Share

December 10, 2019; 93 (24) ArticleOpen Access

Migraine progression in subgroups of migraine based on comorbidities

Results of the CaMEO Study

Richard B. Lipton, Kristina M. Fanning, Dawn C. Buse, Vincent T. Martin, Lee B. Hohaia, Aubrey Manack Adams, Michael L. Reed, Peter J. Goadsby
First published November 5, 2019, DOI: https://doi.org/10.1212/WNL.0000000000008589
Full PDF
Short Form
Citation
Permissions

Make Comment

See Comments

Downloads
783

Share

Abstract

Objective To test the hypothesis that statistically defined subgroups of migraine (based on constellations of comorbidities and concomitant conditions; henceforth comorbidities), previously identified using Chronic Migraine Epidemiology and Outcomes (CaMEO) Study data, differ in prognosis, as measured by rates of progression from episodic migraine (EM) to chronic migraine (CM).

Methods The onset of CM was assessed up to 4 times over 12 months in individuals with EM and ≥1 comorbidity at baseline, based on constellations of comorbidities (comorbidity classes). The “fewest comorbidities” class served as reference. Individuals completing ≥1 follow-up survey from the web-based CaMEO Study were included. Covariates included sociodemographic variables and headache characteristics. Sex, income, cutaneous allodynia, and medication overuse were modeled as binary variables; age, body mass index, headache-related disability (Migraine Disability Assessment [MIDAS]), and Migraine Symptom Severity Scale as continuous variables. CM onset was assessed using discrete time analysis.

Results In the final sociodemographic model, all comorbidity classes had significantly elevated hazard ratios (HRs) for risk of progression to CM from EM, relative to fewest comorbidities. HRs for CM onset ranged from 5.34 (95% confidence interval [CI] 3.89–7.33; p ≤ 0.001) for most comorbidities to 1.53 (95% CI 1.17–2.01; p < 0.05) for the respiratory class. After adjusting for headache covariates independently, each comorbidity class significantly predicted CM onset, although HRs were attenuated.

Conclusions Subgroups of migraine identified by comorbidity classes at cross-section predicted progression from EM (with ≥1 comorbidity at baseline) to CM. The relationship of comorbidity group to CM onset remained after adjusting for indicators of migraine severity, such as MIDAS.

Clinicaltrials.gov identifier NCT01648530.

Glossary

AMPP=
American Migraine Prevalence and Prevention;
BMI=
body mass index;
CaMEO=
Chronic Migraine Epidemiology and Outcomes;
CI=
confidence interval;
CM=
chronic migraine;
EM=
episodic migraine;
HR=
hazard ratio;
ICHD-3=
International Classification of Headache Diseases, 3rd Edition;
LCA=
latent class analysis;
MIDAS=
Migraine Disability Assessment;
MSSS=
Migraine Symptom Severity Scale

Migraine is a complex neurologic disease with multiple environmental and genetic risk factors and substantial phenotypic heterogeneity.1 The natural disease course of episodic migraine (EM) can involve the progression to chronic migraine (CM), remission to less frequent or no migraine, or change of headache type.2 Increased headache day frequency, medication overuse, depression, and cutaneous allodynia are associated with progression to CM.3,4 MRI5,6 and gene association studies1 have previously identified subgroups of migraine with features predisposing to progression to CM.

Migraine is also associated with concomitant conditions or comorbidities (conditions occurring with migraine at a rate greater than expected based on chance alone7) (hereafter comorbidities) such as asthma,8 rhinitis,9 depression and anxiety,10 and chronic pain disorders,11 with asthma,12 depression,13 and noncephalic pain disorders14 identified as predictors of progression to CM. Results from previous research using cluster analysis suggest that people from subgroups of migraine defined based on comorbidity profiles may differ in prognosis, response to treatment, and underlying biology.15

The Chronic Migraine Epidemiology and Outcomes (CaMEO) Study was a web-based survey study using cross-sectional modules with longitudinal follow-up assessments. It was designed to characterize the natural course of migraine and describe related comorbidities and patterns of treatment in a representative US sample of people with migraine.16 Previously, data from the CaMEO Study were modeled using latent class analysis (LCA) to identify subgroups of migraine based on comorbidity profiles.7 These subgroups differed in demographic profiles, disability, and headache characteristics. The present analysis uses discrete time hazard modeling to estimate comorbid health predictors of progression from EM to CM.

