Migraine-associated symptoms change with improvement in youth with migraine: A cohort study

Abstract

Aim

To determine how migraine-associated symptoms change over time, and if they are predictive of outcome at first follow-up visit. Symptoms associated with headache—photophobia and phonophobia, nausea, and/or vomiting—are required criteria for the diagnosis of migraine. Individuals with migraine also report high rates of other symptoms (e.g. lightheadedness, difficulty thinking). We have developed cluster analysis of migraine-associated symptoms (CAMS), which is a composite representation of an expanded set of migraine-associated symptoms that is associated with headache burden. It is unknown if CAMS can change with treatment response or predict outcomes.

Methods

We conducted a secondary analysis of prospectively collected clinician-guided patient questionnaire data from youth (6–17 years) with migraine who were seen at a specialized pediatric headache clinic. Data were collected at time of presentation and at the first follow-up within 30 to 90 days from the first visit. CAMS were implemented to assess associations between 11 migraine-associated symptoms. We determined if CAMS changed over time and as a function of treatment outcome, and if CAMS could be used to predict outcome at first follow-up.

Results

Youth with migraine (n = 8008, 67.1% female, median 13 years old) were included in this study. CAMS revealed that the number of reported migraine-associated symptoms decreased at follow-up, particularly for those who improved. Resolution of nausea and vomiting were most common. CAMS at the initial visit differed between youth who improved and youth who did not, but the effect was clinically negligible and could not reliably predict outcome (AUC 0.53).

Conclusion

Migraine-associated symptoms as captured by CAMS changed over time based on whether youth noted improvement or not but was not predictive of who improved at follow-up.

Keywords

migraine pediatric vestibular dizziness sensory hypersensitivity

Introduction

The presence or absence of symptoms associated with headache play an important role in diagnosis of primary headache disorders according to the International Classification of Headache Disorders (ICHD-3).¹ The main symptoms that differentiate migraine from tension-type headache include photophobia and phonophobia, nausea, and/or vomiting. However, many other symptoms have been associated with migraine including osmophobia,² dizziness (both lightheadedness and vertigo),^3–7 and cognitive symptoms,^3,8 among others. For individuals with migraine, it is of at least equal importance to find effective management strategies for associated symptoms as for head pain.^9,10

Determining how migraine-associated symptoms relate may identify endophenotypes of migraine to support more personalized care. We have developed cluster analysis of migraine-associated symptoms (CAMS) to capture the relationships between the presence or absence migraine-associated symptoms. This approach is based on multiple correspondence analysis (MCA), which explains variance across multiple migraine-associated symptoms using a fewer number of dimensions.^11,12 We have found that approximately 23% of the variance in migraine-associated symptoms relates to the presence or absence of symptoms: if an individual reports one symptom associated with headache, they are likely to report several. The second CAMS dimension that explains about 13% of the variance shows that symptoms cluster into ICHD migraine symptoms (nausea, vomiting, photophobia, phonophobia) and a non-ICHD migraine cluster including symptoms of non-specific global neurologic dysfunction (difficulty thinking, visual disturbance, lightheadedness), and vestibular symptoms (spinning, double vision, ear ringing; Figure 1). We found that CAMS correlated with other migraine burden metrics: the greater number of symptoms, and the more non-ICHD symptoms an individual reports are both associated with greater headache frequency and greater headache-related disability.

Figure 1.

CAMS model. These data were originally presented in Patterson Gentile and colleagues (2024) and include cross-sectional data from the initial visit of youth with headache of any diagnosis from Children's Hospital of Philadelphia and CCHMC combined headache registries. Patients in this study are a sub-group of those used in the development of CAMS. The location of the symptom in this three-dimensional space represents its relationship to other symptoms. The size of the bubble indicates the percentage of youth included in the model who reported that symptom. Symptoms that are close together are more likely to occur. The first dimension (x-axis) that explains the most amount of variance captures no symptoms (left), few symptoms (middle), and many symptoms (right). The second dimension captures the second most variance (y-axis) separating symptoms into those included in the ICHD + osmophobia and those not included in the ICHD.

