From the Outcomes Research and Management, Merck & Co., Inc., West Point, PA (Drs. Ng-Mak, Chen, Hu, and Mr. Ma); Primary Care Network, Inc., Springfield, MO (Dr. Cady); RTI-Health Solutions, Research Triangle Park, NC, USA (Dr. Bell).
Address all correspondence to Dr. Daisy S. Ng-Mak, 770 Sumneytown Pike, P.O. Box No. 4, MS WP39-166, West Point, PA 19486.
Background.—While treating migraine early when the headache is mild is believed to link to improved treatment outcomes, it is not clear whether patients can correctly self-identify a headache as a migraine at onset in real-world settings.
Objective.—This study aims to assess the likelihood that patients can correctly self-identify a headache as a migraine at onset, and to evaluate cues that patients use to correctly identify migraine attacks.
Methods.—Adult migraineurs were recruited from 14 headache clinics across the United States. Patients recorded their headache experiences via an electronic diary daily over a period of 30 days. On days when they experienced headaches, patients were asked to recall the types of headache they experienced at both onset and peak. Patients also identified cues for deciding whether the headache was a migraine or not. Using identification of migraine at headache peak as the criterion, we examined the sensitivity and specificity of migraine identification at onset. We employed generalized estimating equation (GEE) to evaluate factors identified at headache onset that predicted migraine identified at headache peak.
Results.—Of the 192 enrolled patients, 182 patients recorded a total of 1197 headache episodes over 30 days. At headache onset, 888 episodes were deemed by patients as migraine and 309 episodes not migraine; a majority (92%) of these early migraine identifications were confirmed at headache peak. Sensitivity and specificity of self-identification of migraine at onset were 91% and 97%, respectively. A number of factors at headache onset were predictive of a migraine identified at peak: sensitivity to light (OR = 3.1, 95% CI: 1.9–5.0), headache severity (OR = 2.0, 95% CI: 1.4–2.8), nausea symptoms (OR = 2.6, 95% CI: 1.5–4.5), and visual disturbance (OR = 2.3, 95% CI: 1.1–4.9). Patients who ruled out tension-type headache at onset were twice (OR = 2.0, 95% CI: 1.5–2.8) as likely to conclude a migraine at peak.
Conclusions.—Most migraineurs in tertiary care settings can correctly self-identify a headache as a migraine at onset. Factors such as headache severity, presence of nausea, visual disturbance, sensitivity to light, and no tension-type headache, appeared to augment the correct identification.
Controlled clinical trials testing treatment of migraine attacks early with triptans while the pain is mild have consistently demonstrated improved treatment outcomes. However, before advocating treating migraine while headache is still mild, it is important to understand whether migraine patients can recognize a headache as a migraine during the early phase of a migraine attack in real-world settings. If they can, what criteria do migraine patients use to conclude that their headache is indeed a migraine? Since migraine attacks are generally treated in an outpatient setting, identification of migraine is ultimately the critical step for patients in determining treatment needs.
While the literature of premonitory symptoms on prediction of migraine attack is scant, it points to the possibility that migraine patients can recognize migraine during the early phase of headache or possibly by utilizing migraine symptoms that emerge prior to onset of migraine. “Prodromes” or premonitory symptoms include fatigue, depressive symptoms, lack of concentration, stiff neck, and craving for certain food. Some patients experience these premonitory symptoms as early as 2–48 hours before the onset of a migraine attack. About 33% of headache clinic patients reported premonitory symptoms.1 Among headache episodes reporting premonitory symptoms, 72% were concluded as migraine.2 Nevertheless, the potential to intervene at the early phase of migraine is underscored.
