Smetana GW. The Diagnostic Value of Historical Features in Primary Headache SyndromesA Comprehensive Review. Arch Intern Med. 2000;160(18):2729-2737. doi:10.1001/archinte.160.18.2729
Existing criteria for the diagnosis of headache do not indicate which clinical features are most important to establish or exclude primary headache diagnoses.
To determine the value of history taking in the evaluation of patients with primary headache syndromes.
A MEDLINE search of English-language studies published since 1966 and a hand search of bibliographies from retrieved articles. Included studies cited the sensitivity of clinical and historical features for the diagnosis of migraine, tension-type, and cluster headaches. Studies were classified according to their use of the International Headache Society criteria or other criteria as the reference standard for diagnosis.
The features most predictive of migraine, when compared with tension-type headache, are nausea, photophobia, phonophobia, and exacerbation by physical activity. Positive likelihood ratios (95% confidence intervals) are 19.2 (15.0-24.5), 5.8 (5.1-6.6), 5.2 (4.5-5.9), and 3.7 (3.4-4.0), respectively. Headache precipitants are present equally in patients with migraine and tension-type headache, with the exception of food triggers, which confer a positive likelihood ratio of 3.6 (2.8-4.6) for the diagnosis of migraine. Visual aurae are present in 84% of patients with migraine with aura. Cluster headache is strictly unilateral, is periorbital in 80% of patients, and lasts less than 1 hour in 54% of patients.
The features most predictive of the diagnosis of migraine, when compared with tension-type headache, are nausea, photophobia, phonophobia, exacerbation by physical activity, and an aura. Cluster headache is a distinct clinical syndrome.
HEADACHE is part of the human experience. The near universal nature of headache poses a challenge for clinicians seeking the few worrisome headache types among all patients with headache. As an illustration of its ubiquity, the lifetime prevalence of headache in a study of 740 randomly selected individuals was 99% in women and 93% in men.1 In an early survey of more than 1 million individuals in the general population, headache was the most frequent current physical complaint among men and women.2 Although most headaches are benign, they may lead to significant disability. In a report of lifetime headache histories in unselected volunteers, 41% of men and 50% of women experienced severe headaches that rendered them unable to work.3
Despite the frequency of headache in ambulatory practice, many physicians are uncomfortable evaluating patients with headache and may be concerned about misdiagnosing the rare patient with a worrisome cause. A disparity often exists between the severity of headaches and the need to worry. For example, patients with uncomplicated migraine do not have a pathologic cause of headache, but may have some of the most severe and disabling forms of headache. This disparity heightens physician anxiety and can lead to overuse of imaging studies to reassure the patient and physician.
The most comprehensive and widely cited headache classification scheme is the International Headache Society (IHS) classification and diagnostic criteria for headache disorders, cranial neuralgias, and facial pain.4 This classification establishes 13 broad categories and 128 distinct headache syndromes. Operational diagnostic criteria exist for each syndrome. These criteria aim to standardize headache classification and to improve the homogeneity of patient groups in research studies. Among headache specialists using these criteria to diagnose primary headaches, there is a high degree of interobserver reliability.5 The practicing clinician, however, may find these criteria too complicated and restrictive for use in the day-to-day office evaluation of patients. In addition, these criteria do not indicate which clinical features are most helpful to establish or exclude a particular diagnosis.
Many previous review articles have proposed an approach to the patient with headache. No authors have determined the diagnostic value of individual clinical and historical features through a comprehensive survey of published clinical series. The purposes of my review are to summarize a large body of literature and to determine the value of history taking in the evaluation of patients with primary headache syndromes. I will review the sensitivity, specificity, and likelihood ratios of specific clinical features.
I performed a MEDLINE search of English-language articles published from January 1, 1966, through May 31, 1999. Search terms included migraine, tension-type headache, cluster headache, clinical features, symptoms, history, aura, precipitants, triggers, family history, and motion sickness. Additional articles were identified through a hand search of the bibliographies of retrieved articles, previous reviews, textbooks, and monographs. Included studies cited the sensitivity of specific historical features among patients with migraine, tension-type, or cluster headache.
Primary headaches are clinical diagnoses that are based on results of history taking alone. Clinical series require a reference standard(gold standard) for diagnosis. The choice of reference standard directly affects the sensitivity of clinical features. Recent studies have generally used the IHS operational criteria to classify patient diagnoses. Before the publication of the IHS criteria in 1988, authors used varying definitions. I classified all articles based on the authors' use of the IHS classification, other criteria, or indeterminate reference standard for inclusion of patients.
