A critical appraisal of the International Classification of Headache Disorders Migraine Diagnostic Criteria based on a retrospective multicenter cross-sectional headache registry study in youth

Abstract

Objectives:

We used Cluster Analysis of Migraine-associated Symptoms (CAMS) to critically evaluate current ICHD-3 migraine-associated symptoms criteria.

Background:

Diagnostic criteria play a central role in guiding clinical trial inclusion, and therefore available treatments. Migraine and tension-type headaches (TTH) are differentiated in the International Classification of Headache Disorders (ICHD-3) by many headache characteristics, including associated symptoms. A diagnosis of probable migraine indicates some but not all features of migraine are met. Photophobia and phonophobia, or nausea and/or vomiting, are required to meet a diagnosis of migraine; however, CAMS – a model that describes associated symptoms across youth with headache – indicates that a broader range of symptoms contain information about migraine burden.

Methods:

In this multisite retrospective cross-sectional study, we evaluated ICHD-3 migraine criteria. Youth ages 6 to 17 years with migraine (including probable migraine) or TTH were included in the analysis. We used CAMS to evaluate the migraine-associated symptom criterion. With CAMS as a guide, we evaluated how changes to the migraine-associated symptom criterion altered who met the diagnosis of migraine.

Results:

Of the 9,017 participants included in this study, 66.7% were female and had a median age of 13 years [IQR 10, 15]. Most participants had migraine or probable migraine (99.0%), and the remainder had TTH (1.0%). A sizable percentage (10.1%) of youth under the umbrella diagnosis of migraine were diagnosed with probable migraine because they did not meet migraine-associated symptom criterion D; however, many in this group reported several non-ICHD migraine-associated symptoms. We explored alterations to criterion D based on CAMS. Allowing for photophobia or phonophobia recategorized 55.6% of youth as having migraine, though some only had one symptom. Including lightheadedness or lightheadedness and spinning recategorized 19.7% and 25.8% of youth with migraine, respectively, but all of those who were recategorized had at least two migraine-associated symptoms.

Discussion:

The ICHD-3 captures the most prevalent migraine-associated symptoms; however, many youth with probable migraine who do not meet full criteria due to insufficient associated symptoms nonetheless experience multiple non-ICHD migraine-associated symptoms. Changes to criterion D should be considered for the ICHD-4.

PLS:

In order for clinicians to make a diagnosis of migraine based on International Classification of Headache Disorders - Third Edition (ICHD-3) criteria, patients must have headache-associated nausea, vomiting, or light and sound sensitivity (Criterion D). However, migraine can also be associated with other symptoms like lightheadedness, difficulties with thinking, and vertigo. In this study of youth ages 6–17 with migraine, probable migraine, or tension-type headache, we looked at how altering the ICHD-3 Criterion D changed participants’ diagnosis based on their symptoms; overall, our results suggest that the next version of ICHD should account for the fact that many youth who do not experience nausea, vomiting, or light and sound sensitivities still experience many other migraine associated symptoms.

INTRODUCTION

The current International Classification of Headache Disorders (ICHD-3) differentiates tension-type headache (TTH) from migraine based on pain characteristics and the presence versus absence of associated symptoms: nausea and/or vomiting or photophobia and phonophobia.¹Yet, it has long been recognized that many other commonly reported symptoms of migraine are not included in diagnostic criteria. This is purposeful: the ICHD has been simplified over time for ease of use across multiple contexts;² however, the selection of associated symptoms for diagnostic criteria is largely based on expert opinion. Additional symptoms, including lightheadedness, fatigue, and difficulty thinking, have been suggested for ICHD,³ and the debate continues over whether there is sufficient evidence for vestibular migraine (currently in the appendix) to be considered a migraine subtype.^1,4–6 Furthermore, the ICHD symptoms that differentiate migraine from TTH are not treated with equal weight: nausea or vomiting alone meet criteria for a migraine diagnosis, while both photophobia and phonophobia are needed, and presence of only one of these symptoms can still meet criteria for TTH.

Large headache registries now provide data that can be used to refine ICHD criteria. We have developed and validated Cluster Analysis of Migraine-Associated Symptoms (CAMS) – a method using multiple correspondence analysis (MCA) to uncover relationships between an expanded set of commonly reported migraine-associated symptoms.^7,8 We found that youth with headache reported high rates of non-ICHD migraine-associated symptoms, including lightheadedness, difficulty thinking, vision changes, and spinning sensation. Further, we observed a separation of ICHD and non-ICHD symptoms. Clustering of sensory hypersensitivities and vestibular symptoms was also observed.