Methods

Study design

The CaMEO Study (ClinicalTrials.gov identifier: NCT01648530) was an Internet-based study with cross-sectional modules and longitudinal follow-ups designed to assess a range of data including comorbidities and migraine characteristics, including progression of migraine to CM over 1 year.16 Recruiting and screening occurred between September and October 2012. Participants were recruited using quota sampling, to ensure that participants of the invited sample were balanced against key population demographics, from an Internet research panel (Research Now, Plano, TX) that included 2.4 million active US members. A total of 489,537 members of the panel were invited to participate; 80,783 (16.5%) responded, 58,418 (72.3% of respondents) provided usable surveys for analysis (i.e., ≥20% of the survey was complete and headache status could be determined), and 16,789 (28.7% of those respondents with usable data) met study inclusion criteria, reflecting modified International Classification of Headache Diseases, 3rd Edition (ICHD-3)17 criteria for migraine.

Study participants

Respondents were eligible for inclusion in the study if they met CaMEO Study criteria and volunteered to participate.16 Migraine criteria were assessed using the validated American Migraine Study/American Migraine Prevalence and Prevention (AMPP) Study diagnostic screener,18,–,20 which uses a modification of the ICHD-3 migraine criteria,17 and has been demonstrated previously to have a sensitivity of 1.00 and a specificity of 0.82 for the diagnosis of migraine when used in self-report populations.18 Respondents with migraine were defined as those meeting modified ICHD-3 migraine criteria, and those with CM also had ≥15 headache days per month averaged over the preceding 3 months based on Silberstein-Lipton criteria.21 This approach, using patient self-report, has been demonstrated to have a sensitivity of 0.90 and a specificity of 0.83, with a good to excellent intertest reliability (κ = 0.76) for CM.22 Although participation rates in the Internet-based CaMEO Study were modest compared to those of the mail survey–based AMPP Study (16.5% vs 64.9%, respectively), formal comparisons of baseline demographics and headache characteristics of the study populations demonstrated that they were similar, leading to the conclusion that the CaMEO Study respondents were representative of the US population with migraine.23

Assessments

The Screening, Core, and Barriers to Care modules together represent baseline assessments from the CaMEO Study cohort and provided data for respondent demographics and headache characteristics and treatment. Comorbidities were collected at stage 2 of the CaMEO Study via a questionnaire designed using validated instruments where possible and a list of questions on common conditions (Comorbidities/Endophenotype module). Self-reports of 62 different symptoms and conditions were available for the analysis and those conditions judged by clinical experts to require a medical diagnosis for reliable reporting were assessed based on self-report of a physician diagnosis. Data from Screening, Core, Barriers to Care, and Comorbidities/Endophenotype modules were used to inform the LCA and identify subgroups of people with migraine based on comorbidity profiles, providing the main variable of interest for this current analysis. Cross-sectional data from the Core/Snapshot modules collected every 3 months for 12 months were used in this analysis to inform progression from migraine to CM.

Standard protocol approvals, registrations, and patient consents

Data included in this analysis were from the CaMEO Study. The CaMEO Study was approved by the institutional review board of the Albert Einstein College of Medicine, which waived written informed consent for study volunteers who had the right to accept or refuse participation in the survey.

Latent class analysis

Using LCA, we identified 8 subgroups of migraine based on patients reporting ≥1 comorbidities.7 Preliminary LCA modeling was undertaken with 62 comorbidity variables and patients reporting ≥1 comorbidity, identifying the 8-class model as the model of best fit. Subsequently, variable reduction was undertaken to help enhance the interpretability of the model, leaving 22 comorbidity variables. LCA modeling in combination with clinical judgment identified the 8-class model as having the best fit to the data and the most distinctive classes. The classes identified by LCA and being further investigated in this report are as follows: class 1, many comorbidities (most comorbidities); class 2, respiratory/psychiatric (resp/psych); class 3, respiratory/pain (resp/pain); class 4, respiratory; class 5, psychiatric; class 6, cardiovascular; class 7, pain; class 8, few comorbidities (fewest comorbidities) (figure).

