Pediatric migraine and the International Headache Society (IHS) criteria

Migraine is one of the most common causes of headaches in children and adolescents resulting in missed school days and in disruption of other activities. In epidemiologic studies, the overall prevalence of migraine in children ranges from 3.2% to 10.6% depending on the diagnostic criteria, the age distribution of the study sample, and other methodologic features. ^[1-4] With proper diagnosis there is an enormous opportunity for early and effective treatment. Given the high prevalence and the substantial individual and societal cost of pediatric headache disorders, validating an appropriate system for their diagnosis and classification is an important priority for clinical practice and clinical research. ^[5,6]

Over the last three decades there have been several proposed definitions of pediatric migraine. Vahlquist, ^[7] followed by Bille, ^[8] defined migraine as paroxysmal headaches separated by symptom-free intervals and accompanied by at least two or four of the following features: (1) unilaterality, (2) nausea, (3) visual aura, and (4) family history of migraine. In 1962, the Ad Hoc Committee on Classification of Headache provided a description of migraine but did not specify which features had to be present to make a diagnosis. ^[9] Since that time, researchers such as Prensky and Sommer, ^[10] Congdon and Forsythe, ^[11] and others ^[12,13] have proposed minor revisions to the Vahlquist criteria.

In 1988, the International Headache Society (IHS) proposed a new set of criteria for migraine headaches based on international expert consensus (Table 1). ^[14] Though the IHS criteria represent a seminal advance, they were developed primarily for headache disorders in adults. Their only modification for pediatric migraine was to reduce the duration requirements from 4 to 72 hours in adults to 2 to 48 hours for children under age 15 years. Though widely used in clinical research, the IHS criteria may be too restrictive for pediatric clinical practice. ^[15,16] Revising the criteria to account for possible developmental differences encountered in children and adolescents with migraine may increase their utility. ^[15,16]

The objectives of this study were to assess the validity of the IHS criteria for migraine without aura in children and, if necessary, to suggest revisions to the definition. We emphasize migraine without aura because migraine with aura is characterized by distinctive episodes of focal neurologic dysfunction that are relatively easy to recognize. We used the clinical diagnosis of pediatric neurology headache specialists as the gold standard based on the theory that the case definition should correspond, at least in part, to the views of the specialist. We hoped that pediatric headache specialists would use diagnostic rules that were sensitive to the high degree of variation in migraine characteristics and yet sufficiently specific to capture a valid subgroup of headache sufferers.

Methods.

We reviewed the records of 253 pediatric headache patients consecutively evaluated by pediatric neurologists at the Montefiore Headache Unit from 1987 through 1991. Eligible subjects had to be aged 18 or less at the time of their initial consultation. A detailed headache history was obtained by the physician using a semi-structured interview and a standardized case record. The symptoms recorded in the medical record were used to assign an IHS-based diagnosis of migraine without aura using a diagnostic algorithm. Clinical diagnoses assigned by the physician were used as the gold standard for evaluating the IHS diagnoses. The number of headache attacks, duration of headaches, pain features, associated symptoms, diagnostic work-up, and treatment response were noted for each patient. After eliminating patients with migraine with aura (n = 27) and missing data (n = 59), the sample was reduced to 167 subjects. For this group, we divided both IHS and clinical diagnosis into two groups-migraine without aura and all other headache diagnoses.

The sensitivity, specificity, positive predictive value, and positive likelihood ratio of the IHS-based case definition were calculated using the pediatric neurologist's diagnosis as the gold standard. We also calculated the sensitivity and specificity of individual symptom features and alternative symptom-based case definitions. We compiled likelihood ratios for alternative symptom-based case definitions. Table 2 shows the formulae used to compute these statistics. Sensitivity here refers to the proportion of patients with medically diagnosed migraine without aura who met the IHS criteria. Specificity refers to the proportion of patients who have not been medically diagnosed with migraine without aura and who did not meet the IHS criteria. Positive predictive value refers to the proportion of patients who meet the IHS definition of migraine who actually have the disease. True positives are the number of patients with medically diagnosed migraine without aura who met the IHS criteria (see a in Table 2). A positive likelihood ratio for migraine is the odds of meeting the IHS migraine (without aura) criteria among those patients with a clinical diagnosis of migraine compared with the odds of meeting the IHS migraine criteria among patients without a clinical diagnosis of migraine. True negatives are the number of patients who have not been medically diagnosed with migraine without aura and who did not meet the IHS criteria (see d in Table 2). Similar statistics were computed using the symptom features in place of the IHS-based case definition.