In this study, we used CAMS—a model developed from all youth with headache regardless of primary headache disorder diagnosis to provide a composite description of associated symptoms—to critically evaluate the migraine-associated symptom criterion based on the ICHD-3. We hypothesized that using data from multisite headache registries could refine ICHD criteria in a data-driven manner to improve diagnostic criteria.

METHODS

This was a cross-sectional retrospective study based on Children’s Hospital of Philadelphia (CHOP) and Cincinnati Children’s Hospital Medical Center (CCHMC) headache registry data. This post hoc analysis was motivated by findings in our multisite validation CAMS study,⁸ which showed relationships between an expanded set of migraine-associated symptoms. All the participants included in this study were in the CAMS multisite validation study.

Participants:

Youth between the ages of 6 and 17 years who started a headache intake questionnaire between November 2022 and early December 2023 from CHOP general neurology and headache clinics, and intake responses from the CCHMC pediatric headache clinic from 2010 until June 2023 were included. The inclusion timeframe was short for CHOP because form completion greatly improved after November 2022 due to quality improvement initiatives.⁹ Reliability of the CHOP headache registry has been previously reported; that dataset overlapped with the current study by two months (November 2022 – December 2022).⁹ A prior single center study at CHOP did not overlap with this dataset, though some of these data were included in the supplemental materials.⁷

CHOP survey data were collected and managed in REDCap (Research Electronic Data Capture)^10,11 hosted by the institution. Each institution’s Institutional Review Board approved the extraction of the data from the electronic health record into a research registry, with a waiver of verbal or written informed consent and assent to maximize generalizability. Race and ethnicity data were compared to 2020 Pennsylvania and Ohio census data, respectively.^12,13 Youth were included if they had a diagnosis of migraine (including probable migraine) or TTH and had data to assess ICHD-3 criteria A – D for migraine diagnosis. Diagnoses were collected at CCHMC as part of their questionnaire database, while CHOP diagnoses were based on the patient response algorithm.⁹ Those with new daily persistent headache or secondary causes of headache (e.g. post-traumatic headache or headache related to chronic health conditions) were excluded. We also excluded participants if ICD codes entered on the problem list or visit diagnosis suggested another headache diagnosis (e.g. post-traumatic headache, new daily persistent headache). We did not include those with an undefined headache because the primary goal of this paper was to determine how many participants with a diagnosis of probable migraine did not meet full migraine criteria due to insufficient associated symptoms.

CAMS:

The CAMS model developed in our multisite validation study⁸ was used to calculate CAMS scores across youth with TTH, probable migraine, and migraine. Importantly, the model was developed from all comers with headache. Matlab (Mathworks, Natick, MA) was used to produce custom written code to perform CAMS.¹⁴ CAMS uses multiple correspondence analysis (MCA) based on an indicator matrix using publicly available code¹⁵ to assess associations between the presence or absence of 11 migraine-associated symptoms (photophobia, phonophobia, osmophobia, nausea, vomiting, lightheadedness, spinning, difficulty thinking, vision changes, and ear ringing).⁷ MCA reduces the dimensionality of the variance contained in binary categorical data in Euclidean space,¹⁶ similar to principal component analysis for continuous data. Since those with TTH were a small percentage of the sample used in the generation of CAMS, we developed a separate CAMS model for participants with TTH matched by age, sex, and race to participants with probable migraine and migraine to demonstrate the stability across different headache diagnoses (supplemental materials).

ICHD-3 Migraine Criteria:

Diagnosis that was used for study inclusion was made by a headache specialist at CCHMC, and from an ICHD-3 based algorithm applied to patient responses at CHOP.⁹ Headache features reported in the questionnaires were used to determine if ICHD-3 migraine criteria A-D were met (see Table 1).

Table 1:

ICHD-3 criteria and determination of whether ICHD-3 criteria were met by site.

	Criterion A	Criterion B	Criterion C	Criterion D
ICHD	At least 5 migraine attacks	Attack duration of at least 2 hours	≥ 2 of the following: (1) non-holocephalic (bilateral allowed), (2) pulsating quality, (3) moderate/ severe intensity, (4) exacerbated by or causing avoidance of routine physical activity	photophobia and phonophobia, or nausea or vomiting associated with headaches
CCHMC	HA frequency of >4 days/month, OR 1 day/week or greater, OR ≤ 3/month and started prior to current age.	Longest headache duration ≥ 2 hours	Non-holocephalic Usual HA severity moderate or severe Throbbing, pulsating, pounding Exercise worsens or triggers HA	Light/sound sensitivity occur with headache Nausea associated with headache Vomiting associated with headache
CHOP	” How many headaches have you had in your life.” “Many” indicated 5 or more	Headache lasts hours or longer (categories included “seconds”, “minutes”, “hours”, “1 – 3 days”, “more than 3 days”. *Those greater than 3 days were used to capture status migrainosus	Non-holocephalic Usual HA severity moderate or severe Throbbing, pulsating, pounding Exercise worsens or triggers HA	Light/noise triggers or light/sound sensitivity occur with headache Nausea associated with headache Vomiting associated with headache