The notion of treat early with triptans for acute migraine while the headache is mild was first suggested by Cady et al.3 In a retrospective analysis of protocol violators of a large, randomized, placebo-controlled study, compared with placebo, patients taking sumatriptan 50 mg without waiting for headache pain progressing to moderate or severe intensity had a significantly higher pain-free rate at 4 hours after dosing.3 Subsequent ad-hoc analyses of large trials provided additional evidence for improved treatment efficacy in which patients treated their migraine at the onset of headache when pain was mild, as compared with moderate/severe pain.4,5
Randomized placebo-controlled studies provided additional evidence for improved drug efficacy when patients treated their migraine at the first sign of pain while it was mild.6–9 Stratified analyses showed a greater proportion of pain free at 2 hours after treatment for mild attacks than for moderate or severe attacks.8 A recent randomized, double-blind, placebo-controlled study showed that pain intensity was relatively more important than timing in achieving pain free at 2 hours after dosing.10
Apart from clinical trial evidence, a prospective observational study lends support for better treatment effectiveness when the pain is mild. Reporting treatment outcomes via an interactive voice response system within 24 hours of migraine resolution, patients treated with rizatriptan 10-mg tablet or orally disintegrating tablet as soon as they experienced headache were significantly more likely to achieve pain relief after treatment,11 as compared with those who delayed treatment until headache became severe.
As a whole, an increasing body of literature shows that treating early with triptans while the pain is mild may confer better treatment outcomes. However, alternative study designs have been proposed to further support this recommendation.12,13
While the extant literature suggests that treating early while the headache is mild enhances treatment outcomes, the potential benefits should be balanced against the potential risks of medication overuse and/or misuse. Migraine patients should treat early only if they are capable of reliably self-identifying a headache as a migraine at onset. The aims of this study are 2-fold: (1) to evaluate the extent to which self-identification of a headache as a migraine at onset is consistent with self-identification of a headache as a migraine at peak; and (2) to assess headache cues/factors that are used by migraine patients to identify a headache as a migraine.
Study Design.— This study was a multicenter, observational follow-up study. Migraine patients were recruited from 14 headache clinics in the United States during the first half of 2005. At study entry, participants were asked to sign written consent and complete a baseline questionnaire. Participants were provided with a hand-held electronic diary and were asked to record their headache experiences daily over the course of 1 month. Physicians and patients were encouraged to treat as they normally would. The study was approved by the Schulman Associates Institutional Review Board, Inc.
Study Population.— Migraine patients, aged 18 years or above, with a physician-diagnosis of migraine with or without aura were eligible to participate. Eligible patients could experience 2–8 migraines per month, with a maximum of 15 headache days per month. Patients were either prescribed with or given samples of an oral triptan medication for acute migraine attacks. Patients with health conditions that have contraindications to the prescribed migraine medications, per the product's insert, were excluded from the study. Patients who were treated with a parenteral migraine medication were also excluded.
Baseline Questionnaire.— Patients completed a baseline questionnaire that captured demographic characteristics, migraine history, migraine frequency and severity, impact of migraine attacks, treatment decisions in the last 3 months, general beliefs about migraine and migraine treatment, and use of medications.
Follow-Up Diary Questionnaire.— Enrolled patients were asked to keep an electronic diary daily over the next 30 consecutive days. Each diary day captures the occurrence of 1 headache episode if applicable. Whenever a headache episode occurs, information regarding the headache onset and headache at peak times, severity and patient-classification of headache, symptoms associated with the headache, medications used, reasons for and against taking medications, pain relief strategies, disability due to headache, and patient satisfaction are documented. Nonresponse to any questions in the electronic diary was not allowable. Patients were not able to proceed to the next question without providing response to the prior question. Patients were instructed to complete the diary between 5:00 p.m. and their bedtime. An alarm was set to remind patients of completing the diary if patients failed to record diary daily. All recorded information was transmitted via a cradle connected to the central data depository center daily at night over the study period.
Measurement of Study Variables.— Patients provided answers to 2 identical sets of questions regarding certainty of migraine, headache-associated symptoms, headache severity at onset and at peak per recorded headache episode.