I considered individual symptoms and historical features to be diagnostic tests for a given headache diagnosis. Sensitivity was defined as the proportion of patients with the stated headache diagnosis who had the particular clinical feature. Specificity was defined as the proportion of patients without the stated headache diagnosis who did not have the particular clinical feature. For migraine and tension-type headaches, sufficient data existed to calculate likelihood ratios for clinical features and headache precipitants. The calculated values are necessarily subject to incorporation bias, as no diagnostic reference standards exist that are independent of clinical features (ie, pathological or laboratory confirmation). The positive likelihood ratio indicates the increase in the odds of a particular headache diagnosis if a given historical feature is present and is defined as sensitivity/(1 − specificity). The negative likelihood ratio indicates the decrease in the odds of a particular diagnosis if a given historical feature is absent and is defined as (1 − sensitivity)/specificity. Confidence intervals were calculated according to the method of Simel and colleagues.6
I determined the heterogeneity of data across individual studies using the criteria of Panju and colleagues.7 I considered studies of individual features to be homogeneous if the χ2 test of sensitivity excluded statistically significant heterogeneity (P>.05) or if the range of sensitivities across the studies was 15% or less. Using these criteria, significant heterogeneity existed for many clinical features. This reflected the large number of studies in the pooled data and the selection of patients from varying populations by different authors. Rather than restricting my analysis to a limited subset of patients who were identified in an identical manner by the authors, I chose to pool data even when significant heterogeneity existed. This allowed a more comprehensive review of clinical series published during 4 decades.
Physicians may classify all headaches as primary or secondary.8- 10 Primary headaches are not associated with underlying abnormalities, whereas secondary headaches are due to a pathologic cause. The principal primary headache syndromes are migraine, tension-type, and cluster.
The diagnosis of primary headache syndromes rests on a carefully obtained history. Physicians may be more confident of a correct diagnosis as the number of typical clinical features increases. The history and physical examination should be sufficiently detailed to reassure the patient and the physician that no worrisome cause is present. Important historical elements include pain character, location, onset, precipitants, aura, other associated symptoms, duration, frequency, and time course. Results of the neurologic examination are normal in patients with primary headaches who are not currently experiencing a headache and do not add to the diagnostic certainty. An exception is the occasional persistence of a partial Horner syndrome among asymptomatic patients with a history of cluster headache.
The history and the results of general physical and neurologic examinations correctly identify patients who need no further diagnostic evaluation. For example, in a review of patients with migraine or unspecified headache and normal neurologic examination results, the American Academy of Neurology found the incidence of clinically important findings on neuroimaging studies to be 0.4% and 2.4%, respectively.11 These data illustrate the importance of taking a careful history to confidently exclude pathologic causes of headache.
The migraine headache syndrome is important to recognize to minimize unnecessary diagnostic studies and to identify correctly patients who will benefit from effective therapy. Migraine headaches are common and result in significant suffering and disability. Stewart and coworkers12 used the IHS criteria (Table 1) to study the prevalence of migraine in the United States. The overall prevalence was 17.5% of women and 5.7% of men. Moderate or severe disability, defined as requiring bed rest or severely restricting working ability, respectively, was reported by 46% of migraine sufferers. Other groups have reported similar results.13,14
The age at onset varies by sex. Breslau and colleagues15 studied migraine incidence in 1007 unselected individuals in the general population and reported a mean age at onset of any migraine of 13.7 years for men and 17.6 years for women. Steiner and colleagues16 found that only 8% of 479 patients with migraine experienced onset after 40 years of age. Patients with migraine-like headaches occurring for the first time after 40 years of age deserve particular diagnostic diligence to exclude other potential causes of headache.
The IHS criteria distinguish 15 separate migraine diagnoses. The following 3 migraine diagnoses are common in clinical practice: migraine without aura, migraine with aura, and migraine with typical aura. Table 2 summarizes the sensitivity, specificity, and likelihood ratios of individual headache features for the diagnosis of migraine when compared with tension-type headache. I derived these values from pooled data of published series.17- 34
The principal diagnostic features of migraine are throbbing pain, unilateral pain, duration from 4 to 72 hours, nausea, photophobia, phonophobia, and exacerbation by routine physical activity. Among these features, nausea, exacerbation by physical activity, and photophobia are the most frequent. Despite the origin of the word migraine from hemicrania, unilateral pain is the least sensitive criterion.