To date, all studies of CAMS have been cross-sectional. Therefore, it is unknown if CAMS are fixed or can change over time and with overall improvement of migraine symptoms. Furthermore, it is unknown if CAMS can be used as a prognostic biomarker. To address these open questions, we pursued a secondary analysis of prospectively collected longitudinal from the Cincinnati Children's Hospital Medical Center (CCHMC) headache program. We hypothesized that CAMS (i.e. the number and type of migraine-associated symptoms) would change with improvement and could be used to predict the likelihood of improvement.

Methods

This was a secondary analysis of prospectively collected longitudinal standardized patient questionnaire cohort data that was filled out under the guidance of a headache specialist from a large headache registry.

Participants

Youth aged 6 to 17 years old with a diagnosis of migraine seen for intake and at least one follow-up visit within 30 to 90 days of the first visit at the CCHMC pediatric headache specialty clinic from the start of collection in 2010 until June 2023 were included. Diagnosis was determined by a headache specialist, based on ICHD-3 migraine criteria.¹ Both migraine and probable migraine diagnoses, and migraine with aura and migraine without aura were included. The CCHMC Institutional Review Board approved the extraction of the data from the electronic health record into a research registry, with a waiver of verbal and written informed consent and assent to maximize generalizability. Participants received multi-disciplinary headache care at the initial visit that included education on headache, and an integrative treatment plan based on provider assessment. These management plans may have included preventive pharmacologic therapy (most commonly amitriptyline), pharmacologic management of acute headache worsening, neuromodulatory devices for acute and/or preventive treatment, and non-pharmacologic management strategies like counseling on healthy headache habits and, in some cases, cognitive behavioral therapy.

Data collection

CCHMC data were based on a clinician-guided standardized questionnaire. Answers at presentation were used for our prior studies on multisite validation of CAMS and critical appraisal of the ICHD-3.^12,13 Demographic information including age, sex, race, and ethnicity were collected. The presence or absence of 11 migraine-associated symptoms needed to calculate CAMS was collected at presentation and follow-up. These symptoms included photophobia, phonophobia, osmophobia, nausea, vomiting, lightheadedness, difficulty thinking, vision changes, spinning, neck pain, and ear ringing. Additional migraine burden metrics at initial visit to define the patient population included the following: (a) average headache severity (0–10 scale), (b) headache-related disability as determined by PedMIDAS score, which had a possible range of 0 to 240,¹⁴ and (c) number of headache days per month. Frequency of bad headache was collected, but there was a high rate of missingness as it was a later question addition, so this question was not included in the analysis. The presence of aura was also included as this is a defining feature of the ICHD that distinguishes different subgroups of migraine. Specific treatments at the individual level were not assessed. The primary outcome was whether patients perceived that overall, they were better, there was no change, or worse at the follow-up compared to the initial visit. The “no change” and “worse” categories were combined due to the small numbers of those who reported worsening. We referred to those who perceived they were better as “improvement’ and those who perceived they were the same or worse as “no improvement.”

Data analysis

Matlab (Mathworks, Natick, MA) was used to produce custom written code for all analyses, which we have made publicly available.¹⁵

CAMS

We applied the multisite validated CAMS model to this longitudinal dataset to determine relationships between migraine-associate symptoms (Figure 1).^11–13 CAMS is based on MCA using an indicator matrix adapted from a publicly available Matlab function¹⁶ that reduces the dimensionality of binary data (i.e. the presence or absence of 11 migraine-associated symptoms) in Euclidean space.¹⁷ Both the symptoms themselves and the symptoms of individual participants can be represented in this space. The closer one symptom is to another, the more likely they are to co-occur. The original CAMS model was developed including youth with headache regardless of headache diagnosis to explore the relationship between associated symptoms and headache in youth in a broader context, removing the bounds placed by current diagnostic criteria. The first dimension represents overall symptom number, with negative values corresponding to no or few symptoms, and positive values corresponding to many symptoms. The second dimension clusters symptoms into those included in the ICHD (positive values) and those not included in the ICHD (negative values). While the multicenter validated CAMS model originally had three dimensions, we focused on the first two dimensions for analysis because these two dimensions correlated with other migraine burden metrics. Data from the initial visit of participants in this study were used in the development of the multisite CAMS model.

