Abstract
Background: Olfactory hallucinations (phantosmias) have rarely been reported in migraine patients. Unlike visual, sensory, language, brainstem, and motor symptoms, they are not recognized as a form of aura by the International Classification of Headache Disorders.
Methods: We examined the clinical features of 39 patients (14 new cases and 25 from the literature) with olfactory hallucinations in conjunction with their primary headache disorders.
Results: In a 30-month period, the prevalence of phantosmias among all patients seen at our headache center was 0.66%. Phantosmias occurred most commonly in women with migraine, although they were also seen in several patients with other primary headache diagnoses. The typical hallucination lasted 5–60 minutes, occurred shortly before or simultaneous with the onset of head pain, and was of a highly specific and unpleasant odor, most commonly a burning smell. In the majority of patients, phantosmias diminished or disappeared with initiation of prophylactic therapy for headaches.
Conclusions: We propose that olfactory hallucinations are probably an uncommon but distinctive form of migraine aura, based on their semiology, timing and response to headache prophylaxis.
Introduction
Over 30% of migraine patients experience aura in association with their recurrent headache attacks (1). Visual phenomena are the predominant manifestations of migraine aura and are a form of organic hallucinosis (2). Olfactory hallucinations, or phantosmias, are well known to occur in temporal lobe epilepsy and schizophrenia, and have also been documented in patients with cerebral aneurysms (3,4), intracranial mass lesions (5), Parkinson's disease (6), and head trauma (7), and in patients receiving caffeine infusions (8).
Phantosmias have infrequently been reported to occur in patients with primary headache disorders. Unlike visual, somatosensory, language, motor, and brainstem symptoms, olfactory hallucinations in the setting of an acute migraine attack are not a recognized aura symptom by the second edition of the International Classification of Headache Disorders (ICHD-2) (9), although some authors have suggested their inclusion (10). Olfactory aura is probably rare, reported by only one patient in a detailed study of 952 consecutive migraine patients seen in a single tertiary care center, giving a point prevalence within migraine patients of 0.1% (2). Most previous reports of patients with primary headache disorders and phantosmias did not elucidate the finer details of these patients’ presentations, and many did not mention medications taken or coexisting medical and psychiatric conditions.
In clinical practice we have occasionally evaluated patients with a variety of primary headache disorders who report phantosmias in association with their discrete headache attacks or, in the situation in which patients experience continuous head pain (such as certain forms of chronic daily headache), with their headache exacerbations. The objective of this study was to gain a greater understanding of this phenomenon by performing a detailed analysis of the aggregated patients reported in the literature and those encountered at our center, which treats a diverse urban patient population as well as a large number of regional and national referrals. We also aimed to estimate the prevalence of phantosmias across all primary headache disorder patients evaluated at our center.
We hypothesized that: 1) most headache patients experiencing phantosmias have migraine, 2) phantosmias have similar semiology to other aura types in their timing and evolution, and 3) phantosmias become less frequent with prophylactic therapies aimed at suppressing the underlying headache disorder.
Methods and design
This study was conducted in two parts. The first part was conducted as a retrospective chart review of adult patients (age ≥18 years) seen at the Montefiore Headache Center (MHC) starting from 1 September 2008 until 31 December 2010. All new patients complete a standardized questionnaire that captures detailed demographic, headache and associated symptoms, and past medical, psychiatric, social, and family history characteristics, which has been described elsewhere (11). There was no question specifically querying for the presence of smells before or during headache attacks, so patients either wrote the symptom in the space available for ‘other’ symptoms or discussed it with their physician directly. Active depression and anxiety were ascertained using the Patient Health Questionnaire (PHQ-9) (12) and Generalized Anxiety Disorder (GAD-7) (13) scales. Investigations such as neuroimaging, blood tests, electroencephalography, and lumbar puncture were requested as clinically indicated to exclude secondary etiologies as part of standard clinical care. Primary headache disorders were diagnosed according to the criteria set forth by the ICHD-2. Management decisions of all patients were made at the discretion of the treating physicians as part of standard clinical care.