Figure
Figure Conditional probability that a member of a comorbidity class would self-report having given comorbidity/concomitant condition

The heat map is based on the comorbidity classes of migraine as derived by latent class analysis of Chronic Migraine Epidemiology and Outcomes (CaMEO) comorbidities/endophenotype data. The heat map was created by assigning the color green to the cell that holds the minimum probability value for all comorbidities and classes. The cell that holds the median probability is colored yellow, and the cell that holds the maximum value is colored red. All other cells are colored proportionally. Psych = psychiatric; PTSD = posttraumatic stress disorder; Resp = respiratory; SR = self-reported symptoms; SR-PD = self-reported physician diagnosis of condition. Reproduced with permission from Lipton RB, Fanning KM, Buse DC, et al. Identifying natural subgroups of migraine based on comorbidity and concomitant condition profiles: results of the Chronic Migraine Epidemiology and Outcomes (CaMEO) Study. Headache 2018;58:933–947.7

Discrete time hazard analysis

The discrete time hazard model is an extension of the proportional hazard model, incorporating conditional odds of an event occurring in a series of subsequent discrete time periods. This model accommodates censored data (i.e., when information about the event is incomplete because it is missing or occurs outside of the observation period). For this analysis, a Cox model with a discrete time variable was used, as data were collected at 3-month intervals. The main statistic reported is the hazard ratio (HR).

Analysis approach

The LCA-derived comorbidity classes were used as the basis for discrete time hazard models examining time to CM onset across the 8 comorbidity classes over 4 follow-up periods (i.e., 3, 6, 9, and 12 months after the baseline survey). The analysis sample included those with ≥1 of 62 comorbidities included in the LCA and ≥1 wave of follow-up data beyond baseline. As the analysis sought to predict CM onset, those who had CM at baseline were also excluded.

A range of sociodemographic and headache characteristics were used as covariates, including age, body mass index (BMI), sex, income (<$50,000 [40.9% of respondents] vs ≥$50,000 [59.1% of respondents]), race, Migraine Disability Assessment (MIDAS), allodynia (Allodynia Symptom Checklist [<3 indicates no allodynia vs ≥3 indicates mild or worse allodynia24]), the Migraine Symptom Severity Score (MSSS), and medication overuse at baseline. The MSSS is a composite index that incorporates the frequency of 7 primary migraine features (unilateral pain, pulsatile pain, moderate or severe pain intensity, routine activities worsen pain, nausea, photophobia, and phonophobia). Possible responses were 1 = never, 2 = rarely, 3 = less than half the time, or 4 = half the time. The overall MSSS score ranges from 7 to 28 and was calculated by summing scores for each of the 7 headache features assessed. Medication overuse was assessed as regular overuse of acute migraine treatment for >3 months including acetylsalicylic acid, nonsteroidal anti-inflammatory drugs, and acetaminophen [paracetamol] for ≥15 d/mo and ergotamines, triptans, opioids, or combination analgesics on ≥10 d/mo.17 Sex, income, allodynia, and medication overuse were modeled as binary variables, while age, BMI, MIDAS, and MSSS were modeled as continuous variables, with age and MIDAS being modeled based on 10-year age and 10-point groupings, respectively. The LCA-derived comorbidity class was the main predictor variable of interest, with class 8 (fewest comorbidities class) serving as the reference group. The comorbidity classes were included as a single variable.

Three different modeling approaches were undertaken to ensure validity of the results. Discrete time models were run in a nested fashion; sociodemographics were entered first, followed by headache characteristics and treatment variables (headache-related disability [MIDAS], MSSS, allodynia, and medication overuse), which were entered one at a time, then removed before adding the next variable. These models were assessed to determine if headache characteristics accounted for the majority of the variation in progression to CM observed. Variables that did not significantly contribute (i.e., p values > 0.05) were trimmed from final models. After sociodemographic and headache variable models were completed, all significant variables were entered into a set of forward and backward stepwise models. Forward stepwise regression is an automated process that first enters the variable whose inclusion gives the most statistically significant improvement in model fit and then continues to add variables until there is no further significant improvement in fit. Backwards regression starts with all variables included in the model and takes out the variable whose removal results in least deterioration in model fit, continuing until no additional variable can be removed without significant loss of model fit. p < 0.05 was considered statistically significant. Missing data were negligible (<5%) and no imputation measures were employed. All analyses were performed using IBM SPSS Statistics, version 20.0 (IBM, Armonk, NY; 2011).