CHOP clinician-corrected questionnaires:

Differences noted between CCHMC and CHOP were hypothesized to be due at least in part to the strategy of data collection: CCHMC collected clinician-guided responses, while CHOP collected unedited patient responses. To test this, a subset of questionnaires from CHOP was used where the clinician (either general neurologist or headache specialist) updated or clarified a part of the form. These clinicians tended to show high rates of form use. Our prior work on the CHOP headache registry assessing provider form use of patient questionnaire responses indicated that 20.9% updated answers on the form, providing a clear indication that they reviewed the form. Based on our prior study, providers who used the form were more likely to be part of the headache program and practice in a suburban satellite location than the urban main campus location.⁹

Updated ICHD-3 Migraine Criteria:

Based on CAMS, we developed three changes to the ICHD-3 migraine-associated symptoms (criterion D). We then assessed how these alterations impacted diagnostic categorization of youth with probable migraine who met all ICHD-3 criteria for migraine except criterion D. Alternatives to criterion D are summarized (Table 2).

Table 2:

Alternate criterion D and rationale.

	Alternate 1	Alternate 2	Alternate 3
Criterion	Photophobia OR phonophobia or nausea or vomiting	2 of 3: photophobia, phonophobia, lightheadedness; or nausea or vomiting	2 of 4: photophobia, phonophobia, lightheadedness, spinning; or nausea or vomiting
Rationale	Photophobia and phonophobia were the most common symptoms reported, and highly likely to co-occur based on CAMS. However, comparison of patient response and clinician response charts, it was common for a patient to not report both symptoms, but for a clinician to identify both symptoms on their assessment.	Lightheadedness was added since it was the most common non-ICHD symptom based on CAMS dimension 2.	Spinning was added since it was the most common vestibular symptom based on CAMS dimension 3.

Statistical Analysis:

No a priori sample size calculations were done. Matlab (Mathworks, Natick, MA) was used to produce custom written code to perform statistical analysis.¹⁴ Age was rounded to number of birthdays and was reported as median values with interquartile ranges given non-normally distributed data. Categorical variables were reported as proportions. Univariable binomial regression analysis was used to compare age, sex, race, ethnicity, and site across those excluded due to missing data to those included in the final analysis. These variables were chosen because they demonstrated differences in CAMS in other work.⁸ In our multisite validation study, we observed similar differences in CAMS scores when those who were White were compared to those who were Black, Asian, another, or unknown race.⁸ We hypothesized that patterns may represent differences between the most represented racial group and underrepresented racial groups in headache medicine. To statistically capture these differences, we compared youth of color to youth who were White. CAMS scores were graphically represented as an overall range and as a probability density function estimate. Proportions of participants who met current versus alternate ICHD criterion D were reported. To compare rates of migraine versus probable migraine across race and site, Chi squared testing was used. Data fulfilled the following assumptions: data were organized as count variables into categories that were mutually exclusive, study groups were independent, two categorical variables were being compared, and the value of cell expected counts was 5 or more for at least 80% of the cells. A p-value of < 0.05 was defined as significant.

RESULTS

The 10,721 participants who were part of the multisite CAMS validation study were assessed for inclusion.⁸ There were 996/10,721 (9.3%) participants who were excluded because they had a diagnosis other than migraine, probable migraine, or TTH or did not meet an ICHD-3 headache diagnosis. An additional 4.6% (66/1,429) of CHOP respondents were excluded because their provider had assigned an ICD code of NDPH, PTH, or trigeminal autonomic cephalalgia. Of the remaining participants, 6.6% (642/9,659) were excluded because they did not have sufficient documentation of ICHD-3 migraine criteria A – D. Of the participants included in the final analysis, the median age was 13 years [IQR 10, 15] and 66.7% (6,038/9,017) were female; 85.2% (7,680/9,017) were from CCHMC and 14.8% (1,337/9,017) were from CHOP. Ninety-nine percent (8,924/9,017) had migraine or probable migraine, and 1.0% had TTH (93/9,017). Race and ethnicity matched the catchment areas of the respective centers, based on state 2020 census data.^12,13 Overall, 1.2% were Asian, 10.5% were Black, 5.7% were another or unknown race, and 82.4% were White; 3.1% reported Hispanic ethnicity.

