treatments”, whilst the second as “prolonged treatments”. Some acute treatments in HC and PH were used also to control pain outside exacerbations and were both considered as acute and prolonged treatments.

Drug mean dosage and therapeutic standards for non-pharmacological treatments were considered and summarized, even if not statistically analyzed.

Treatments used in less than five patients or which were clearly ineffective were not pooled in the statistical analysis, even if reported. Data regarding treatments used in 5 or more patients are summarized in Table 1, those ones regarding treatments used in less than 5 patients are reported in the Additional file 1: Table S1.

Globally, 65 articles were considered for the statistical analysis [6‐70]. Indomethacin was referred to as the most widely used treatment for HC. Melatonin was used in 17 patients, gabapentin and topiramate were utilized in 13 patients, onabotulinumtoxinA (OnabotA) in 12 patients and celecoxib in 11 patients. The other drugs were used in less than 10 patients. Supraorbital nerve blockade (SONB) was used on 17 patients, great occipital nerve blockade (GONB) on 15, occipital nerve stimulation (ONS) on 14 patients and minor occipital nerve blockade (MONB) on 6 patients.

Other drugs rather than indomethacin were used before indomethacin was given in 60% of cases, but only in the 20% of cases data were good enough to be considered (data not shown). Alternatively, since indomethacin was stopped in the 30% of cases because of its related AEs, other treatments were tried. Pharmacological treatments used in at least 5 patients, are summarized in Table 1 (section A). Statistical comparisons between the odds of responders and complete responders are summarized in Table 2 for the acute treatments and in Table 3 for the prolonged treatments. Data regarding those treatments performed in less than 5 patients are reported in Additional file 1: Table S1 (section A).

Fifty five articles were considered for PH [26, 38, 53, 60, 62, 66, 71‐123]. Indomethacin is the most used treatment (168 patients), followed by verapamil (30 patients), sumatriptan (24 patients) and oxygen (18 patients). Carbamazepine (CBZ) was tried on 15 patients, topiramate on 12 patients, amitriptyline and piroxicam on 5 patients. SONB, MONB and GONB were all used upon 6 patients. Piroxicam and amitriptyline were used upon 5 patients. All other treatments were used on less than 5 patients and were not taken into consideration for the statistical analysis. Treatments used in 5 or more patients are summarized in Table 1 (section B). Statistical comparisons of the odds of responders and complete responders for acute treatments are summarized in Table 4 whilst for the prolonged ones in Table 5. Data regarding those drugs taken by less than 5 patients are summarized in the Additional file 1: Table S1 (section B).

Globally 56, studies were analyzed [124‐179]. The most widely used treatment to control the excruciating and frequent attacks during active phases was lidocaine (36 patients), followed by prednisone (11 patients) and methylprednisolone (7 patients). To prevent the incoming of new active phases the most used treatments were: lamotrigine (84 patients), CBZ (78 patients), indomethacin (48 patients), gabapentin (48 patients) and topiramate (36 patients). All other treatments were used in less than 10 patients.

All these data are summarized in Table 1 (section C), data regarding statistical comparisons between the odds of responders and complete responders are summarized in Table 6 (acute treatments) and in Table 7 (prolonged treatments). Data regarding treatments used in less than 5 patients are reported in Additional file 1: Table S1 -section C.

Due to the infrequent diagnosis of these conditions, only case-reports or small case-series were found in literature and this strongly limits the reliability of the analysis. In many articles responders are not so well identifiable and in a very few ones the partial response was clearly described in terms of reduction of headache frequency, intensity or both, making almost impossible a comparison between the activity of different drugs on these aspects of pain. Treatment safety profile is hard to study too, primarily due to the sporadic report of AEs.