Data availability

Data reported in this manuscript are available within the article. Additional data from the CaMEO Study may be requested at allerganclinicaltrials.com/PatientDataRequest.htm.

Results

Analysis population

A total of 58,418 individuals provided valid returns to CaMEO baseline surveys; 16,763 individuals met case definition for migraine and were sent the CaMEO Study Comorbidities/Endophenotype module, providing a total of 12,810 valid returns.16 We excluded persons with CM at baseline (n = 1,111), persons who did not complete any follow-up survey (n = 2,296), and, per the initial LCA, persons free of comorbidities (n = 745), leaving an eligible sample of 8,658 individuals with EM.

Sociodemographics

The final population for this analysis had a mean (SD) age of 43.2 (14.7) years, a mean (SD) BMI of 28.6 (7.6)kg/m2, and 75.0% were women (table 1). The distribution of individuals across comorbidity classes was variable, from 409 (4.7%) members of class 1 (most comorbidities) to 3,054 (35.3%) members of the reference group (class 8, fewest comorbidities). Demographic and clinical characteristics varied across classes (table 1); class 5 (psychiatric) had the youngest mean (SD) age, 36.3 (12.4) years, and class 6 (cardiovascular) the highest (55.0 [12.2] years). Compared with other classes, members of class 6 (cardiovascular) were more likely to be men (41.4%) and those of class 2 (respiratory/psychiatric) were most likely to be women (85.3%). Individuals in class 1 (most comorbidities) had the highest mean (SD) BMI (32.4 [8.4]kg/m2) and class 8 (fewest comorbidities) the lowest (27.2 [7.0]kg/m2).

View this table:
Table 1

Demographics and clinical characteristics at baseline of the comorbidity classes in the final analysis sample

Headache characteristics

Similarly, headache characteristics varied across the comorbidity classes (table 1). The median MIDAS score was 6.0; 2,960 respondents (34.2%) were MIDAS grade III or IV at baseline. The mean (SD) MSSS score was 15.3 (3.2), with mean (SD) monthly headache days of 3.5 (3.2). Allodynia was reported by 3,844 (44.4%) respondents, and 1,028 (11.9%) met criteria for medication overuse (table 1).

Class 1 (most comorbidities) was most likely to be associated with severe headache-related disability (MIDAS grade IV; 37.0%); class 6 (cardiovascular, 10.9%) and 8 (fewest comorbidities, 12.4%) were least likely. Similarly, allodynia was most likely to be observed in members from class 1 (most comorbidities, 60.1%), and least likely in members from class 6 (cardiovascular, 36.0%), and class 8 (fewest comorbidities, 36.6%). Reported medication overuse was highest in class 1 (most comorbidities, 27.6%) and lowest for class 8 (fewest comorbidities, 7.6%) and class 4 (respiratory, 9.8%).

Discrete time hazard models

Initial CM progression hazard models were performed with only the LCA comorbidity classes and sociodemographic variables (i.e., sex, age, race, income, and BMI) included. Race did not significantly contribute (HR [95% confidence interval (CI)] 1.05 [0.82–1.35]) so it was trimmed from the model. The HR for sex was also not significant (1.20 [95% CI 0.99–1.45]); however, it was left in the model because sex differences contribute to migraine epidemiology and comorbidity profiles.25 The final sociodemographic model included age, which was associated with a 14% decrease in the hazard of progressing to CM over 12 months for each change in age by 10 years (HR 0.86 [95% CI 0.81–0.92]).

In the final sociodemographic model, all comorbidity classes had significantly elevated HRs for risk of progression to CM from EM, relative to the fewest comorbidities class (table 2). HRs for CM onset ranged from 5.34 (95% CI 3.89–7.33; p ≤ 0.001) for the most comorbidities class to 1.53 (95% CI 1.17–2.01; p < 0.05) for the respiratory class (table 2). The HR for CM onset was second highest in the class with a combination of respiratory and pain comorbidities (respiratory/pain class: HR 3.64 [95% CI 2.67–4.98]).