Statistical analysis

No a priori sample size calculations were done. Sample size was based on available data, which was sufficient given our experience with prior analysis.^11–13 Categorical variables were reported as percentages. Age was reported as number of birthdays and was reported as median values with interquartile ranges given non-normally distributed data. CAMS’ scores were represented as mean with standard error of the mean, though full distribution curves were also presented. For all statistical modeling of CAMS, significance was defined by Bonferroni correction for two CAMS variables as p < 0.025. Hypothesis testing was two-tailed. Mixed effects models were used to evaluate the change in CAMS based on outcome (improved vs. not improved) with visit (initial and follow-up) treated as a repeated measure. Models for CAMS used a normal distribution. For CAMS dimension 2, analyses were run with all respondents, and then a sensitivity analysis was performed that only included those who reported at least one associated symptom. This was done because those with no symptoms did not have meaningful values in these two dimensions that captured symptom clusters, as described in our previous studies.^11–13

To determine the predictive value of CAMS, binomial logistic regression models with improvement versus no improvement as the primary outcome and CAMS dimensions 1 and 2 at the initial visit as the primary predictors were used to generate receiver operating characteristic (ROC) curves and area under the curve (AUC) was calculated, with 0.5 defined as no discrimination, 0.5 to 0.7 defined as poor discrimination, 0.7 to 0.8 as good discrimination, and >0.8 as excellent discrimination. Rates of missing data were overall low and are reported.

Results

A total of 8675 patient responses were assessed for inclusion; 667 were excluded because they had a primary diagnosis other than migraine, most commonly probable migraine (n = 269), post-traumatic headache (n = 174), new daily persistent headache (n = 130); all other diagnoses occurred with low frequency (n < 50). This left 8008 patients included in the final analysis. Patients were a median age of 13 years old [IQR 10–15] and were 67.1% female (for demographics see Table 1). Of all participants, 19.1% had migraine with aura.

Table 1.

Patient demographics.

Characteristic	N = 8008
Median Age yrs [IQR]	13 [10–15]
Sex (%)
Female Male Unknown	5375 (67.1); 2631 (32.9); 2 (<0.1)
Race ± Ethnicity (%) Total American Indian Asian Black Middle Eastern Mixed race Other Pacific Islander White Unknown	Hispanic145 (1.8)3 (<0.1)0 (0.0)0 (0.0)6 (<0.1)2 (<0.1)22 (0.3)0 (0.0)64 (0.8)48 (0.6)	Non-Hispanic7863 (98.2)1 (<0.1)70 (0.9)784 (9.8)6 (<0.1)28 (0.3)203 (2.5)2 (<0.1)6751 (84.3)18 (0.2)
HA characteristics
Migraine w/ aura (%) Migraine w/out aura (%) HA frequency Median [IQR] HA severity Median [IQR] PedMIDAS Median [IQR]	1581 (19.7)6427 (80.3)16 [9, 30]6 [5, 7]28 [1, 61]

HA = headache; w/ = with; w/out = without. Data missingness was <3%.