Migraine Identification.— Dichotomous migraine identifications at onset and at peak were created. If patients responded “probably a migraine” or “definitely a migraine” to the question “At headache onset today and before you took any medications, how certain were you that this headache was or was not a migraine?,” the variable “migraine identification at onset” was coded positive. “Migraine identification at onset” was coded negative if responses to this question were “probably not a migraine” or “definitely not a migraine.”“Migraine identification at peak” was created in a similar fashion. A correct migraine identification was defined as consistent migraine identification at headache onset and at headache peak.
Headache Pain Severity and Headache-Associated Symptoms.— For the measurement of pain severity, patients responded to this following question “How would you classify the pain severity at headache onset today?” in a 4-point Likert scale (0 = no pain, 1 = mild, 2 = moderate, and 3 = severe).
With regard to headache-associated symptoms, patients checked off any symptoms that they had at the headache onset or at the headache peak from a list of 10 symptoms (loss of appetite, nausea, vomiting, sensitivity to light, sensitivity to odors, sensitivity to noise, dizziness, sparkling/flashing/colored lights, blurred or double vision, and aura). A new variable “visual disturbance” was created by electing any positive response to the following 3 symptoms: sparkling/flashing/colored lights, blurred or double vision, and aura.
Tension-Type Headache.— At the baseline, patients checked if they also suffered from tension-type headache in addition to migraine.
Statistical Analysis.— To evaluate whether migraine patients can identify a headache as a migraine at onset, sensitivity and specificity of identifying a headache as a migraine were computed as measures of validity for migraine identification. Sensitivity is defined as the percentage of affirmative responses of migraine identifications at onset given the affirmative migraine identifications at peak. Specificity is defined as the negative responses of migraine identifications at onset given negative migraine identifications at peak.
To assess factors that are used by migraine patients to identify a headache as a migraine, univariate analysis was first performed to provide a comprehensive description of predictors, including symptoms at onset, headache severity at onset, and tension-type headache and their relationship with the correctness of migraine identification.
Statistical significance of differences of each predictor between correctly identified episodes and incorrectly identified episodes was assessed using generalized estimating equation (GEE).
The effects of migraine-associated symptoms at the onset, headache severity at the onset, and tension-type headache on migraine identification at peak were examined. Multivariate analysis was performed using GEE, taking into account of the correlated nature of repeated headache episodes within patient. Compound symmetry as the covariance structure of the GEE model was specified. Intraclass correlation coefficient, which measures the correlation among repeated headache episodes within patient was reported. Covariates such as age, gender, and education level were also controlled in the GEE analysis.
All analyses were performed using SAS statistical software V8.14
Table 1 displays the demographic characteristic for the study participants. Enrolled patients were in their productive years (mean age = 42 years, SD = 11.2), predominantly being females (85.4%). Over half of the sample (58.5%) was college or post-college graduates and 58% were full-time employees. Mean age of first diagnosed with migraine was 27 years (SD = 11.2). In addition to migraine, 58.2% of these patients also suffered from tension-type headaches.
Table 1.—. Sample Demographic Characteristics
Frequency (N = 185)
Age in years, range 18–66 years, mean (SD)
Less than 8th grade
Some high school
High school graduate
Employed full time
Employed part time
Unable to work because of migraine
Unable to work because of other health reasons
Age first diagnosed with migraine, years: mean (SD)
Suffer from other types of headaches
Tension type headaches
A total of 192 migraine patients enrolled in this study. Seven patients did not record any diary and 3 patients did not have headache at all during the 30-day follow-up period. One hundred eighty-two patients recorded 1197 headache episodes via an electronic diary over 30-day study period and 970 (81%) of these episodes were deemed as migraine headaches identified at peak. The distribution of migraine attack frequency is shown in Figure 1. Over the study period of 30 days, patients experienced on average 5.3 migraine episodes (SD = 4.3). Nine patients reported 15 or more attacks (total number of attacks = 162) during the 30 days follow-up period. The likelihood of migraine episode occurred within 24 hours of other episodes was low (4.8%).