View this table:
Table 2

Discrete time hazard to chronic migraine onset: reference group fewest comorbidities class (n = 3,054 [fewest comorbidities class]/8,658 [total], 35.3%)

Adjusting for headache covariates (e.g., headache-related disability [MIDAS], MSSS, allodynia, and medication overuse) independently resulted in attenuation of the relationship between comorbidity classes and progression to CM; however, the influence of each covariate on progression to CM relative to class 8 (fewest comorbidities) remained significant (table 3). For example, adjustment for headache-related disability (MIDAS) attenuated the HR for the most comorbidities class (from 5.34 [95% CI 3.89–7.33] to 3.95 [95% CI 2.85–5.48]), as did adjustment for medication overuse (from 5.34 [95% CI 3.89–7.33] to 4.01 [95% CI 2.92–5.51]). Adjustment for MSSS and allodynia also attenuated the HR for the most comorbidities class, although to a lesser degree than that observed with headache-related disability (MIDAS) and medication overuse (table 3).

View this table:
Table 3

Five separate models for discrete time hazard to chronic migraine onset in comorbidity classes of migraine in individuals with episodic migraine at baseline

In forward stepwise models, headache-related disability (MIDAS) came into the model first as it provided the most statistical improvement in model fit (table 4). Medication overuse was included next, followed by comorbidity class variable, allodynia, income, age, sex, and then MSSS. Only BMI and race were not added in the forward stepwise model. In the final forward stepwise model, all comorbidity classes were attenuated; however, class 1 (most comorbidities) retained the highest risk of progression to CM, 3 times higher than the reference fewest comorbidities class (HR 3.01 [95% CI 2.17–4.18]). The addition of age tended to increase the HR for the comorbidity classes; for example, the HR for most comorbidities increased to 3.02 (95% CI 2.17–4.20) from 2.49 (95% CI 1.83–3.39; table 4). Cardiovascular and pain classes lost statistical significance in the forward stepwise models, except after addition of age, where the HRs retained significance (cardiovascular 1.46 [95% CI 1.00–2.13]; pain 1.48 [95% CI 1.01–2.17]).

View this table:
Table 4

Forward stepwise model for the discrete time hazard to chronic migraine onset in comorbidity classes of migraine in individuals with episodic migraine at baseline

The results of the backward stepwise model converged on those of the forward stepwise model (table 5), with most comorbidities, respiratory /pain, and psychiatric classes having the highest risk of progression of all comorbidity classes.

View this table:
Table 5

Comparison of the results of backward and forward stepwise models

Discussion

Discrete time analysis demonstrated that LCA-derived comorbidity classes of migraine, composed of naturally occurring constellations of comorbidities and concomitant conditions, were associated with different rates of risk of progression from EM to CM over 12 months, and results converged regardless of which modeling approach applied. When adjusting only for sociodemographic variables, all comorbidity classes were associated with a statistically significant risk of progression to CM. However, there were differences among comorbidity classes. For example, members of the most comorbidities class were approximately 5 times more likely to progress to CM than members of the fewest comorbidities class. The respiratory class was the least likely of the 7 comorbidity classes to progress to CM over time; however, members of this class were still 1.5 times more likely than members of the fewest comorbidities reference group to progress to CM.

The addition of individual headache features in general, and headache-related disability (MIDAS) in particular, attenuated the risk of progression for all comorbidity classes, most prominently for the most comorbidities class. Nonetheless, even with the addition of all the headache features in the final step of the forward stepwise model, the HR for the progression to CM remained >1 for all comorbidity classes, with the lowest HR never falling below 1.41 (95% CI 0.96–2.05, cardiovascular class). This suggests that the effect of comorbidity class membership on progression to CM is moderately large, and relatively independent of headache characteristics.

We did not adjust for headache day frequency specifically; however, the observation that the addition of headache-related disability (MIDAS) to the model attenuates the HRs for many of the comorbidity classes is likely due, at least in part, to the correlation between headache days and total MIDAS score.26 Importantly, the addition of headache-related disability (MIDAS) and therefore partially adjusting for underlying headache days did not fully explain the association among comorbidity classes and risk of progression to CM. Of note, although the HR for MIDAS per 10-point change is only 1.11, this represents risk of progression to CM for each 10-point change in MIDAS. For a 20-point change in MIDAS, the HR would be 1.22; for a 40-point change in MIDAS, the HR would be 1.49. Indeed, headache-related disability (MIDAS) as an individual covariate explained more variability in the progression to CM than any other covariate, and was the first covariate entered into the forward stepwise model. Similarly, the HR for age was based on risk of progression to CM for each 10-year change in age, while the HR for MSSS and BMI was based on risk of progression to CM for each 1-point change in MSSS or BMI. In contrast, allodynia and medication overuse were modeled as binary variables and HRs represent increased risk of progression for those with vs without allodynia or with vs without medication overuse.