In practice, clinicians often must choose between migraine or tension-type headache as the correct diagnosis for a given patient. The features most specific for migraine are nausea, phonophobia, and photophobia. The positive likelihood ratios of these features for the diagnosis of migraine are 19.2, 5.2, and 5.8, respectively. The specificity and positive likelihood ratios for these 3 features are high, in part, because of restrictions imposed by the IHS criteria on the diagnosis of tension-type headache. These criteria require the absence of nausea; if photophobia or phonophobia is present, the other must be absent (Table 1). There are not enough clinical series of tension-type headache before the publication of the IHS criteria to determine if the specificity would be equally high using less restrictive criteria. Throbbing and unilateral pain, which are often thought to be specific for migraine headache, are less so, with specificities of 75% and 82%, respectively.
Nausea, photophobia, phonophobia, exacerbation by physical activity, and throbbing nature each confer a negative likelihood ratio of less than 0.4. The absence of any of these features makes the diagnosis of migraine less likely.
If one restricts the analysis to those studies that used the IHS diagnostic criteria, the sensitivity and specificity of clinical features are not substantially different from those derived from the pooled data of all studies. This finding suggests that, despite many different pre-IHS diagnostic criteria, early studies included reasonably homogeneous populations of patients with migraine and tension-type headache.
Patients with migraine headache often identify triggers for their migraine attacks. Table 3 details the sensitivity of these triggers. The most common migraine precipitants are stress, certain foods, missing a meal, and menses. Among women with menstrual triggers, migraines most commonly occur during the week before the onset of menses.34,36,42,43 Migraines during menses are the next most frequent. Less frequent precipitants include lack of sleep, perfume or odors, and weather change.
Among potential precipitants, only certain foods are significantly more prevalent among patients with migraine than among patients with tension-type headaches. Food triggers are noted by 49% of patients with migraine, for a positive likelihood ratio of 3.6. Specific food triggers include cheese, chocolate, and alcohol. Cheese is the most frequent precipitant. Chocolate is the most specific trigger and has a positive likelihood ratio of 7.1 when patients with migraine are compared with those with tension-type headache. These data are limited, however, by the relatively small series of patients with tension-type headaches.
Drummond41 looked at this question from the opposite perspective. He questioned 600 patients about self-reported precipitating and relieving factors for their headaches and included patients with migraine, cluster, and tension-type headache. Among precipitating factors, menstrual variation, glare, stress, and alcohol each significantly predicted a diagnosis of migraine.
Clinicians should seek a history of headache precipitants to guide counseling on nonpharmacologic therapy for headaches. With the exception of food triggers, however, these features are equally prevalent among patients with migraine and those with tension-type headache and are of limited diagnostic value. The pooled data support the exclusion of headache precipitants from the IHS criteria for migraine syndromes.
One third of all patients with migraine experience an aura before their headache. In a study of 4000 unselected individuals in the general population, the lifetime prevalence of migraine without aura was 14.7%; that of migraine with aura was 7.9%.46Table 4 summarizes the features of migrainous aurae based on a comprehensive review of published series.
The two important historical features of the aura are the subjective description and its duration. Visual aurae are most common. Among patients with aurae, 74% report positive visual phenomena. Most common are zigzags, fortification spectra, and stars or flashes. Negative visual phenomena are reported by 56% of patients. These include scotoma and hemianopsia. Disturbances of visual perception are the least frequent and include foggy vision, tunnel vision, the perception of looking through heat waves or water, mosaic, and, rarely, complex visual hallucinations.49
Sensory, aphasic, and motor aurae occur in 20%, 11%, and 4% of patients, respectively. One establishes a diagnosis of migraine with typical aura when one of these aurae or a homonymous visual disturbance is present.4 These aurae rarely occur in isolation. In one study, 99% of these aurae occurred in conjunction with a visual aura.48 Sensory aurae involved the hand in 96% and face in 67% of cases. Involvement of the trunk and lower extremities was uncommon. The sensory aura was unilateral in 84% of cases. Commonly, the aura began in the hand, progressed to the arm, and then the face and tongue. Motor aurae occurred in the same affected areas, and always in concert with visual and sensory aurae. Aphasic aurae included paraphasias, impaired production of language, and impaired comprehension of language.
The duration of a migraine aura is stereotypic. The IHS criteria require each aura feature to be longer than 4 minutes and shorter than 60 minutes in duration.4 In fact, 70% of visual aurae last less than 30 minutes (Table 4).