The first choice treatment for HC is indomethacin: for the management of recurrent exacerbations indomethacin should be the first choice drug, according with the higher effectiveness than all other treatments (see Table 2), which should be reserved to patients who don’t tolerated indomethacin. SONB has a similar proportion of responders but the lower odds of pain-free patients suggest that this technique is worse and more effective in diminishing pain rather than abolishing it [180]. It should also be considered that SONB has been tested only in a smaller number of patients than indomethacin and currently the experience on the use of these techniques is scarce, both for long-term availability (mean follow-up time = 93 days-data not show) and AEs profile. Celecoxib has an odds of responders lower than indomethacin but higher than GONB and an odds of complete responders higher than GONB and SONB, so it appears a better therapeutical approach than the last two in patients who don’t tolerate indomethacin. Piroxicam is comparable to celecoxib in terms of effectiveness, mirroring a similar action, as also stated by other studies [181]. GONB and MONB usefulness in relieving HC exacerbations seems to be negligible, like the usefulness of those treatments available for CH attacks, like SC sumatriptan and oxygen inhalation. This confirms that, despite the clinical over-lapping of HC and CH, the underlying pathogenetic mechanisms should be different, thus justifying a different pharmacological response [182]. HC management on long-time periods is unscheduled, but medications have been introduced trying to prevent pain recurrence. The prolonged use of drugs which were effective exacerbation control is a common practice and drugs like indomethacin, piroxicam and celecoxib are frequently used in HC patients outside active phases, even for many months: in our sample the duration of indomethacin assumption ranged between 5 and 1440 days, whereas from 18 to 540 for celecoxib. For piroxicam those data were not available, but its use for “many months” was reported in 5 patients out of 7. The stoppage of these drugs was due to AEs, mainly gastro-intestinal (GI), in the 70% of cases. The development of serious AEs is the main reason for which indomethacin, piroxicam and celecoxib should not be continued for many months outside exacerbations, even if the dose is titrated to the lowest possible or a preventive therapy with a proton pump inhibitor is started. SONB and GONB were both used even for the prevention of HC exacerbations, but GONB seems of no effect and SONB has a low odds of pain-free patients, denoting a partial action. The incoming of GI AEs and the low effectiveness of GONB and SONB impose the use of other drugs to control pain.

Gabapentin, topiramate, melatonin and OnabotA seems to be comparable in terms of effectiveness even if, considering the p-values of these comparisons (p = 0.063), a better action for gabapentin and topiramate than melatonin should be hypothesed. ONS should be a reliable option besides pharmacological techniques, as also confirmed from a recently published statement from the European Headache Federation [183]. The usefulness of verapamil in HC is scarce, since it has a lower odds of responders than indomethacin, OnabotA, topiramate, ONS and gabapentin and an odds of complete responders lower than all other treatments, except the non-pharmacological ones and melatonin.

The question on the tolerability of these treatments remains open and the unfair data about AEs make any comparison doubtful. Anyway, from the available literature, celecoxib and piroxicam should have a similar AEs profile than indomethacin with an even higher risk of cardiovascular side-effects with celecoxib [184], but a lower risk of renal AEs according to its higher COX-2 selectivity [185]: celecoxib and other COX-2 selective NSAIDs should be avoided with cardiovascular co-morbidities, but should be chosen after indomethacin in patients with renal diseases of with gastro-intestinal co-morbidities.

PH is another member of the so-called indomethacin-responsive headaches [1] and, in fact, indomethacin is undoubtedly the best treatment even for this condition. The activity of other treatments is low both for the acute treatment and for the prolonged one. Piroxicam emerges as the best treatment besides indomethacin for exacerbations management, according to the higher odds of pain-free patients than oxygen and sumatriptan. The usefulness of this last two drugs is almost null and this confirms once again the differences in TACs’ pathogenesis besides their clinical similarity [182]. Even when used for PH control outside active phases indomethacin is the most effective treatment. Even so, since PH is frequently chronic and indomethacin assumption for long periods of time may cause a wide range of AEs, this usually lead to the discontinuation of therapy in about 27% of cases. This imposes the use of different treatments to control the pain, but other tested drugs seems to be of little use with the most effective being rofecoxib, which has been retired from the international market because of its cardiac side-effects [186]. Piroxicam seems to be the most effective treatment other than indomethacin, even if the possibility of having GI AEs remains [187] and, like indomethacin, its use should be avoided for long periods of time. Since the hypothesized overlapping between PH and migraine pathogenesis [15], two well-known migraine prophylaxis such as topiramate and amitriptyline have been tried for PH, with comparable and moderate results. Topiramate and amitriptyline are also comparable to piroxicam and verapamil in terms of effectiveness, even though the latter shows a not-significant higher odds of responders and complete responders. CBZ usefulness seems to be low and the null number of complete responders should discourage its use for PH management.