At follow-up, most patients (64.2%) reported that they perceived overall improvement since their initial visit. Linear mixed effects models with repeated measures were generated to determine how CAMS changed over time in youth who perceived improvement at follow-up compared to those who did not (Figure 2; Table 2). Overall, the number of reported migraine-associated symptoms decreased between initial and follow-up visits (CAMS 1). This reduction was greater for those who perceived improvement (mean estimate initial visit 0.24 [95%CI 0.18–0.31] vs. follow-up −1.50 [95%CI −1.57 to −1.42]) than those who did not (0.44 [95%CI 0.35–0.53] vs. −0.48 [95%CI −0.57 to −0.38]). Symptoms reported at follow-up were more heavily weighted towards non-ICHD symptoms than the initial visit (CAMS 2). This shift toward non-ICHD symptoms was more pronounced for those who perceived improvement (0.15 [95%CI 0.11–0.19] vs. −0.55 [95%CI −0.59 to −0.52]) compared to those who did not (0.05 [95%CI −0.00 to 0.10] vs. −0.41 [95%CI −0.46 to −0.37]).

Figure 2.

Change in CAMS as a function of time and outcome. Ridgeline plots representing the probability density function of CAMS 1 (left panel), CAMS 2 (right panel) from the initial visit (solid) and the follow up visit (dotted) for youth who reported improvement (black) compared to youth who reported no improvement (gray) in migraine symptoms between the initial visit and follow up. Triangles show the mean CAMS values for initial (solid) and follow up (open) visits. Sx = Symptoms.

Table 2.

Change in CAMS between initial visit and follow-up.

	No Improvement		Improvement		p-Value
	Initial	Follow-up 30–90 days	Initial	Follow-up 30–90 days	Outcome	Outcome:Time
CAMS 1	0.44[0.35–0.53]	−0.48[-0.57 to -0.38]	0.24[0.18–0.31]	-1.50[-1.57 to -1.42]	p < 0.001	p < 0.001
<12 yrs	-0.35[-0.50 to -0.21]	-1.41[-1.57 to -1.26]	-0.47[-0.56 to -0.37]	-2.18[-2.28 to -2.07]	p < 0.001	p < 0.001
12 yrs+	0.83[0.72–0.93]	-0.02[-0.14 to 0.10]	0.68[0.60–0.76]	-1.08[-1.17 to -0.98]	p < 0.001	p < 0.001
CAMS 2	0.05[-0.00 to 0.10]	-0.41[-0.46 to -0.37]	0.15[0.11–0.19]	-0.55[-0.59 to -0.52]	p = 0.458	p < 0.001
CAMS 2 Sensitivity analysis	0.04[-0.02, 0.09]	-0.35[-0.40, -0.30]	0.12[0.07, 0.16]	-0.44[-0.48, -0.40]	p = 0.866	p < 0.001
< 12 yrs	0.59[0.50–0.68]	-0.03[-0.11 to 0.06]	0.53[0.47–0.59]	-0.35[-0.41 to -0.29]	p < 0.001	p < 0.001
12 yrs+	-0.21[-0.27 to -0.15]	-0.60[-0.65 to 0.55]	-0.08[-0.13 to -0.04]	-0.69[-0.72 to -0.64]	p = 0.392	p < 0.001

	No Improvement		Improvement		p-Value
	Initial	Follow-up 90–120 days	Initial	Follow-up 90–120 days	Outcome	Outcome:Time
CAMS 1	0.36[0.23–0.49]	-0.65[-0.80 to -0.51]	0.16[0.07–0.26]	-1.57[-1.67 to -1.46]	p < 0.001	p < 0.001
CAMS 2	0.09[0.02–0.17]	-0.39[-0.46 to -0.33]	0.15[0.10–0.21]	-0.51[-0.56 to -0.46]	p = 0.473	p < 0.001