Migraine identification at peak was used as a criterion for making definitive migraine identification. As shown in Table 2, at headache onset, 888 episodes were judged by patient as migraine and 880 (99%) of these early migraine episodes were confirmed at peak. In 309 headache episodes, which were deemed as nonmigraine at onset of headache, 219 (71%) were confirmed by patients as nonmigraine episodes at headache peak. Overall, 1099 headache episodes (92%) of these early migraine were confirmed at headache peak. The sensitivity and specificity of migraine identification at onset were 91% and 97%, respectively. Providing equal weights to all contributing patients, additional analysis pertaining to the validity measures across the first headache attacks was conducted. Among 182 first headache attacks, 128 out of 129 (99.2%) headaches that were self-identified as migraines at onset were confirmed at peak. Thirty-five out of the 53 (ie, 66%) headaches that were self-identified as migraines at onset were confirmed at peak. The sensitivity and specificity of self-identification at onset among the first headache attacks were 97% and 88%, respectively.
Table 2.—. Migraine Identification at Onset and at Peak
Migraine Identification at Peak
Sensitivity of migraine identification at onset = (880/970) × 100%= 91%.
Specificity of migraine identification at onset = (219/227) × 100%= 97%.
Across all headache episodes, sensitivity and specificity of self-identification at onset were also examined in 2 subpopulations: patients with or without tension-type headaches (data not shown). Sensitivity of self-identification of migraine at onset was larger for migraineurs without tension-type headaches (95%) than patients with tension-type headaches (86%). Patients of either subpopulation were similarly well in identifying nonmigraine at the onset (specificity measures for patients with tension and without tension type headaches were 97% and 96%, respectively).
Potential factors that differentiated the correctness of migraine self-identification are shown in Table 3 and Figure 2. Patients' characteristics with respect to age, age at first migraine diagnosis, education, and employment status were similar for both correctly self-identified (episodes = 1099) and incorrectly identified episodes (episodes = 98) (data not shown). No tension-type headaches were more likely (OR = 2.4, 95% CI = 1.4–4.1) to be associated with correctly identified episodes, as compared with incorrectly identified episodes. Severe headache pain endured at onset raised the likelihood of correctly identifying migraine episodes (OR = 2.6, 95% CI = 1.8–3.8).
Table 3.—. Associations Between Absence of Tension-Type Headache and Headache Severity at Onset With Accuracy of Migraine Identification
95% CI for OR
aGeneralized estimating equation was performed to test for statistically significant difference between episodes of correct and incorrect migraine identification.
No tension headache
Headache severity at onset
Across all comparisons of headache symptoms at the onset, with the exception of vomiting, correctly identified episodes were associated with significantly more migraine-associated symptoms such as loss of appetite, nausea, visual disturbance, sensitivity to light, sensitivity to noise, sensitivity to odor, and dizziness than incorrectly identified episodes (Fig. 2).
Frequencies of migraine-associated symptoms at onset and at peak are presented separately for correctly identified episodes (Table 4, left column) and incorrectly identified episodes (Table 4, right column). In general, migraine-associated symptoms were reported more frequently at peak than at onset. When the magnitude of migraine-associated symptoms difference between peak and onset was expressed in ratio (ie, peak/onset: the closer to 1, the smaller the difference), among correctly identified episodes, migraine-associated symptoms presented at peak were most likely also presented at onset (ie, ratio ranged from 1.1 to 2.0). In contrast, among incorrectly identified episodes, migraine-associated symptoms were more likely to be presented at peak than at onset (ie, ratio ranged from 2.0 to 3.3.). The same patterns were observed when the data were analyzed in patients experiencing 4 or more episodes, which suggested good internal consistency (data not shown).