Table 5 summarizes the sensitivity, specificity, and likelihood ratios of common historical features among patients with migraine compared with patients without headache. Among patients with migraine, 58% have a family history of migraine, and the positive likelihood ratio of this finding is 5.0. These patients may underestimate the frequency of migraine in their family. When family members are interviewed, the frequency of a family history of migraine is 1.6-fold greater than that suggested by patient history alone.64
A patient's life history provides clues to the diagnosis of migraine. In 3 early studies, a history of recurrent childhood vomiting attacks existed in 30% of adult patients with migraine and only 13% of adults without migraine.30,31,56 Lanzi and colleagues57 reported the frequency of associated conditions in children with migraine. Cyclic vomiting and motion sickness were each 3.6 times more likely among children with migraine than among control subjects. In a prospective study of 968 children, those in whom disabling headaches developed by 5 years of age were 2.8 times more likely to experience motion sickness than those who did not have disabling headaches.65
These associations remain controversial because of the variability of case definitions. Authors have used terms such as the periodic syndrome, abdominal migraine, and cyclic vomiting with imprecision. The IHS has acknowledged the view that some of these presentations are childhood migraine equivalents, but has not included these syndromes among the named operational diagnoses.4 When diagnostic uncertainty exists, clinicians will find a history of childhood vomiting attacks or motion sickness helpful, as each increases the likelihood that a given patient's headaches are migraines.
Questionnaire studies provide an external validation of the pooled clinical data in this review. Rasmussen and colleagues1 developed a questionnaire that included all of the principal IHS criteria for migraine and tension-type headache and compared this instrument with a clinical interview by an experienced neurologist.66 The sensitivity, specificity, and positive and negative predictive values of the instrument for the diagnosis of migraine were 51%, 92%, 50%, and 93%, respectively. Two other published questionnaires replicated the IHS criteria for migraine and provided no insight into which clinical features were most sensitive or specific for the diagnosis of migraine.67,68
Henry and colleagues32 developed an IHS-based questionnaire and algorithm that lay interviewers used in epidemiological studies. The sensitivity and specificity for the diagnosis of migraine were 95% and 78%, respectively. In a follow-up study of 166 patients with chronic headache, duration from 4 to 72 hours, photophobia and phonophobia, unilateral location, and pulsating quality were most sensitive for the diagnosis of migraine.69 Sensitivities were 74%, 66%, 65%, and 64%, respectively. Nausea or vomiting, unilateral location, and pulsating quality were the most specific, at 93%, 85%, and 83%, respectively.
The results of these studies32,69 differ from my findings by the lower sensitivity of nausea for the diagnosis of migraine and by a higher specificity of unilateral location and pulsating quality. They classified patients without migraine as having nonmigraine headache but did not require that these patients meet IHS criteria for tension-type headache. As a result, a greater frequency of phonophobia and photophobia existed among these patients (51%) than in studies of patients with a diagnosis based on IHS criteria for tension-type headache (13% in Table 2). Despite their decision not to use the restrictive IHS criteria for the diagnosis of tension-type headache (Table 1), these researchers still found a high specificity of nausea for the diagnosis of migraine. This finding supports my conclusion, based on the use of pooled data, that nausea is a specific finding for the diagnosis of migraine.
Cluster headache is a distinctive syndrome that physicians may easily distinguish from migraine. Table 1 lists the IHS operational criteria for cluster headache. Most series of patients with cluster headaches predate the development of the IHS criteria. The most commonly used diagnostic criteria among early studies were the ad hoc criteria.70 These criteria differ from the more recent IHS criteria only by the absence of miosis, ptosis, duration, frequency, and total number of attacks.
Table 6 summarizes the sensitivity of clinical features. Cluster headache is 6 times more common in men than in women. The mean age at onset is older than that for migraine and averages 30 years.75- 77,80 In contrast to migraine, family histories of migraine or other primary headaches are present in only 26% of patients. Pain is most commonly periorbital; the next most common locations are frontal and temporal. The headache is strictly unilateral. The headache occurs only on the right side in 48% of patients, only on the left side in 38%, and alternates in 14%.
The IHS criteria require that untreated headaches last from 15 to 180 minutes, although headache duration is most commonly 30 to 60 minutes. This is distinctly shorter than a migraine attack, which most commonly lasts 4 to 24 hours. Fifty-four percent of all attacks occur during sleep. Cluster headaches derive their name from their periodicity or tendency to cluster over time. Each cluster lasts on average 1 to 2 months, and two thirds of patients experience 1 to 2 clusters per year. Trigeminal neuralgia, another cause of periorbital pain, differs from cluster headache by its much more brief duration of a few seconds and by the absence of autonomic symptoms.