The second part of the study entailed a literature review by using the online databases PubMed and Google Scholar and hand searching available issues of Cephalalgia and Headache. We excluded reporting patients with secondary etiologies to their headache and olfactory hallucinations from both our case series and literature review. In a few cases from the literature a primary headache diagnosis was not noted. In these cases, we have identified the most likely diagnosis based on the description of headache symptoms, and the diagnosis is listed as ‘probable’ in Table 2. Approval from Montefiore Medical Center's Institutional Review Board was obtained before undertaking this study.
Results
Patients captured and prevalence
Montefiore Headache Center patient characteristics
Literature Review Patient Characteristics
Demographic information
Twenty-nine patients were female and eight male, giving a female to male ratio of 3.6:1. Two cases from the literature did not report gender. Mean age at presentation was 37.2 years (range: 8–58). Mean age of headache onset was 21.1 years for the 24 patients for whom data was available. For a number of patients (Patients 7, 14, 24, 25, 32, and 40) the headache onset was recorded as occurring during ‘childhood’ or ‘adolescence’ without a specific age noted, so the actual mean age of headache onset is probably somewhat younger. Mean age of phantosmia onset was 34.6 years, with data available for 14 patients.
Primary headache diagnoses
Of the 39 patients with olfactory hallucinations, 33 (84.6%) were diagnosed with migraine, 3 (7.7%) with cluster headache, and one each (2.6%) with hemicrania continua (HC), new daily persistent headache (NDPH), and unspecified chronic daily headache. Among the migraine patients, 12 (36.4%) had migraine with typical aura (seven visual, two visual and sensory, and three unspecified) and 21 (63.6%) had migraine without aura.
Timing and duration of olfactory hallucinations
Using the ICHD-2 criteria for typical aura as a framework, we divided the duration of phantosmia symptoms into three categories: less than 5 min, 5–60 min, and greater than 60 min. Data regarding the duration of hallucinations was available for 12 of the 14 patients in from our clinic, and for 13 of the 25 cases in the literature; overall we present data for 25 of the 39 patients in the series. Of these, one patient had phantosmias lasting less than 5 min, three had phantosmias lasting more than an hour, and 16 had phantosmias that invariably lasted 5–60 min. An additional five patients had phantosmias that varied in duration from attack to attack, lasting anywhere from a few minutes to several hours. Thus 64.0% of patients for whom duration of phantosmia was recorded (41.0% of all patients in the series) fell within the 5–60 minute window in every attack, and 84.0% (53.8% of all patients in the series) fell within that window in some of their attacks.
Regarding the temporal relationship between phantosmias and headache, 25 patients (64.1%) experienced phantosmias before the onset of headache. Of these, ten patients invariably experienced olfactory hallucinations 60 min or less before the onset of headache. For eight patients who experienced phantosmias before headaches, the time interval was not recorded, though from the phrasing of the reports (e.g. ‘headaches were preceded by a smell…’) it can be reasonably inferred that the interval was less than an hour. Four patients' phantosmias occurred within a variable time frame ranging from minutes to hours before headache. All together, including those patients for whom a numerical interval was not specified, 56.4% of patients had phantosmias less than 1 hour before headache onset in at least some attacks. Three patients (7.7%) had phantosmias more than an hour (range: 2–72 hours) before the onset of their headache. Fourteen patients (35.9%) experienced phantosmias during their headache, including the three patients with chronic daily headache (NDPH, HC, and unspecified chronic daily headache), all of whom experienced phantosmias concurrent with headache exacerbations.
Data were available for 12 patients (six from MHC, six from the literature) regarding timing of phantosmias with their other typical aura symptoms. In four patients, phantosmias occurred simultaneously with a typical visual aura; in one of these patients, the phantosmias could occur immediately after the visual aura as well. In two patients, phantosmias preceded headache, and typical sensory auras started during the headache phase. In three patients, phantosmias occurred during the headache phase, just as the visual or sensory aura concluded. In one patient, phantosmias occurred during the headache phase as visual or sensory aura continued. In two patients, phantosmias and typical visual or sensory auras did not occur in the same attacks.