Exacerbation of pain by routine physical activity, one of the IHS-defining symptoms (see Table 1) was not uniformly ascertained. As a consequence, we implemented two IHS definitions. Definition 1, termed the "IHS-restrictive," requires two of three pain features, effectively assuming that no one has exacerbation with physical activity (Table 3). The second definition, termed the "IHS-loose," requires one of three pain features (see Table 3), effectively assuming that everyone has the pain feature. Using this approach, we can estimate the upper and the lower bounds of sensitivity and specificity (see Discussion). This approach was used by Solomon and Lipton in a previous report. ^[17]

Results.

In our final sample of 167 children and adolescents, the mean age was 13.1 years (range, 6 to 18, SD 3.3). Eighty-eight children (52.7%) received the clinical diagnosis of migraine without aura, and 79 (47.3%) were in the other headache group. The migraine-without-aura group was 49% females. The other headache group was 54.4% females. Diagnoses in the other headache group included episodic and chronic tension type headache (37.1%) and others (10.2%).

Discussion.

We compared the IHS and clinical diagnoses of migraine without aura in a patient sample with primary headache disorders (excluding migraine with aura) recruited from a subspecialty center. Using the restrictive IHS definition, the overall level of agreement of 53% is comparable with the 44 to 61% level of agreement in prior reports. ^{[15,16,18,19]} We also evaluated the sensitivity and specificity of the IHS criteria using the clinical diagnosis as the gold standard. Though the excellent specificity (92.4%) was encouraging, the sensitivity was very low. A minority of children diagnosed by a pediatric neurologist met the IHS criteria even using our IHS-loose definition. Estimated sensitivity of the IHS criteria in other studies ranged from 47 to 53%. ^[15,19]

The low sensitivity of the IHS criteria might arise if the study clinicians were excessively free in diagnosing migraine or if the IHS criteria are excessively restrictive. ^[19] From a clinical perspective, we would say that clinicians were excessively free if their diagnoses did not predict natural history or treatment response. We would say that the IHS criteria were excessively restrictive if many patients who should be diagnosed with migraine, who responded to migraine therapy, did not meet the case definition. The IHS criteria were developed to ensure the selection of homogeneous samples for research, but the restrictive nature of these criteria decreases their utility in clinical practice, where a clinician must assign a diagnosis as a prelude to treatment. ^[20] There are other difficulties in applying strict criteria in childhood when prostration, nausea, repeated vomiting, and abdominal discomfort may overshadow the complaint of headache. ^[21] In addition, young children and their parents may not be able to describe the pain or the associated features of the attacks in the terms required by the IHS criteria.

Several investigators compared IHS diagnoses with diagnoses assigned based on other sets of criteria. Metsahonkala and Sillanpaa ^[22] compared the IHS definition of migraine with six other standard sets of criteria including those of Vahlquist. ^[7] They found that 95 children had IHS migraine, 106 had Vahlquist migraine, and 83 had both. They concluded that the IHS definition was more restrictive than the Vahlquist definition though there were bidirectional differences in classification. Other studies reached similar conclusions. ^[2,19,23]

The IHS definition is longitudinal in that at least five lifetime attacks are required. In this study, 98.9% of medically diagnosed migraine patients fulfilled this requirement. Reducing the number of headache attacks required for the definition of migraine did not significantly change the sensitivity of the case definition. Similar results are reported by Wober-Bingol et al. ^[23] as 95.2% of their patients with migraine without aura had five or more headache episodes when diagnosed. In referral samples, patients are likely to have had multiple attacks prior to seeing a headache specialist but criterion would be more problematic in primary care setting.

The IHS criteria permit a shorter headache duration for patients under 15 years (2 to 48 hours) than for those aged 15 years or over (4 to 72 hours). In our patient population with medically diagnosed migraine without aura, 28 (32%) had headaches lasting less than 2 hours, and in 8 (28%), headaches lasted less than 1 hour (4 of these patients would have been otherwise diagnosed as having migraine). Bille ^[8] reported that 27% of children with migraine had some headaches with a duration of less than an hour. Mortimer et al. ^[2] found that 11% of their migraine cases had duration of less than 2 hours. Metsahonkala and Sillanpaa ^[22] reported that 81% of their patients had headache duration of less than 2 hours, and 35% had headache duration of less than an hour. Similar results were reported by others. ^[15,18,23] These results led some authors to argue that the minimum headache duration required in the IHS criteria should be reduced. ^[15,18] Based on our findings and prior report, ^[18] we support the suggestion of Winner et al. ^[15] to decrease the minimal duration requirement to 1 hour. Our results indicated that further decreases in the minimal duration produced small gains in the sensitivity but reduced the specificity of the definition (see Table 3).