Assessment of Selection Bias

We performed a demographic comparison of youth who were included in the analysis to youth who were excluded due to missing data to help assess for selection bias. We found in our multisite validation study of CAMS, that participants who completed the CHOP questionnaire were younger, less likely to be Black or another or unknown race compared to White, more likely to be Non-Hispanic than those who were assigned but did not start the questionnaire; the groups did not differ by sex.⁸ Within this study, those who were included because they had complete ICHD-3 criteria did not differ by sex, age, Asian or another or unknown race compared to White race, or ethnicity (p > 0.07) from those who were excluded due to missing data. However, those with complete data were more likely to identify as Black compared to White (estimated odds ratio (OR) = 1.39, [95%CI 1.04, 1.87], p = 0.028) and were more likely to be from the CHOP dataset (estimated OR = 4.12, [95%CI 2.77 – 6.13], p < 0.001). The overrepresentation of CHOP with complete data is likely due to the way diagnosis of migraine was made based on ICHD criteria rather than clinician diagnosis in the CCHMC dataset.

CAMS and Current ICHD-3 Criteria

We assessed migraine diagnostic criteria A – D for the 8,924 participants who were given a diagnosis of migraine or probable migraine; diagnosis was based on provider assessment at CCHMC and based on diagnostic algorithm at CHOP (Figure 1a).¹ Of those, 83.1% (7,419/8,924) met all criteria for migraine, 16.1% (1,436/8,924) met all but one of the criteria consistent with probable migraine (Figure 1b), and 0.8% (69/8,924, all from the CCHMC dataset) met fewer than 3 ICHD criteria for migraine. Most youth with probable migraine did not meet criterion D (10.1%, 903/8,924), followed by criterion B (3.8%, 336/8,924). Very few youth did not meet criteria for migraine due to pain quality characteristics or number of lifetime attacks.

Figure 1:

Percentage of participants with a migraine diagnosis stratified by meeting all diagnostic criteria and those with probable migraine diagnosis separate by criterion they did not meet. (a) ICHD-3 criteria for migraine without aura (top) and TTH (bottom). Migraine with aura has similar diagnostic criteria to migraine without aura, differing in the presence versus absence of aura. (b) Number of participants who met all Criteria A – D for migraine (black), and three of four Criteria A – D consistent with probable migraine. Those who did not meet Criterion A (darkest gray), Criterion B (dark gray), Criterion C (medium gray), or Criterion D (light gray) indicating the percentage of those with a diagnosis of probable migraine who did not meet that criterion. Of note, 69/8,924 (0.8%) of participants who were included under the umbrella diagnosis of migraine did not meet at least 3 of 4 documented ICHD criteria.

We assessed differences in ICHD-3 diagnosis and provider given diagnosis across the centers to determine the impact of differing approaches to data collection, which is summarized in Table 3. For CCHMC (clinician-guided), the rates of having a migraine compared to probable migraine diagnosis were higher than for CHOP (patient response questionnaires). This difference shrunk but remained significant when CHOP clinician-edited responses were used. Most patients were given a provider diagnosis of migraine whether they met ICHD-3 criteria for migraine or probable migraine.

Table 3:

ICHD-3 questionnaire-based diagnosis and provider-based diagnosis by site for participants under the umbrella diagnosis of migraine. CHOP provider diagnoses are based on ICD codes, CCHMC are based on clinician entered diagnoses. Probable migraine is not a specific ICD code diagnosis, therefore none of the CHOP participants were given a specific diagnosis of probable migraine. ICHD-3 criteria for CHOP clinician responses were determined by the clinician updated form. Dx = diagnosis. ICD = International Classification of Diseases. TTH = tension-type headache.

	ICHD-3 probable migraine	ICHD-3 migraine
CCHMC Total (%)	1,012/7,609 (13.3%)	6,597/7,609 (86.7%)
Provider Dx (Questionnaire entered)
Migraine	948/1,012 (93.7%)	6,425/6,597 (96.9%)
Probable migraine	64/1,012 (6.3%)	172/6,597 (2.6%)
CHOP patient responses Total (%)	424/1,246 (34.0%)	822/1,246 (66.0%)
Provider Dx (ICD codes)
Migraine	309/424 (72.9%)	754/822 (91.7%)
Headache syndrome	60/424 (14.2%)	42/822 (5.1%)
TTH	36/424 (8.5%)	15/822 (1.8%)
No headache diagnosis	19/424 (4.5%)	11/822 (1.3%)
CHOP clinician responses Total (%)	72/365 (19.7%)	293/365 (80.3%)
Provider Dx (ICD codes)
Migraine	44/72 (61.1%)	275/293 (93.9%)
Headache syndrome	13/72 (18.1%)	10/293 (3.4%)
TTH	13/72 (18.1%)	4/293 (1.4%)
No headache diagnosis	2/72 (2.8%)	4/293 (1.4%)
Statistical Comparisons
CCHMC vs. CHOP patient form ICHD-3	χ2 = 337.55, p < 0.001
CCHMC vs. CHOP clinician form ICHD-3	χ2 = 20.46, p < 0.001
CHOP patient vs. clinician form ICHD-3	χ2 = 18.19, p < 0.001
CHOP patient vs. clinician form provider Dx	χ2 = 5.29, p = 0.152