Linear mixed effects models treating timepoint as a repeated measure were calculated for each CAMS dimension. Marginal means with 95% CI are reported. Values are on a linear scale. The absolute value or sign of a CAMS score does not hold specific meaning but rather captures the amount of variance accounted for across a dimension. Because of this, interpreting changes in CAMS can be challenging. To help provide an intuition, we will provide value landmarks. For CAMS 1, larger positive values indicate greater symptom number, while small negative numbers indicate only a couple of symptoms are present, and a value of −5.33 indicates no migraine-associated symptoms. For CAMS 2, No symptoms correspond to a value of −1.22. Values above −1.22 indicate symptoms associated with the ICHD with increasing positive values indicating more ICHD symptoms, while values lower than −1.22 indicate more non-ICHD symptoms including vestibular symptoms and brain fog, with the greatest number of non-ICHD symptoms being associated with the lowest negative values. We performed a sensitivity analysis only on patients who reported at least one migraine-associated symptom at the initial visit and follow-up (n = 7010). This is because “no symptoms” are located centrally in CAMS 2 values that could have limited identifying changes in symptom clusters. Missing data rates for all metrics were <3% at both time points. HA = headache; mo. = month.

It is important to consider where no migraine-associated symptoms lie within each CAMS dimension when interpreting these results. For CAMS 1, this is straightforward with no migraine-associated symptoms falling to the far left, corresponding to the most negative CAMS’ value possible. Thus, no additional analysis is needed. However, for CAMS 2, the value corresponding to no migraine-associated symptoms was located centrally near a value of 0. This means that youth with more classic migraine-associated symptoms are in the positive CAMS 2 space, while vestibular symptoms are in the more negative CAMS 2 space, with those with no migraine-associated symptoms sitting in between these two clusters. As a result, the change in CAMS 2 could be explained by a shift toward no migraine-associated symptoms, as opposed to a change in the composition of the symptom clusters, toward more non-ICHD symptoms. Therefore, sensitivity analysis was performed on youth with at least one migraine-associated symptom at both visits (n = 7010). This did not significantly change the outcome of the results, supporting that this was a true shift toward non-ICHD symptoms. This shift appears to be driven primarily by a reduction in the number of patients who reported nausea and vomiting between initial and follow-up visits (Table 3). In those who reported improvement, nausea was present in 70.2% at initial visit compared to 37.6% at follow-up. In those who reported no improvement, nausea was present in 69.8% at initial visit, compared to 50.2% at follow-up. Similarly, those who reported improvement, vomiting was present in 36.5% at initial visit compared to 13.8% at follow-up. In those who reported no improvement, vomiting was present in 34.2% at initial visit, compared to 19.6% at follow-up.

Table 3.

Percentage of youth reporting migraine-associated symptoms at initial presentation and follow-up.

	No improvement		Improvement
	Initial	Follow-up	Initial	Follow-up
Photophobia	86.2	75.8	85.2	66.5
Phonophobia	80.1	67.3	80.4	56.9
Osmophobia	23.1	24.0	20.8	17.3
Nausea	69.8	50.2	70.2	37.6
Vomiting	34.2	19.6	36.5	13.8
Neck pain	44.4	37.9	38.6	30.5
Lightheadedness	59.2	50.5	54.6	37.9
Vision changes	31.7	27.0	30.2	19.1
Difficulty thinking	47.1	39.0	47.0	30.5
Spinning	28.7	25.3	27.0	16.9
Ear ringing	19.4	16.7	18.1	11.7

Predicting improvement based on migraine-associated symptoms at presentation

We then determined if CAMS could predict improvement at follow-up. To accomplish this, binomial logistic regression models were produced with improvement versus no improvement as the primary outcome, and CAMS as the primary predictor. The two CAMS dimensions were included as predictors in the same model to represent changes in reported migraine-associated symptoms. Those who reported improvement endorsed fewer migraine-associated symptoms at the initial visit (CAMS 1 OR of improvement with increasing value 0.97 [95%CI 0.95–0.99], p = 0.001) and primarily reported symptoms included in the ICHD-3 (CAMS 2 OR of improvement 1.05 [95%CI 1.02–1.09], p = 0.004). While statistically significant, the magnitude of differences in CAMS 1 and CAMS 2 was clinically negligible and had poor discrimination in predicting improvement (AUC 0.53).