Character and hedonic quality of hallucinations
Odors described by patients
Twenty-nine patients (74.4%) experienced conventionally unpleasant smells, five patients (12.8%) smelled pleasing scents, and two patients' (5.1%) perceptions were neutral. Most patients (82.1%) described their hallucinations as specific, recognizable smells, using descriptors such as ‘coffee’, ‘skunk’, or ‘burning cookies.’ Only seven patients (17.9%) described smells as vague or unidentifiable, using words like ‘sweet’, ‘unusual’, or ‘repulsive.’ Although the smells were diverse, most fall into a few broad categories, for example burning, fecal/decay, foodstuffs, chemical, and floral. The most common categories were burning, which was reported by 26.1% of patients, and rot, which was reported by 17.9%.
Primary hedonic quality of phantosmias by number of patients
MRI, CT and EEG findings
Ten of the 14 patients from our center had MRIs performed. Of these, eight had normal studies and two had findings that are described in Table 1 and were considered incidental. Five of the 25 patients described in the literature had MRIs, all of which were normal. One patient from our center and ten patients from the literature had head CTs, all of which were normal. An additional literature patient had a ‘full radiographic evaluation’ that was normal. Four of the patients in our series and 11 from the literature underwent EEG. Of these, all but one (Patient 26) were normal. Her findings are described in Table 2 and were not considered to be epileptiform in nature.
Osmophobia
Overall, 11 of 14 MHC patients (78.5%) reported osmophobia in conjunction with their headache attacks. Those with osmophobia constituted 11 of 12 patients with a diagnosis of migraine; data were unavailable for one migraine patient. The two patients who denied osmophobia had diagnoses of cluster headache and NDPH. None of the literature reports noted the presence or absence of osmophobia.
Medical and psychiatric comorbidities
Thirteen of the 14 patients from our center had medical comorbidities, which are elaborated in Table 1. Medical comorbidities were described in four patients in the literature; four literature patients were specified as healthy aside from their headache disorder, and for the remainder this information was not recorded. The most common medical comorbidities were hypertension and fibroids (n = 3), followed by hyperthyroidism and unspecified head injury (n = 2). Other comorbidities appearing in single patients included HIV, cervical dystonia, polycystic ovarian syndrome, menorrhagia, endometriosis, and benign positional vertigo of childhood.
Psychiatric comorbidities were described in four of our patients and five patients from the literature. Among the MHC patients, the most common psychiatric comorbidities were depression and anxiety, which coexisted in two patients. One patient had panic attacks and one had a history of post-partum depression. Of the five patients from the literature (all of whom were from the same case series (15)), four had major depression and one had generalized anxiety disorder.
Self-reported mood symptoms
Among our 14 patients, eight (57.1%) reported that their headaches were accompanied by cognitive and/or affective symptoms, with seven reporting one or more of the following: irritability, difficulty concentrating, memory problems, or confusion. The remaining patient reported an ‘unsettled feeling.’
In ascertaining depressive symptomatology, PHQ-9 scores were available for only eight patients, of whom four scored in the normal range (0 to 4), one scored as mild depression (5 to 9), and three scored as moderate depression (10 to 14). Only one of these three patients had a known history of depression; one had a history of panic attacks, and the other had no psychiatric history. In ascertaining anxious symptomatology, GAD-7 scores were available for seven patients. Four patients scored in the normal range (0 to 4), two scored as mild anxiety (5 to 9), and one scored as moderate anxiety (10 to 14).
Response to treatment
In our series, nine of 14 patients (64.3%) responded to medical therapy, with reduction in the frequency of both headache and phantosmia symptoms, while five (35.7%) did not improve with treatment. Successful medication trials included topiramate (n = 3), nortriptyline (n = 2), and verapamil (n = 2), which had no failures. Gabapentin treatment yielded one positive response and two treatment failures.
Two patients – Patient 9, who was treated with Petasites hybridus root extract, and Patient 11, who was treated with verapamil and magnesium oxide – described complete resolution of phantosmias and only partial resolution of headaches with treatment. Patient 9 also described improvement of her visual and sensory aura out of proportion to the improvement in her headaches. In no case did headache improve with treatment without concurrent improvement in phantosmia.