Migraine attack duration:

An attack duration of 2 hours or more was used to meet criterion B, which is recommended by the ICHD for those under 18 years old. Attack duration was collected differently between CHOP and CCHMC registries: CCHMC collected information on the shortest, average, and longest attack durations, while CHOP asked more broadly whether headaches lasted seconds, minutes, hours, 1 – 3 days, or more than 3 days. The longest attack duration from the CCHMC dataset was used to meet criterion B, while any attack duration of hours or more was used to meet criterion B in the CHOP dataset. A small percentage (3.8%) of those under the umbrella diagnosis of migraine were diagnosed with probable migraine because they did not meet requirements for attack duration (criterion B). Despite differences in questions, there were similar rates of having probable migraine due to attack duration at CCHMC (3.6%, 280/7,678) and CHOP (4.5%, 56/1,246). To explore attack duration further, we looked at shortest, average, and longest headache attack durations from CCHMC respondents who had these three durations captured and met all other criteria for migraine (7,315/7,678, 95.3%; Figure 2). Only 45.1% (3,300/7,315) of those with migraine or probable migraine would have met criterion B if the shortest attack duration was used. This percentage increased to 82.8% (6,060/7,315) when the average attack duration was used and 96.1% (7,031/7,315) when the longest attack duration was used. These data indicate that attack duration can be variable on the individual level.

Figure 2:

Duration of shortest, average, and longest headache attacks from CCHMC. The dotted line shows the 2-hour cut-off for those under 18 years currently recommended by the ICHD-3.

Migraine-associated symptoms:

Of those under the umbrella diagnosis of migraine, 10.1% (903/8,924) were diagnosed with probable migraine because they did not meet requirements for migraine-associated symptoms (criterion D). Of note, almost all these participants (803/903, 88.9%) were diagnosed with migraine based on provider diagnosis. We first assessed if those who did not meet criterion D reported many migraine-associated symptoms not included in the ICHD-3. We found that 20.4% (184/903) reported none of the 11 migraine-associated symptoms, 28.8% (260/903) reported one symptom, and 50.8% (459/903) reported two or more symptoms. Therefore, half of those who did not meet a migraine diagnosis due to not having enough migraine-associated symptoms reported two or more symptoms when an expanded list was considered. To better understand the composition of these symptoms, we graphically represented the distribution (both range of values and probability density) of CAMS symptom number and CAMS migraine ICHD symptoms for those with migraine, probable migraine due to not meeting criterion D, and TTH (Figure 3). Those with ICHD migraine-associated symptoms lie on the upper part of the graph, while those with probable migraine who did not meet criterion D lie on the lower part of the graph. Those with TTH occupied the space in the left section of the graph, indicating most had few or no associated symptoms; however, there were some individuals with a diagnosis of TTH who reported non-ICHD migraine-associated symptoms. Specifically, 46.2% (43/93) of respondents diagnosed with TTH reported no symptoms, 32.3% (30/93) reported one symptom, and 21.5% (20/93) reported two or more symptoms. Overall, the graphical representation of migraine, probable migraine, and TTH diagnoses showed that migraine captures those who report the most migraine-associated symptoms, with probable migraine being second, and TTH showing the fewest number of symptoms. There is overlap between migraine and probable migraine, and (to a lesser extent) TTH in the number of migraine-associated symptoms when a broader number of symptoms is considered. Given that there were substantially fewer respondents diagnosed with TTH compared to migraine and probable migraine, we re-ran CAMS on a smaller sample where participants with TTH were matched with those with probable migraine and migraine. Importantly, this yielded very similar dimensions to the main CAMS model. This indicates that CAMS is valid for those with TTH even when they make up a small percentage of the overall sample, though differences were seen and described in more detail (see supplemental materials).

Figure 3:

(A) Range of CAMS 1 scores (x-axis) and CAMS 2 scores (y-axis) for participants with TTH (white), probable migraine who did not meet migraine criteria D (gray), and migraine (black). (B) Probability density function estimate of CAMS 1 and CAMS 2 scores; the volume of each bar is the relative number of observations, defined by $v_i=\frac{c_i}{N{A}_i}$.

Updating ICHD Migraine Associated-Symptom Criterion

We next explored how changing the ICHD-3 migraine-associated criterion would alter who was given a migraine diagnosis using CAMS as a guide (Figure 4). We left nausea and vomiting as sole symptoms that would meet criteria for migraine given that they can be the sole presenting symptom, particularly in young children, which is supported by CAMS in our multisite validation study.⁸ We assessed the impact of changing the existing ICHD-3 criteria in three different ways.