Comparison between older and younger youth

CAMS change across age with older youth reporting more migraine-associated symptoms overall, and more non-ICHD symptoms. We performed sub-analyses only on youth 11 years old and younger (n = 2905), and only on youth 12 years old and older (n = 5103) to determine if there were age-related differences in how CAMS changed over time. This did not change the outcome of the results, except that there was a significant difference in CAMS 2 based on outcome alone for younger children (Table 2). CAMS had poor discrimination in predicting improvement for both age groups (AUC = 0.52 for younger; AUC = 0.53 for older).

Analysis of 90- to 120-day follow-up

Our primary analysis focused on patients who returned for follow-up between 30 and 90 days. To address this limited follow-up window, we performed an analysis on 3787 patients who had follow-up at 90 to 120 days to determine if changes in CAMS over time was similar at longer follow-up periods. Of those, 1587/3787 (41.9%) also had a follow-up visit between 30 and 90 days, and therefore were included in the original analyses. Despite this being a later follow-up period, and an overlapping but different cohort, outcomes were the same: the number of reported migraine-associated symptoms were overall lower at follow-up and trended towards more non-ICHD symptoms, particularly for those who reported improvement (Table 2). CAMS at the initial visit also had poor predictive power for who would improve at follow-up for this later time period (AUC = 0.52).

Discussion

Longitudinal assessment of CAMS revealed that the experience of migraine-associated symptoms, like other migraine burden metrics, can change over time and improve with effective management. However, CAMS were not able to predict perceived improvement between the initial and follow-up visits.

We found that nearly two-thirds of youth reported improvement after being seen in a specialty headache clinic, consistent with improvement in treatment and placebo groups from the CHAMP trial.¹⁸ A reduction in the number of migraine-associated symptoms reported was associated with overall improvement, with nausea and vomiting being the most likely to resolve. Notably, 11.5% of participants reported no migraine-associated symptoms at follow-up compared to 1.2% at the initial visit. We interpret this to mean that migraine-associated symptoms were substantially improved and no longer bothersome to the patient who experienced improvement, rather than migraine-associated symptoms being completely absent from all headache attacks if the patient were to experience a more severe headache attack. We speculate that nausea and vomiting improved more than other symptoms at least in part due to the effective acute treatment. Acute therapies are available that specifically target nausea and vomiting, unlike other associated symptoms. It may also be the case that symptoms like photophobia and phonophobia are more likely to be present both interictally and during attacks. The observation that CAMS improves with overall improvement makes it a promising candidate for an outcome measure in clinical trials. To date, clinical trial outcomes for migraine preventive therapies have focused on head pain including frequency, severity, and headache-related disability.¹⁹ However, people with lived experience consistently report that migraine-associated symptoms are an important aspect of migraine to treat effectively.⁹ Indeed, migraine-associated symptoms are already included in clinical trial design for migraine therapies focused on treating acute headache worsening with inclusion of resolution of “the most bothersome symptom.”²⁰

Youth who reported fewer migraine-related symptoms (i.e. lower CAMS 1 score) and fewer non-ICHD symptoms (i.e. higher CAMS 2 score) at presentation were statistically more likely to report improvement at follow-up. However, this significance was likely a result of our large sample size and is not clinically significant because CAMS could not reliably predict whether a patient reported improvement at follow-up. This suggests that the starting point of migraine-associated symptoms for each patient is not a strong indicator of whether they will improve. This finding is consistent with our prior work that did not find any headache characteristic predicted improvement in youth with new daily persistent headache, which included multiple metrics of headache burden including headache frequency, severity, and disability.²¹ However, our findings are contrary to models that used clinical data to reliably predict treatment response outcomes to both acute and preventive therapies in adults with migraine.^22–24 Predictive models have included headache characteristics, co-morbidities, and neuroimaging and primarily relied on machine learning techniques, which may be superior to standard logistic regression used in this study.²⁵ There may also be differences between the pediatric and adult populations. It is also important to note that we did not assess response to specific therapies. Headache program care is standardized with patients being offered a preventive medication and generally choosing amitriptyline, as well as non-pharmacologic interventions including healthy headache habit guidance and sometimes cognitive behavioral therapy. Therefore, it is possible that individuals with certain associated symptoms are more responsive to specific pharmacologic therapies, though this was out of the scope of the current study.