To stop SUNCT and SUNA exacerbations, lidocaine (intravenously or subcutaneously) seems to be the most effective treatment and is now emerging as a novel option for chronic pain syndromes [188]. Its effectiveness is unquestionable, but paranoid idealization, depressive thoughts and cardiac arrhythmias were registered as AEs: this imposes the careful and shortest use of this drug only for the worst cases and the patient’s continuous monitoring with a 12-lead ECG registration and sequential blood pressure measurements during the treatment [189]. In our sample the time of use ranged between 2 to 10 days (data not shown). Steroids represent a less effective but safer options for stopping attacks, with methylprednisolone presenting a better action than prednisone, even if not significantly. As previously discussed for lidocaine, steroids should be given intravenously for the shortest time as possible: from literature it is well-known that they can have a wide range of AEs, which can be prevented by reducing the duration of infusion to the time necessary for the ceasing of painful exacerbations [190]. In our sample the mean time of infusion was 8 ± 4.32 days (data not shown) The usefulness of phenytoin should be considered negligible.

Lamotrigine is the best drug for the prevention of the incoming of new active phases, but seems to be more suitable in reducing attacks frequency rather than abolishing them completely: it has an odds of partial responders higher than all other drugs, but the odds of complete responders are comparable, with the exceptions of CBZ and indomethacin, which efficacy is scarce. Non-pharmacological techniques have an odds of responders comparable to lamotrigine and, moreover, ONS has even a higher odds of complete responders. Lamotrigine has also a similar AEs profile than other treatments except for gabapentin, confirming the available literature [191].

Verapamil, gabapentin and topiramate have similar effectiveness, with gabapentin showing a better AEs profile, even if the number of reported AEs is too poor to let a reliable comparison. CBZ appears less useful in treating SUNCT and SUNA than gabapentin and topiramate, according with the lower number of complete responders. Indomethacin usefulness in these conditions is sometimes reported, but should be considered as negligible: an occasional benefit of this drug in SUNCT or SUNA should rise the question of a diagnostic mistake with HC, PH or a secondary headache, imposing the reconsideration of the initial diagnosis, following scheduled diagnostic algorithms [192].

Recently, non-pharmacological techniques has gained importance in the treatment of these disorders, but the experience with these treatments is scarce and the long-term follow-up of patients is often lacking in many studies. From the available data ONS has emerged as the best technique and this result is in accordance with the findings in CH, were ONS is the only class-A evidence treatment for the American Headache Society (AHS) [193]. Moreover, even the European Headache Federation (EHF) has confirmed the effectiveness and safety of this technique in SUNCT and SUNA, pointing out that 4 patients out of 6 analyzed were nearly pain free with mild facial paresthesia as the principal AE [183].

From the reviewed literature, ONS has demonstrated an almost complete effectiveness and a good safety profile, but it has been tried only on 7 patients. Ventral tegmental area deep brain stimulation has shown a similar effectiveness, but adverse events were reported in the 100% of cases and should be reserved to the refractory cases. Finally, GONB appears to be less effective but also safer than the previous techniques and should be considered as a reliable alternative in patients with episodic forms.

In Fig. 2 the ORs of complete responders and the relative IC95% are visually summarized for all diseases. ORs are calculated as the odds of pain-free patients for the indicated treatments split by the odds of pain-free patients for the most used treatment for every disease.