Figure 4:

CAMS symptom number and CAMS ICHD migraine symptom scores for individual participants with probable migraine who did not meet criteria due to insufficient migraine-associated symptoms (gray) under current ICHD-3 diagnostic criteria, and the changes in diagnostic category for participants based on the new criterion D (black). Points are displayed at an opacity of 70% to visualize areas where many points overlap.

In the first iteration, we allowed for photophobia or phonophobia to meet criteria for migraine, which cast the widest net. The rationale for this change was that photophobia and phonophobia overlap substantially in CAMS, and both are indicative of sensory hypersensitivity. Additionally, we found differences in the rates of photophobia and phonophobia between CCHMC (clinician-guided questionnaire) and CHOP (patient response questionnaire): of those who reported either photophobia or phonophobia, 79.3% reported both in the CCHMC dataset compared to 66.1% in the CHOP dataset (χ² = 85.77, p <0.001). This may indicate that upon additional clinician-guided questioning, both sensory hypersensitivities were found to be present. This hypothesis is further supported by the finding that for CHOP questionnaires that were updated by a clinician (29.8%, 398/1,337), the rate of both photophobia and phonophobia in those who reported either symptom increased to 74.4%. This change would capture individuals with more pronounced sensory hypersensitivity within one modality. Allowing for photophobia or phonophobia moved 55.6% (502/903) of those with probable migraine due to not meeting migraine-associated symptom criterion to meeting a diagnosis of migraine; 29.2% (147/502) of these respondents reported only one migraine-associated symptom with the remaining 70.6% (356/502) reporting two or more. However, 11.5% of respondents (104/903) with two or more non-ICHD migraine-associated symptoms were still excluded from a diagnosis of migraine with this change.

In the second iteration, we included lightheadedness, which was the most common symptom reported of the non-ICHD migraine symptoms and required two of three symptoms to fulfill criterion D. Of those with probable migraine, 19.7% (178/903) – all of whom by definition had at least two migraine-associated symptoms – met diagnostic criteria for migraine, capturing more individuals with non-ICHD migraine-associated symptoms. A sizeable number (31.1%, 281/903) who reported two or more migraine-associated symptoms were still excluded from a diagnosis of migraine.

In the third iteration, we included both lightheadedness and spinning to capture both non-specific global neurologic and vestibular symptoms and required two of four symptoms to fulfill criterion D. Of those with probable migraine due to not fulfilling criterion D, 25.8% (233/903) met criteria for migraine – all of whom had two or more migraine-associated symptoms. Still, many (25.0%, 226/903) with two or more symptoms were not captured with this change. The second and third iterations cast a narrower net but selectively included youth with a greater number of non-ICHD migraine-associated symptoms.

Race and ICHD-3 diagnosis

In our studies that describe CAMS development,^7,8 we found a weak but consistent pattern that youth who identified as either Black, Asian, or another or unknown race reported fewer migraine-associated symptoms than youth who identified as White and possibly more migraine-associated symptoms not included in the ICHD-3. We aimed to explore this observation further. Given that similar CAMS patterns were seen for underrepresented groups in headache medicine, including those who identified as Black, Asian, or another race, as well as those with unknown race compared to youth who identified as White, a statistical comparison was made between youth who identified as White and People of Color.

Under the current ICHD-3, 16.0% (1,184/7,419) of participants with a diagnosis of migraine identified as People of Color, 83.9% (6,222/7,419) identified as White, and 0.2% (13/7,419) had missing information about race. Of the 903 participants who met a diagnosis of probable migraine because they did not meet one migraine-associated symptom criterion, 23.8% (215/903) were People of Color, 75.9% (685/903) were White, and 0.3% (3/903) had missing information about race. There was a significant difference between probable migraine and migraine across race (χ² = 35.77, p < 0.001). The three alternative iterations of criterion D included a greater number of individuals under the diagnosis of migraine but did not alter this difference (χ² > 27.67, p < 0.001).

Furthermore, we observed that differences in ICHD-3 across race were not consistent across sites. There were no differences between youth who identified as People of Color and youth who identified as White in the CCHMC dataset for the current ICHD criterion D or proposed alternatives (χ² ≤ 3.03, p > 0.088). By comparison, these differences were present in the CHOP dataset under current ICHD criterion D and proposed alternatives (χ² > 9.85, p < 0.002). To help determine if this effect could be explained by differences in a clinician-guided questionnaire versus an unguided patient-reported questionnaire, we looked at clinician updated forms from the CHOP dataset. We found that the differences in ICHD-3 migraine versus probable migraine diagnosis between respondents who were People of Color and respondents who were White were not present in the clinician updated forms (χ² = 0.33, p = 0.563), supporting this hypothesis.