The utility of unilaterality as a diagnostic criterion has also been challenged. Many authors report that unilateral pain is less common in pediatric migraine. ^{[15,18,19,23]} For example, Metsahonkala and Sillanpaa ^[22] reported that 67% of their migraineurs had unilateral pain. Gallai et al. ^[18] found that 41.5% of their IHS migraineurs had unilateral pain. In our sample, we found that unilateral pain had a sensitivity of 34% and a specificity of 86% (see Table 4). Although bilateral headache is common in children, unilaterality has a positive predictive value for migraine of 85%. For this reason, we believe that unilateral pain should remain a migraine-defining feature in children. The limitation in sensitivity is offset by making it one of several alternative pain features.

Other migraine characteristics such as intensity and pulsatile quality of pain are sometimes difficult to ascertain in pediatric patients. In our study, pulsating quality had a sensitivity of 70.5%; sensitivity reported in the literature varied from 36.7 to 86.3%. ^[3,18,22,23] Similarly, the sensitivity of moderate to severe intensity in our study was 71.6%, whereas the reported sensitivity varied from 57.8% ^[18] to 97.1%. ^[3] We believe that these pain features should remain part of the IHS definition in children as long as they are accurately ascertained. Associated symptoms such as nausea, vomiting, and photophobia had good sensitivity and specificity. Phonophobia, with or without photophobia, was deemed important because of its high specificity and positive predictive value; its low sensitivity can be offset by using it as one of several alternative features.

None of the alternative case definitions we tested provided high sensitivity, specificity, and positive predictive value. Definitions 1 and 2 (Table 3) provide substantially higher sensitivity than the IHS at the price of specificity. Definition 6 provides a substantial gain in sensitivity with an enormous compromise in specificity. None of these criteria accurately replicate the decisions of the study clinicians.

Several factors limit our study. We did not uniformly collect data on pain exacerbation during routine physical activity. In other studies, this feature occurs in 29 to 81% of patients. ^{[16,18,22,23]} Many young patients may not be able to clearly report this feature. Our restrictive IHS definition assumed that no one had this feature and yielded sensitivity of 27%. Our loose IHS definition assumed that everyone had this feature and yielded an increase in sensitivity of 39% with little influence on specificity. Accurate ascertainment of this feature would predict only a modest improvement in classification. Failure to ascertain this feature had a limited effect on the present results. The study is also limited by the selection bias that leads pediatric patients to consult our headache center. Results need to be reassessed in general pediatric and population-based settings. Finally, our methodology is dependent on the diagnostic skills of the pediatric neurologists who work at our center.

We conclude that the IHS criteria for the diagnosis of migraine without aura have poor sensitivity and high specificity when applied to a sample of children using the clinical diagnosis as a gold standard. Several IHS-defining features, such as minimal duration of 2 or 4 hours, unilateral pain, or phonophobia, were present in a small proportion of pediatric patients. Developmental differences in headache characters that change and evolve as children age may limit the utility of adult criteria. Reporting difficulties and clinical heterogeneity may limit the use of any symptom-based system.

In the absence of a gold standard, diagnostic validity of migraine must be approached by searching for convergent evidence. Optimal diagnostic criteria depend upon the diagnostician's objectives. Our candidate criteria for migraine without aura (definitions 1 and 6) better correspond to pediatric clinical practice, are relatively simple to use, and are more inclusive than the IHS criteria. Despite these improvements, these definitions are still inadequate in sensitivity (definitions 1 and 6) and in specificity (definition 1). Future work might focus on alternative case definitions based on behavior rather than symptoms (i.e., going to a dark room, squinting, or turning off the lights instead of photophobia). Inclusion of additional or alternative diagnostic features (pallor, flushing, sensitivity to odors) may be helpful. Finally, restrictive criteria for use in research may be supplemented by less restrictive criteria for use in practice.