Among the literature patients, response to treatment was described for ten of 25 patients. Of these, nine responded to medical therapy, while the tenth (Patient 33) failed medical therapy but responded to combined occipital–supraorbital neurostimulation. Seven of the nine patients responded to propranolol, and two of those patients had previously tried carbamazepine, with no improvement. One patient (Patient 34) was diagnosed with hemicrania continua and responded to indomethacin. The final patient, Patient 37, reported a decrease in phantosmias but increase in headaches with valproic acid, but responded well to phenytoin, with improvement in both phantosmias and headaches.
Overall, response to prophylactic medication was recorded for 24 patients in the series. Of these, 18 (75%) responded positively, with improvement in headaches and diminution or disappearance of phantosmias. Six did not respond to medication, although one of these did respond to neurostimulation. In no patient was improvement seen in headaches without concurrent improvement in phantosmias.
Discussion
We have presented 14 new cases and reviewed 25 published cases of olfactory hallucinations occurring with primary headache disorders, of whom the preponderance (84.6%) suffered from migraine. Diagnostic tests revealed no secondary etiology of headache and no alternative explanation for olfactory hallucinations, such as epilepsy or mass lesions, although the diagnostic testing received varied widely. Among the patients at our center, ten had MRI and four had EEG, while across the entire cohort 15 underwent EEG and 27 had neuroimaging with either CT or MRI.
Although olfactory hallucinations are an uncommon finding in migraine patients, the 33 cases of migraine reviewed here demonstrate that it undoubtedly occurs. At our center, the prevalence of phantosmias among all headache disorder patients was 0.66%. When applied to the worldwide prevalence of migraine in the worldwide adult population, estimated at 11% (27), nearly 5 million individuals may have migraine with olfactory hallucinations. This prevalence estimate is probably an overestimation when applied to the general population, as patients seen in tertiary care are likely to have more complicated and severe headache disorders, although the lack of a specific questionnaire item for smells could also lead to its underestimation. However, this estimate indicates that this clinical problem may be encountered by many practitioners treating headache patients.
Although other disorders must be ruled out in atypical or suspicious cases, in very straightforward presentations of migraine extensive diagnostic testing is probably unnecessary. We also present isolated cases of phantosmia occurring in patients with other headache disorders, including cluster headache, HC, and NDPH. Though typical auras have rarely been reported in patients with non-migraine headache disorders, little is known about the pathophysiology or epidemiology of aura in these conditions. Aura and headache may be regulated by separate sets of genes (28).
Olfactory hallucinations as a form of migraine aura
The ICHD-2 describes three cardinal characteristics of typical migraine auras: 1) complete reversibility, 2) duration of 5–60 min and 3) gradual development. In addition, headache should start during or within 60 min of resolution of aura symptoms (9). In this series, phantosmias were always reversible, satisfying the first criterion. The temporal relationship was clearly present in our patients, with 56.4% experiencing phantosmia within 60 min before headache and 35.9% experiencing phantosmia coincident with their headache. Interestingly, the latter number is far higher than the 4.7% reported for other types of aura (29).
In most cases (21 of 25 for which data is available), olfactory aura fulfills criterion 2 in some or all attacks, with a few outliers experiencing hallucinations of longer duration. Prolonged visual and sensory auras are well known, and were initially included as a separate diagnostic code in first edition of the ICHD. The authors of the ICHD-2 removed this classification because the vast majority of patients with prolonged auras had other attacks that fulfill the typical classification, and thus prolonged aura was not viewed as a truly separate phenomenon. Our data demonstrate a similar pattern for olfactory aura, with five patients experiencing both prolonged and typical auras. Three patients in the series reported exclusively prolonged auras, and although such prolonged auras are occasionally seen in typical aura types, they may warrant more rigorous search for secondary causes such as a mass or arteriovenous malformation. In regard to the one patient whose aura lasted only a few seconds (Patient 1), although the time course is concerning for temporal lobe epilepsy, she had a normal EEG and an otherwise completely typical presentation for migraine, including a visual aura of scintillating scotoma. We are therefore comfortable ascribing her olfactory hallucinations to migraine, despite its unusually short duration.