DISCUSSION

Diagnostic criteria are used to define discrete pathologic conditions, inform treatment, and guide clinical trial design. Our findings largely support current ICHD-3 migraine diagnostic criteria. We found that criterion A (number of migraine attacks) and C (head pain quality and characteristics) rarely were the sole features that excluded youth from meeting a full diagnosis of migraine. This indicates these features are inclusive to migraine diagnosis. Criterion B (attack duration) was a more common criterion to exclude youth from a full diagnosis of migraine but was still relatively uncommon. Expanding the headache duration to attacks as short as 1 hour has been proposed for the pediatric population,^18,19 but has not been supported by the ICHD-3.^1,3 Our findings indicate that a cut off of 2 hours for those under 18 years old excludes a small subset of individuals; however, our findings also emphasize the importance of using the longest attack durations as a guide for this criterion.

The most common reason youth were excluded from a full diagnosis of migraine was due to insufficient migraine-associated symptoms (criterion D). CAMS provides composite scores with which to assess criterion D. CAMS supports the selection of photophobia, phonophobia, nausea, and vomiting as ICHD migraine-associated symptoms because they are prevalent and tend to co-occur. However, CAMS also demonstrated that half of youth with a diagnosis of probable migraine due to not meeting criterion D experience multiple non-ICHD migraine-associated symptoms.

Casting a wider net for ICHD migraine-associated symptoms

The distinction between probable migraine and migraine is critical because migraine is a much more studied entity. The advantage of broadening criteria for migraine diagnosis is that it provides access to effective treatments and clinical trials to a larger group of individuals.

Our findings support allowing photophobia or phonophobia to fulfill the migraine-associated criterion. Because photophobia and photophobia are very common, this change identified the most youth with multiple non-ICHD migraine-associated symptoms. Furthermore, rates of both photophobia and phonophobia were higher in the CCHMC clinician-guided questionnaires and the CHOP clinician updated questionnaires compared to the unreviewed CHOP patient responses. This observation indicates that those with sensory hypersensitivity may have one modality that is more impacted than another, leading them to report only photophobia or phonophobia, but both can be identified with targeted questioning. This targeted questioning would not be needed if photophobia or phonophobia met the migraine-associated symptom criterion, and thus would improve ease of use. Furthermore, it would necessitate the removal of photophobia or phonophobia being permitted to meet a diagnosis of TTH, which is currently suggested in the appendix of the ICHD-3.¹ Such a change would create a clearer distinction between TTH and migraine and prevent individuals from being misdiagnosed with TTH when migraine-associated symptoms are present.

ICHD diagnosis and equitable care

Critical evaluation of ICHD diagnostic criteria is important for equitable care. Prior work has identified differences in ICHD diagnosis of migraine across race – a social and cultural construct. This can have implications for treatments offered.^20–22 Differences in migraine-associated symptoms were found between youth who were White and People of Color in unreviewed patient responses to the CHOP dataset but not the clinician-guided CCHMC questionnaires or the CHOP clinician-corrected responses. This finding indicates that racial disparities may be addressed in part by clinician-guided questionnaires. It is important to note that youth who identified as People of Color were compared to youth who were White in the main analysis because all underrepresented groups showed the same pattern of reporting fewer migraine-associated symptoms and more non-ICHD symptoms that impacted their chances of being diagnosed with migraine. However, it is critical to recognize that distinct racial identities with unique social and cultural experiences are represented within this broader group. A limitation of our analysis is that we did not include additional social determinants of health that intersect with other demographic factors (e.g., socioeconomic indicators),¹⁷ which should be considered when interpreting the results. Further study is needed to better understand the intersection between social, structural, and/or cultural influences that may contribute to differing reports of migraine-associated symptoms.

The importance of non-ICHD-3 migraine-associated symptoms

Though inclusion of lightheadedness and spinning in migraine diagnostic criteria did not lead to substantial shifts in those who received a diagnosis of migraine, these symptoms are still important. An expanded assessment of migraine-associated symptoms through CAMS differentiates milder from more severe headache burden. The presence of non-ICHD symptoms correlates with greater headache frequency, and a greater number of migraine-associated symptoms overall correlates with greater headache-related disability and severity based on our studies developing and validating CAMS.^7,8 We also found individuals with TTH and probable migraine diagnoses who report multiple non-ICHD migraine-associated symptoms. Indeed, we found blurred lines exist between migraine, probable migraine, and TTH in CAMS based on current diagnostic criteria. It may be that TTH, probable migraine, and migraine represent a continuum of headache severity,²³ consistent with prior work that supports this “continuum hypothesis” for adolescents, young adults, and those with frequent headache.²⁴ Alternatively, it may be that we have not fully identified the criteria that cleanly separate TTH from migraine, despite being separate entities with distinct pathophysiology, as suggested by the “spectrum hypothesis.”²⁵