Others have reported an association between allodynia and certain subgroups of migraine.6,27 Given the association between allodynia and central sensitization,28 it might be expected that allodynia would attenuate HRs for members of classes with pain comorbidities (i.e., respiratory/pain and pain classes). While allodynia attenuated the HR, to a degree, the magnitude of the effect was relatively uniform across comorbidity classes. Medication overuse, like headache-related disability (MIDAS), was a powerful predictor of progression (HR 4.06 [95% CI 3.41–4.84]), confirming the findings of others.29,–,31 The addition of medication overuse attenuated the HR for the most comorbidities and respiratory classes in particular. It is unclear from our analysis whether medication overuse is acting as a confounder (medication overuse is associated with comorbidity class and risk of progression) or a mediator (persons in particular comorbidity classes, for example the psychiatric class, tend to overuse medication, which in turns increases the risk of progression). We did not explore the effect of the type of medication being overused and the risk of progression for the subclasses of migraine; this could be usefully explored in the future.

It is interesting to note that when pain and respiratory conditions occurred in relative isolation, they had relatively low HRs (approximately 1.4) for progression to CM in forward stepwise models compared with the other subgroups. When these comorbidities occurred with other conditions, they were associated with higher HRs for progression to CM. For example, the HR for the respiratory/psychiatric subgroup was 1.87 (95% CI 1.40–2.50) and that for the respiratory/pain subgroup was 2.57 (95% CI 1.88–3.51). This suggests that when certain comorbidities occur together, the risk for progression to CM is increased.

As discussed previously, the rationale for the apparent grouping of migraine based on comorbidities could be variable.7 It is possible that some of the comorbidities are actually symptoms of migraine per se and not separate conditions, while others may share underlying pathophysiologic mechanisms with migraine.

Study strengths and limitations

An overview of limitations of the primary CaMEO Study has been reported previously, including relatively low response rates to the initial invitation resulting in potential selection bias and all data being self-reported.7,16 Despite these potential limitations, the web-based longitudinal design enabled the collection of comprehensive information on respondents with migraine over 12 months, and for this analysis there were substantial numbers of individuals in each comorbidity class. Nonresponse bias for the CaMEO respondents was assessed through comparison of demographics and disease severity between respondents and nonrespondents using data from a follow-up survey to nonrespondents.16 Characteristics were found to be similar across respondents and nonrespondents; however, the percentage of individuals with CM and those with headache in the last 30 or 90 days was higher among respondents than nonrespondents. Furthermore, the low response rate to the nonrespondents' survey leaves open the possibility of nonrespondent bias. However, the baseline demographics and headache characteristics of the CaMEO population and the AMPP population were similar,23 demonstrating that, since the AMPP Study sample, with a response rate of 64.8%, is considered representative of the US population with migraine, the CaMEO data could also be generalizable to the US population with migraine. Another limitation is the relatively brief duration of follow-up.

It should also be noted that we elected to include BMI as a continuous variable rather than dichotomizing it in 2 groups: obese and less than obese. Generally, the continuous variable BMI provides a more sensitive adjustment because information is coarsened when it is dichotomized. It is possible that including BMI as a continuous variable may have resulted in missing a potential effect of obesity. One large population study showed that BMI was not associated with migraine prevalence after adjustments for age, race, and education; in contrast, increasing BMI, using normal weight as a reference, was associated with high attack frequency and other clinical features of migraine.32 The greatest effects tended to occur in those with a BMI of 30–34.9 kg/m2 (obese) or a BMI ≥35 kg/m2 (morbidly obese). A more recent meta-analysis of 12 studies indicated that both obesity (BMI ≥30) and underweight (BMI <18.5) were associated with the risk of migraine.33

There are a number of additional limitations. It should be noted that we relied on self-reported medical diagnosis of comorbidities and concomitant conditions, which could lead to either underascertainment or overascertainment of specific comorbidities.7,34 A more robust approach might have relied on systematic diagnostic assessments, an important step for a future study. In addition, a form of Berkson bias could influence our results. Individuals with more severe migraine may have visited their health professional more frequently and, as a result, might have been more likely to have been diagnosed with a concomitant condition and fall into the most comorbidities class. In this circumstance, severe migraine and not the comorbidity class could have been the driver of progression. However, for most of the comorbid subgroups, risk of progression remained elevated after adjustment for measures of headache severity.