It is important to acknowledge the limitations of our study. This dataset is subject to selection bias because it only included those who presented to an initial visit and follow-up to a specialty headache clinic at a single center. This likely biases the sample to towards youth with more severe presenting symptoms and access to care. Therefore, results may not be generalizable to those with the milder migraine symptoms or those who face barriers to receiving specialty headache care. The CAMS model is limited by the symptoms included in the model (e.g. cranial autonomic symptoms associated with migraine were not assessed). None-the-less, CAMS behave in conjunction with other mainstay metrics used in migraine research including headache frequency, severity, and disability, indicating that it may be capturing an unrepresented dimension of migraine burden.

Conclusions

Experience of migraine-associated symptoms can change over time and decrease with overall improvement in symptoms of migraine. CAMS capture these differences and thus offer a promising candidate outcome in clinical trial design for migraine preventive therapies. However, migraine-associated symptoms as captured by CAMS at initial presentation were not predictive of who improved at follow-up.

Footnotes

Acknowledgements

The authors would like to thank the patients for their time filling out the questionnaire on their headache features. They would like to thank the CCHMC headache providers who guided patients through the intake questionnaire: Drs. Marielle Kabbouche, Joanne Kacperski, Jessica Saunders, and Hope O’Brien, and CCHMC headache fellows.

Abbreviations

Commercial relationships disclosures

C.P.G.: Dr Patterson Gentile is currently funded by the National Institutes of Health/National Institute of Neurological Disorders and Stroke (K23 NS124986) and the CHOP Foerderer Institutional grant.

C.L.S.: Dr Szperka has received research/grant support from the National Institutes of Health/National Institute of Neurological Disorders and Stroke (K23 NS102521), and PCORI. Dr Szperka or her institution have received compensation for her consulting work for Eli Lilly; Teva Pharmaceutical Industries Ltd; Upsher-Smith Laboratories, LLC; and Abbvie.

ADH.: Dr Hershey or his institution have received compensation for serving as a consultant for AbbVie, Amgen, Biohaven, Eli Lilly, Lundbeck, Supernus, Teva, Theranica and Upsher-Smith. His institution has also received research support from Amgen, Biohaven, Eli Lilly, Theranica, Upsher-Smith, and the NIH NINDS/NICHDS.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Ethical approval

The CCHMC Institutional Review Board approved the extraction of the data from the electronic health record into a research registry, with a waiver of verbal and written informed consent and assent to maximize generalizability.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (grant number K23NS124986).

Key findings

Migraine-associated symptoms as captured by CAMS change over time with fewer migraine-associated symptoms, particularly resolution of nausea and vomiting, being associated with improvement.

The number or type of migraine-associated symptoms burden does not reliably predict outcome.

ORCID iDs

Carlyn Patterson Gentile

Christina L. Szperka

Andrew D. Hershey

References

International Headache Society . The international classification of headache disorders, 3rd edition. Cephalalgia 2018; 38: 1–211.

Zanchin

Dainese

Trucco

, et al. Osmophobia in migraine and tension-type headache and its clinical features in patients with migraine. Cephalalgia 2007; 27: 1061–1068.

Hershey

Winner

Kabbouche

, et al. Use of the ICHD-II criteria in the diagnosis of pediatric migraine. Headache. [Online Serial] 2005; 45: 1288–1297. Accessed at: https://onlinelibrary.wiley.com/doi/10.1111/j.1526-4610.2005.00260.x.