Our findings also support the association between different forms of dizziness (lightheadedness and vertigo) and migraine. Overall, rates of dizziness (lightheadedness and/or spinning) were high in both samples, with at least one dizziness symptom reported in 61.7% of CCHMC and 47.8% of CHOP participants with migraine. This rate is 8 to 11 times higher than 5.3 – 5.6% rates of dizziness in children from large population-based studies.^26,27 The relationship between dizziness and headache has also been observed in vestibular clinics specializing in childhood dizziness, where those with central vestibular conditions were 16 times more likely to report headache.²⁸ Over a quarter of patients presenting to vestibular and dizziness clinics are diagnosed with vestibular migraine.^29,30 The current ICHD-3 deals with this association with the diagnosis of vestibular migraine, which is currently in the appendix of the ICHD-3 and, thus, not a confirmed subtype of migraine.^1,4–6 CAMS did show that spinning, vision changes, and ear ringing, along with lightheadedness and difficulty thinking, clustered together across sites⁸ and was consistent with our initial single-center study.⁷ This finding may be consistent with a distinct vestibular migraine subtype, or it may be that dizziness and vestibular symptoms are common manifestations of migraine. Further study is needed to determine if the cluster of vestibular symptoms represents a unique migraine endophenotype. Notably, several participants with lightheadedness and/or vestibular symptoms did not meet full ICHD-3 criteria for migraine, consistent with studies of vestibular migraine in adults.^6,31

Limitations

Our dataset has sources of bias that should be considered when interpreting the results. The study focused on patients who presented to specialized clinics for headache care, which can be impacted by access to care and is often biased toward more severe presentation. This bias likely contributed to few participants (1.0%) in the sample being diagnosed with TTH (nearly all of those were from CHOP), limiting the comparison of TTH and migraine in this study. Furthermore, there was evidence of selection bias, including demographic differences between those who were assigned but did not start from those who completed the CHOP questionnaire, which we reported in prior work.⁹ CHOP clinician-edited forms also differed systematically from those that were not edited: clinicians who used the questionnaire were more likely to be in the headache program or practicing at one of the satellite locations.⁹ There are variables that we did not collect (e.g., primary language) that may have resulted in additional selection bias. Patient responses are also subject to recall bias, bias introduced by the phrasing of the questions, and discomfort with disclosing certain symptoms. Furthermore, some of the survey data did not allow for analysis of ICHD-3 criteria with pinpoint precision. For instance, the number of migraine attacks (CCHMC) and attack duration (CHOP) had to be extrapolated. We were able to explore the impact of patient bias (e.g., the patient interpreted the question differently than was intended) versus clinician bias (e.g., knowledge of current diagnostic criteria for migraine that may have guided directed questioning) because we used both unedited and clinician-edited patient responses across the two datasets. These two forms of bias do indeed appear to yield different results with ICHD diagnosis. Finally, our study focused on the pediatric population, and it is unclear if findings would be similar in adults. Given these limitations, additional work is needed to determine the generalizability of our results.

A major challenge in critically appraising current diagnostic criteria is what to use as the gold standard for diagnosis. Most participants at CHOP and CCHMC who met ICHD-3 criteria for probable migraine were given a migraine diagnosis, which makes the use of provider diagnosis as a gold standard challenging. We interpret this to mean that if all but one criterion for migraine were met, patients would be treated the same way as if they had a full diagnosis of migraine, and therefore, they were given a diagnosis of migraine. The CHOP dataset was also limited in that ICD codes were used that do not include probable migraine as a separate diagnosis, but the same pattern was found when looking at clinician-entered diagnoses within the CCHMC dataset. To help address this broader challenge, one of our future directions is to conduct a longitudinal study to determine if outcomes differ between those with ICHD-3 criteria for probable migraine or migraine and assess the relationship between prognosis and CAMS.

Conclusions

In summary, CAMS provides a broader perspective on migraine-associated symptoms that both supports the ICHD symptoms of migraine for diagnosis and highlights the significance of non-ICHD migraine-associated symptoms as a part of the headache burden. Based on current evidence, we propose casting a wider net for full migraine diagnosis by including photophobia or phonophobia to meet the associated symptom criterion for migraine.

Abbreviations:

CAMS

Cluster Analysis of Migraine-associated Symptoms

CCHMC

Cincinnati Children’s Hospital Medical Center

CHOP

Children’s Hospital of Philadelphia

ICHD

International Classification of Headache Disorders

MCA

Multiple Correspondence Analysis

TTH

Tension-type headache