Table 4.—. Frequency of Headache Symptoms at Onset and at Peak by Correctness of Identified Episodes
Correctly Identified Episodes
Incorrectly Identified Episodes
At Onset Episodes = 1099 (%)
At Peak Episodes = 1099 (%)
Ratio = Peak/Onset
At Onset Episodes = 98 (%)
At Peak Episodes = 98 (%)
Ratio = Peak/Onset
Loss of appetite
Sensitivity to light
Sensitivity to noise
Sensitivity to odors
Table 5 shows factors/cues at onset of headache that were significantly associated with migraine identification at peak as determined by patients. Patients who experienced sensitivity to light at onset were about 3 times more likely to identify migraine at peak (OR = 3.1, 95% CI = 1.9–5.0). Those who had nausea (OR = 2.6, 95% CI = 1.5–4.5) or visually disturbed (OR = 2.3, 95% CI = 1.1–4.9) or experienced severe headache pain at onset (OR = 2.0, 95% CI = 1.4–2.8) doubled the likelihood of identification with migraine at peak. Patients without histories of tension-type headache were twice as likely as tension headache suffers to reach a migraine identification at peak (OR = 2.0, 95% CI = 1.5–2.8). Additional analysis controlled for age, gender, and education did not alter these estimated effects (data not shown). Intraclass correlation coefficient of the estimated model was 0.30.
Table 5.—. Multivariate Analysis* of Factors/Cues at Headache Onset in Predicting Migraine at Peak
Patient Identified Cues
95% CI for OR
*Generalized estimating equation procedure was performed to account for the correlated nature of repeated headache episodes within each patient.
OR = odds ratio; CI = confidence interval.
Sensitivity to light
No tension headache
In this study, patients were able to predict the pathogenesis of symptoms culminating in a period of peak symptomatology that may be indicative of a migraine diagnosis. Our findings underscore the fundamental point of patient-centered care in migraine treatment as advocated by Cady and Dodick.15 With adequate physician–patient communication, educated migraine patients can identify migraine at the onset of headache. A patient-centered stratified care approach, which emphasizes individualized treatment protocol according to the characteristics of a patient's headache attack should be advocated.15
Physicians diagnose patients with migraine by applying the diagnostic criteria developed by the International Headache Society (IHS).16 We considered patient-identified migraine at peak intensity of headache for each headache attack as an internal gold standard. Pain intensity at headache peak can be argued as one of the most relevant and reliable benchmarks for identifying a headache as a migraine for each migraine attack. Once headache sufferers received a migraine diagnosis by their physicians, migraine patients are the ultimate decision makers for migraine identification and medication use for each headache attack.
Our data showed that the vast majority of migraine patients encountered in headache clinics were able to identify a headache as a migraine at onset as indicated by high sensitivity (91%) and specificity (97%) of migraine identifications. Community-based estimates of sensitivity and specificity of self-identification of migraine were much lower: 53.4% and 83.8%, respectively. About half of the IHS-diagnosed migraineurs were self-identified with migraine.17 The difference in the validity estimates is largely due to the divergent sample characteristics. The knowledge gap regarding migraine self-identification between community-based migraineurs and clinic-based migraineurs may be narrowed by appropriate physician-patient education.
Migraine patients should treat early only when they experience a migraine attack. However, in a recent report, 69% of migraine patients would rather wait and delay treatment until they are sure of a migraine attack.18 Our study results showed that headache pain experienced at onset of migraine serves as a good time window of opportunity to intervene.
We identified 3 specific migraine-associated symptoms at headache onset, namely, sensitivity to light, nausea, and visual disturbance, which were predictive of migraine at peak. These headache symptoms at onset have consistently been reported as important factors for migraine self-identification16,19–21 and are the most commonly reported symptoms in migraine community.22 Recently, Mulleners and colleagues19 concluded that self-reported sensitivity to light among migraine patients was a reliable diagnostic predictor for migraine. According to category D of the IHS's alternative method of migraine diagnosis, 2 out of 5 symptoms (ie, nausea, vomiting, photophobia, phonophobia, and osmophobia) experienced during a headache attack are sufficient to be classified as migraine.16 Kelman20 validated this alternative classification method in which 98.2% of his tertiary care patient population was classified as migraine patients. Nausea is a consistent and reliable predictor of migraine. Martin et al21 concluded that nausea is the only 1 variable that met criteria for both 1-variable optimal and 3-variable optimal models for migraine diagnosis.