The third criterion, gradual onset, is inherently difficult to apply to olfactory hallucinations. Visual auras expand or move across the visual field, whereas sensory auras may show a Jacksonian march. It is not clear what the analogous progression would be in olfactory auras, but certainly in nearly all cases, olfactory hallucinations seem to come on all at once and remain static until they disappear. It should be noted, however, that our questionnaire did not directly ask patients whether or not their olfactory aura changed over time in intensity or character, and examiners may not have probed patients on this feature. However, most patients did not volunteer this information.
Although the ICHD-2 requires gradual onset, the ICHD-1 recognized a subtype of acute onset auras, with one study finding that 13% of migraine auras began suddenly (30). As our examination of olfactory hallucinations in migraineurs leads us to believe that they are consistent with aura phenomena in all other respects, we suggest that even if most olfactory auras are truly of acute onset they are not atypical for aura semiology.
Three intriguing cases in this review do show gradual evolution of olfactory symptoms, which may be analogous to the clinical progression seen in most visual and sensory auras. Patient 3 described her aura as beginning with a nondescript ‘sweet’ smell, evolving slowly into a rotting flowers scent. Patient 14 endorsed a clear gradual onset and offset to her phantosmias. Patient 26 had phantosmia as part of a complex set of perceptual and mood disturbances: she perceived a repulsive smell, beginning suddenly and lasting 5 min to 2 hours. After the smell disappeared she experienced anger and aggression, followed by depression, and then anxiety. This emotional progression lasted about an hour and was ‘accompanied by the hearing of familiar voices, especially her mother, berating her’. After these symptoms abated her headache began – usually right-sided and accompanied by nausea and sometimes numbness in the right arm. In light of her complex hallucinations temporal lobe epilepsy was suspected, but her EEG showed non-specific changes and the attacks never involved any alteration of consciousness, were completely unresponsive to carbamazepine, and resolved completely with propranolol.
This was the only case in the series in which the olfactory hallucination was part of a complex perceptual disturbance reminiscent of temporal lobe epilepsy (TLE) (though one other patient hints at a ‘reminiscence’ component – Patient 27, who smelled her grandfather's cigars). It raises the possibility that in migraine, as in TLE, olfactory hallucinations may be linked to emotive symptoms. Sacks reported seeing several migraine patients with hallucinations of smell associated with forced reminiscence and feelings of déjà vu (31). This linkage is logical, considering the proximity of the primary olfactory cortex, which includes the amygdaloid nucleus, uncus, piriform, and entorhinal cortex, to the hippocampus and other limbic structures (32). Cortical spreading depression propagating through the deep temporal structures and generating olfactory, mood, and memory symptoms as it spreads is a plausible mechanism. This phenomenon was proposed by Morrison, who found a 100% correlation between abnormal perceptual experiences, including olfactory hallucinations, and transient mood symptoms during migraine attacks (15). Morrison also found correlation with diagnoses of affective disorders, and further proposed that repeated electrical derangement of the limbic structures during migraine might predispose such patients to depression and anxiety (15).
Our findings did not reproduce Morrison's, with only 54% of patients acknowledging cognitive and/or affective symptoms as part of their headache complex. In terms of comorbidity, among the subset of patients who completed the PHQ-9 and GAD-7 screens, rates of depression were 37.5% and anxiety were 14.3%. The small sample size renders drawing any conclusions difficult. Finally, we found that four of 13 patients with phantosmias had past or current psychiatric diagnoses. Thus our findings do not necessarily support Morrison's theory that neural disruption involved in recurrent phantosmias leads to such enhanced susceptibility.