Patients with cluster headache less frequently describe their pain as throbbing than those with migraine (30% vs 73%, respectively). Patients more often characterize their pain as burning, piercing, or neuralgic. It is usually excruciating and the patient is unable to find a comfortable position or find relief by lying down. Manzoni and coworkers77 noted that, in 80% of observed attacks, the patients looked restless and agitated and felt an impulse to move or change posture. Among the autonomic symptoms, ipsilateral lacrimation is most frequent. Rhinorrhea and a partial Horner syndrome (miosis or ptosis) are next most common. The most commonly reported headache triggers are alcohol and stress. However, these features are relatively insensitive and occur in only 32% and 30% of patients, respectively. These triggers are similarly common in migraine and tension-type headache (Table 3) and are therefore of no value in differential diagnosis.
Tension-type headaches are the most common chronic headaches. Using the IHS criteria, Rasmussen and colleagues1 surveyed 1000 adults from the general population and reported a lifetime prevalence of tension-type headache of 69% in men and 88% in women. Among gainfully employed individuals in the same cohort, 43% of subjects with migraine and only 12% of those with tension-type headache reported missing work in the previous year because of headache.85 However, more total lost workdays were due to tension-type headache than to migraine because of its higher prevalence. Individuals may experience more than 1 primary headache syndrome. In a study of 4000 adults, Ulrich and colleagues25 found that the 1-year prevalence of tension-type headache was the same among individuals with and without migraine.
Table 1 cites the IHS operational criteria for tension-type headache. The criteria essentially define tension-type headache as a chronic recurring headache with few or no migrainous features. As the IHS criteria are restrictive and exclude certain characteristics, studies that report the frequency of clinical features of tension-type headache run the risk of circular reasoning.
Tension-type headaches are commonly bandlike, nonthrobbing, and bilateral. Although the IHS criteria require headache duration to be from 30 minutes to 7 days, Ulrich and colleagues25 found that 82% of tension-type headaches lasted less than 24 hours. Since this duration is similar to that of migraine headaches, duration of headaches is not a useful distinguishing feature. The IHS criteria allow the presence of either phonophobia or photophobia but not both. However, both findings are uncommon in tension-type headache; each occurs in only 13% of patients. Clinicians should entertain the possibility of migraine in the presence of one of these features.
Precipitants of tension-type headache are the same as those of migraine, with the exception of food triggers (Table 3). Although the number of patients studied is small, the available data suggest that the frequency of menses as a trigger of tension-type headache is similar to that of migraine. Food triggers occur in only 14% of patients with tension-type headache and should prompt consideration of the diagnosis of migraine.
The most sensitive clinical features for the diagnosis of migraine are nausea, exacerbation by physical activity, and photophobia. When clinicians consider a differential diagnosis that includes tension-type headache, the most specific features of migraine are nausea, photophobia, and phonophobia. Among potential headache triggers, only certain foods are more common among patients with migraine than among those with tension-type headache. These include cheese and chocolate. Visual aurae predominate among patients with migraine with aura; the most common duration of aura symptoms is less than 30 minutes. Migraine is 3 times as common among women than men, and more than half of patients with migraine have a family history of migraine.
Clinicians may distinguish cluster headaches from migraine by their strict unilaterality, duration of less than 3 hours, clustering over time, and predominantly periorbital location. Ipsilateral lacrimation and rhinorrhea are the most common associated autonomic symptoms. These features are absent in migraine and tension-type headache. Cluster headache is 6 times more common in men than women.
Tension-type headache is the most common primary headache syndrome. This is a recurring bandlike pressing headache with few migrainous features. It may be throbbing or unilateral but is very rarely associated with nausea, photophobia, or phonophobia. Half of patients with tension-type headache identify stress or hunger as headache triggers.
A carefully taken history can establish the diagnosis of migraine, cluster, or tension-type headache with sufficient confidence that no need will exist for additional evaluation. Use of the IHS criteria ensures that patient populations for research studies are homogeneous. Through the use of pooled data from clinical series, I have shown that certain historical features are of particular value to establish or exclude diagnoses of primary headache syndromes.
Accepted for publication April 18, 2000.
I am indebted to Mark Aronson, MD, Erin Hartman, MS, and Russell Phillips, MD, for their review of the manuscript and helpful suggestions, and to Roger Davis, ScD, for statistical advice.
Reprints: Gerald W. Smetana, MD, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 330 Brookline Ave, Boston, MA 02215.