Many migraine patients suffer from more than 1 type of headache pain. A recent Danish population survey shows that about 92%–94% of migraineurs, aged 25–36 years, also suffer from tension-type headache.23 In our study population, 58% of patients suffered from tension-type headache. The difference in the comorbidity rates between our study and the Danish survey is largely due to the restricted age range in the Lyngberg and colleagues' study. With dual ailments, patients can actually differentiate between migraine and other types of headache.24
Some caveats in the interpretation of our findings are noted. Our study patient population was recruited from headache clinics. They are “sophisticated” migraine patients who may be more educated on migraine identification and treatment strategy than other patients from primary care physician offices. Hence, they may be more capable in distinguishing a migraine versus a nonmigraine headache episode. Our study findings, therefore, may not be generalizable to other patient populations in nontertiary care settings.
Prospective nature of the study design was embedded into the retrospective recording of headache experience. Patients did not record real time headache incidences. They were asked to record their daily headache experience anytime during 5:00 p.m. and bedtime. Retrospective reconciliations of headache experiences and headache symptoms might artificially produce congruent scenarios about their migraine identification. This may potentially overestimate the sensitivity and specificity of migraine identification at headache onset.
As part of the inclusion criteria, we did not use the strict IHS diagnostic criteria for recruiting patients. We enrolled physician-diagnosed migraine patients from headache clinics. Using a rather loose migraine diagnosis as an inclusion criterion might have introduced an overrepresentation of false-positive migraine patients. An imminent consequence of flooding the sample with false-positive migraine patients would have lowered the patients' ability to self-identify migraine, and thus might have reduced the sensitivity and specificity of migraine identification. This potential bias is, however, unlikely for 2 reasons. On average, our patient sample had endured migraine for over 15 years. Also, all physicians involved in the study were headache specialists. It is highly unlikely that these patients were misdiagnosed for such an extended period of time. Misdiagnosis, if any, would have been corrected under the care of a headache specialist.
An overarching goal of this study was to understand patients' perspectives on treatment decisions and its rationale. We did not study treatment outcomes in relation to timing of medication use and other clinical endpoints. In hindsight, we would have learned more about the effects of patient's treatment decisions and strategy on treatment outcomes if we have had included treatment outcomes in the study protocol.
Nevertheless, advantages of this study warrant discussion. This is an observational study designed to understand patients' headache experience over a consecutive 30 days. Without interfering with patients' decisions about treatment options, we gathered candid patient perspective regarding migraine treatment. To our knowledge, this is the first observational study designed to observe patients' decisions regarding migraine treatment over time.
The use of an e-diary safeguards data accuracy and integrity. All information was collected daily via an electronic diary. Once the data were recorded and transmitted to the data center, patients could not modify their responses to the diary questions. With the use of preprogrammed hand-held organizer, missing information was practically nonexistent.
The validity of patient-reported migraine identification at onset was enhanced by the identical measurement of migraine identification at onset and at peak. The high sensitivity and specificity of the measurement of migraine identification at onset ensures the validity of this measurement.
In conclusion, the vast majority of migraine patients in tertiary care settings are capable of self-identifying a headache as a migraine by evaluating their migraine-associated symptoms at onset (sensitivity to light, nausea, and visual disturbance), headache severity at onset, and ruling out tension-type headache at onset. For migraine patients who can identify a headache as a migraine at an early phase of an attack, they would benefit by treating early while the pain is mild for their acute migraine attacks.
Acknowledgments: The authors would like to thank Kathleen Foley, PhD, for her contribution in the study design.
Conflict of Interest: This research study was funded by Merck & Co., Inc.
Drs. Ng-Mak, Chen, and Hu and Mr. Ma are employees of Merck & Co., Inc.
Mr. Bell was a former employee of Merck & Co., Inc.
Dr. Cady was one of the investigators of this research study.