Distinguishing olfactory hallucinations in primary headache disorders from other etiologies of phantosmias
Comparison of features of phantosmias in different syndromes
Migrainous olfactory hallucinations are characterized by several features. They occur in people who meet criteria for a primary headache disorder and are temporally related to episodes of head pain. In a few patients, olfactory hallucinations occur at intervals more distant from the onset of head pain, but this is atypical and in such cases, a more rigorous search for other potential causes of phantosmias is probably warranted. Second, these olfactory hallucinations almost invariably last minutes to hours, whereas epileptic auras last on the order of seconds. Third, migrainous olfactory hallucinations are usually highly specific, identifiable smells. In contrast, epileptic olfactory auras are classically described as crude and unrecognizable (33), or at best as ‘sharp, smoky, or disgusting’ (34), although one study of 13 patients with epileptic olfactory auras found that eight of 13 had specific, identifiable hallucinations (35). Interestingly, most of the specific smells recorded in this study (35) were also seen in our series, including sulfur, peanut butter, flowers, burning, and gas. Thus the assignment by a patient of a specific label to a hallucination type is highly suggestive of migrainous phantosmia, but does not exclude epilepsy. Migrainous olfactory hallucinations are usually unpleasant, as are epileptic auras.
Therapeutic response can also be valuable in differentiating migrainous from epileptic phantosmias, aside from some antiepileptic drugs, such as topiramate and gabapentin, which may suppress both conditions. Diminution of phantosmias by beta blockers or calcium channel blockers is highly suggestive of migraine, whereas response to phenytoin or carbamazepine raises the suspicion for epilepsy. Patient 37 reported worsening of headaches with valproate, an anticonvulsant with demonstrated efficacy in migraine (36), but improvement in both phantosmias and headaches with phenytoin, an anticonvulsant with no known utility in the treatment of migraine. However, this patient was atypical: although her phantosmias were often associated with headaches, she also had phantosmias in the absence of headaches. Except for the phenytoin response, her presentation was not consistent with epilepsy, as she had a normal EEG, her phantosmias lasted at least 30 min, and there was no alteration of consciousness.
Phantosmias in psychiatric disease have several characteristics distinguishing them from migrainous phantosmias. In depressed patients, the phantom odor is usually perceived as ‘intrinsic’, that is, emanating from the patient's own body. This is also the case in the olfactory reference syndrome, a psychiatric disorder characterized by persistent hallucinations of a foul odor perceived as arising from the patient in the absence of previous depressive symptoms. Patients with schizophrenia, on the other hand, perceive phantom smells as extrinsic, but believe that the odor is being maliciously forced on them. In all psychogenic causes of phantosmia, there is some degree of incomplete insight into the phantom nature of the smell, whereas migraine patients, after searching unsuccessfully for a source of odor, invariably come to understand that they are experiencing a hallucination (37).
There are rare reports of lesional causes of headache associated with phantosmias, localizing to the left medial temporal lobe. One case was a patient with typical migraine headaches who had a lesion in the left posterior temporal area, thought most likely a cyst or hamartoma (38). The other case was a patient with a primary intracerebral hemorrhage in the left uncinate lobe (39). In both cases patients had normal EEGs, but partial seizures were not definitively excluded. These cases demonstrate that irritation of the temporal lobe can cause both headache and phantosmia.
Possible mechanisms
Cortical spreading depression (CSD) of the olfactory cortex in the temporal lobe and orbitofrontal cortex may underlie migrainous phantosmias, just as CSD in occipital cortex produces visual aura and CSD in parietal cortex produces sensory aura. The fact that phantosmias represent a fully reversible neurological symptom with timing and duration similar to typical auras supports this theory.
Another potential mechanism regarding olfactory hallucinations in migraine patients relates to dopamine. It has been shown that Parkinson's disease patients have increased dopaminergic periglomerular cells in their olfactory bulbs, and this increase in dopamine has been proposed as the mechanism behind their phantosmias (6). Dopamine has a complex role in migraine pathophysiology, but may be especially important in the development of premonitory symptoms before headache attacks (40). The fact that phantosmias in migraine usually occur before headache attacks, and in some cases many hours before headache attacks (unlike typical auras), make a dopaminergic link intriguing.