Iljazi

Ashina

Lipton

, et al. Dizziness and vertigo during the prodromal phase and headache phase of migraine: a systematic review and meta-analysis. Cephalalgia 2020; 40: 1095–1103.

Lempert

Olesen

Furman

, et al. Vestibular migraine: diagnostic criteria (update). J Vestibular Res 2022; 32: 1–6.

Levin

. The international classification of headache disorders, 3rd edition (ICHD III)—changes and challenges. Headache 2013; 53: 1383–1395.

Cohen

Bigal

Newman

. Migraine and vestibular symptoms—identifying clinical features that predict “vestibular migraine”. Headache 2011; 51: 1393–1397.

Gerstein

Wirth

Uzumcu

, et al. Patient-reported experiences with migraine-related cognitive symptoms: results of the MiCOAS qualitative study. Headache 2023; 63: 441–454.

Urtecho

Wagner

Wang

, et al. A qualitative evidence synthesis of patient perspectives on migraine treatment features and outcomes. Headache 2023; 63: 185–201.

10.

Khayata

Farley

Davis

, et al. Beyond pain control: outcome and treatment preferences in pediatric migraine. Headache 2022; 62: 588–595.

11.

Patterson Gentile

Aguirre

Hershey

, et al. Symptoms associated with headache in youth. Cephalalgia 2023; 43: 03331024231187162.

12.

Patterson Gentile

Szperka

Hershey

. Cluster analysis of migraine-associated symptoms (CAMS) in youth: a retrospective cross-sectional multicenter study. Headache 2024; 64: 1230–1243.

13.

Patterson Gentile

Hershey

Szperka

. 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. Headache 2024; 64: 1217–1229.

14.

Hershey

Powers

Vockell

A-LB

, et al. PedMIDAS. Neurology 2001; 57: 2034–2037.

15.

Patterson-Gentile

. Headache Symptom Clusters Repository [online]. Accessed at: https://github.com/carlynpatterson/Headache-symptom-clusters.git.

16.

Trujillo-Ortiz

. Multiple Correspondence Analysis Based on the Indicator Matrix [online]. MATLAB Central File Exchange; 2021. Accessed at: https://www.mathworks.com/matlabcentral/fileexchange/22154-multiple-correspondence-analysis-based-on-the-indicator-matrix.

17.

Panagiotakos

Pitsavos

. Interpretation of epidemiological data using multiple correspondence analysis and log-linear models. J Data Sci 2004; 2: 75–86.

18.

Powers

Coffey

Chamberlin

, et al. Trial of amitriptyline, topiramate, and placebo for pediatric migraine. N Engl J Med 2017; 376: 115–124.

19.

McGinley

Houts

Nishida

, et al. Systematic review of outcomes and endpoints in preventive migraine clinical trials. Headache 2021; 61: 253–262.

20.

Houts

McGinley

Nishida

, et al. Systematic review of outcomes and endpoints in acute migraine clinical trials. Headache 2021; 61: 263–275.

21.

Szperka

Evans

Patterson Gentile

, et al. Headache trajectories in children and adolescents with new onset continuous headache. Cephalalgia 2024; 44: 03331024241282803.

22.

Chiang

Schwedt

Dumkrieger

, et al. Advancing toward precision migraine treatment: predicting responses to preventive medications with machine learning models based on patient and migraine features. Headache 2024; 64: 1094–1108.

23.

Ezzati

Fanning

Buse

, et al. Predictive models for determining treatment response to nonprescription acute medications in migraine: results from the American migraine prevalence and prevention study. Headache 2022; 62: 755–765.

24.

Ezzati

Buse

Fanning

, et al. Predictors of treatment-response to acute prescription medications in migraine: results from the American migraine prevalence and prevention (AMPP) study. Clin Neurol Neurosurg 2022; 223: 107511.

25.

Chen

Zhang

Cao

, et al. Prediction models for treatment response in migraine: a systematic review and meta-analysis. J Headache Pain 2025; 26: 32.