Another possibility involves irritation of the olfactory bulb. Phantosmias have been recently reported in a patient with rhinosinusitis, with definitive resolution of the hallucinations with oral and intranasal steroid therapy (41). The authors postulated that olfactory bulb irritation can lead to paroxysmal olfactory bulb activation (41). This could be a plausible theory for phantosmias in headache disorders, in which cranial autonomic symptoms such as nasal congestion and rhinorrhea may occur. However, in our series and review, autonomic symptoms were not frequently reported. In addition, primary headache disorders with prominent autonomic features (cluster headache, hemicrania continua) comprise the minority (10%) of the 39 patients in our series.
Role of the piriform cortex
One feature of migrainous phantosmias that emerged in our review may provide a clue to more precise localization to the piriform cortex in their generation. The piriform cortex functions as an associative olfactory cortex, linking odor representations formed in lower brain regions (namely anterior olfactory nucleus) to myriad cognitive and behavioral areas (32,42). These connections include links to language and memory centers, making the piriform cortex central to the ascription of names and contexts to smells. Thus the consistently identifiable and highly specific nature of migrainous phantosmias suggests that the piriform cortex may have an essential role in their development. In contrast, the typically amorphous phantosmias of epilepsy have been attributed, via intra-operative stimulation and post-excision studies, to the amygdala (43) and uncus (33).
The implication of the piriform cortex in migrainous phantosmias is especially interesting in the context of the role for the piriform cortex in CSD and olfactory sensitivity in migraineurs. Gursoy-Ozdemir et al. found that the piriform cortex is highly sensitive to the effects of CSD, showing higher levels of matrix metalloprotein activation than other brain regions, including the occipital cortex (44). Demarquay et al., using positron emission tomography (PET) scanning to investigate the olfactory cortex of migraineurs with olfactory hypersensitivity, found hyperactivity of the left piriform cortex in migraineurs compared with controls (45). This area requires further investigation, but taken together, these studies as well as our findings suggest that the piriform cortex may be an important region in migraine pathophysiology and olfactory derangements in migraine.
Directions for further study
The phenomenon of olfactory hallucinations in primary headache disorders deserves further elucidation, as our study was limited by its retrospective nature. Longitudinal, prospective studies using detailed aura and headache diaries would be required to elucidate the symptoms further and establish its true prevalence. Using a control group without migraine or severe headache would also be a useful arm to such a study. A prospective analysis could also address more definitively the low presence of migraine with typical aura (36.4%) in our series, which could be explained in part by the lack of reporting of phantosmias because they are overshadowed by more bothersome visual symptoms. The lack of any prolonged EEG monitoring or ictal EEG was also a limitation, particularly as migraine and epilepsy are comorbid disorders (46) and partial seizures are not definitively excluded by single, routine surface EEG.
Another intriguing avenue for future research is investigation of olfactory sensitivity in patients with migrainous olfactory hallucinations. Studies of olfactory sensitivity in migraineurs in general have found conflicting results (47,48). As visual and sensory migraine auras typically include both positive and negative symptoms, we can theorize that migraine patients with olfactory aura may experience anosmia or hyposmia in the wake of their olfactory hallucinations. This hypothesis could be tested by providing patients with smell test kits and instructions for self-administering the test at home.
Finally, fMRI or PET studies during phantosmias could yield valuable information about the brain regions and mechanisms involved in generating hallucinations. Such studies in patients with visual aura have shown cerebral blood flow changes consistent with CSD (49).
In summary, phantosmias are an uncommon but distinct manifestation of headache disorders. They occur predominately in women with migraine and are usually readily identifiable and unpleasant, with burning the single most common odor described. Their semiology, timing and response to headache prophylaxis suggest that they deserve strong consideration as a form of aura.
Footnotes
Acknowledgements
We would like to thank Oliver Sacks for his candid, interesting, and helpful comments on our study.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflicts of interest
BMG has received personal compensation for activities with Merck & Co, Inc, and has received research support from Allergan, Inc, St Jude Medical, OrthoMcNeil Pharmaceuticals, GlaxoSmithKline, Inc, Merck & Co, Inc, Advanced Bionics, Proethics, and Pfizer, Inc. ERC and MSR have nothing to disclose.