Second, LCA identified comorbid subgroups with overlapping comorbidity profiles; for example, the respiratory/psychiatric and respiratory/pain classes. This requires cautious interpretation. Our results do not reflect the effect of a comorbidity per se on disease progression, but rather the effect of comorbidity class membership on clinical course of disease.

Third, we have adjusted for a number of headache characteristics assuming that they may act as confounders and explain the change in risk of progression to CM in the subgroups. As noted above, some of these features may be mediators; they may be in the causal pathway linking the comorbid subgroup to disease progression.35 In addition, we cannot exclude the possibility that there may be potential mediators and confounders we did not include.

Fourth, as previously reported, we defined our subgroups of migraines based on a cross-section of self-reported comorbidities rather than on other markers of disease.7 We could have selected other measures for identifying subgroups including migraine symptom profiles,36 measures of brain structure,5,6 or treatment response.3,36 Ultimately, subgroups of migraine could also be defined based on longitudinal data using latent trajectory modeling, and the effect of class membership on the treatment response, neuroimaging, and biologic markers could be further explored.

Finally, we only considered progression from EM to CM and not fluctuation between EM and CM. The fact that patients alternate between periods of EM and CM is well-recognized37; the effect of subgroups of migraine on fluctuation between EM and CM is worthy of further exploration.

Despite the limitations outlined above, our results are an important illustration that we can define subgroups of migraine based on cross-sectional comorbidity profiles, and that membership of these subgroups predicts disease progression over time.

By identifying comorbidity classes in migraine and observing their relationship with changes in disease over time, we hope to understand more about the underlying heterogeneity of migraine and identify the genetic and biologic features for each different class. The identification of homogeneous subgroups of migraine will most certainly be an iterative process; this analysis forms an important early step in identification of homogeneous subclasses of migraine. The clinical importance of subgroups we have identified is demonstrated by the differences among classes in the risk of progression to CM over time. The relationship of comorbidity group to CM onset remained after adjusting for indicators of migraine severity and other potential confounders (e.g., headache-related disability [MIDAS], MSSS, allodynia, and medication overuse). When these were added to the model to explain some of the observed differences, comorbidity classes still differed in their risk of progression to CM, suggesting that there are underlying biologic or genetic similarities linking members of each class. As a next step, external validation of comorbidity classes will be undertaken to determine whether we can predict treatment responses to different types of treatment based on comorbidity class.

Study funding

This study was funded by Allergan plc, Dublin, Ireland. The Article Processing Charge was funded by Allergan plc.

Disclosure

R. Lipton serves on the editorial boards of Neurology® and Cephalalgia and as senior advisor to Headache; has received research support from the NIH; receives support from the Migraine Research Foundation and the National Headache Foundation; has reviewed for the NIA and NINDS; serves as consultant, advisory board member, or has received honoraria from Alder, Allergan, Amgen, Autonomic Technologies, Avanir, Biohaven, Biovision, Boston Scientific, Dr. Reddy's Laboratories, Electrocore, Eli Lilly, eNeura Therapeutics, GlaxoSmithKline, Merck, Novartis, Pernix, Pfizer, Supernus, Teva, Vector, and Vedanta; receives royalties from Wolff's Headache (8th Edition, Oxford University Press), Informa, and Wiley; and holds stock options in eNeura Therapeutics and Biohaven. K. Fanning is an employee of Vedanta Research, which has received support funded by Allergan, Amgen, Promius, Eli Lilly, GlaxoSmithKline, and Merck & Co., Inc., via grants to the National Headache Foundation. D. Buse has received grant support and honoraria from Allergan, Avanir, Amgen, Eli Lilly and Company, Teva, and Promius; and is on the editorial board of Current Pain and Headache Reports. V. Martin has been a consultant for Allergan, Amgen, Alder, Avanir, Biohaven, Promius, Supernus, Eli Lilly, and Teva, and a speaker for Allergan, Amgen, Avanir, Eli Lilly, Depomed, and Pfizer. L. Hohaia is an employee of CHC Group, LLC, an ICON plc company, which has received funding from Allergan for manuscript development. A. Manack Adams is a full-time employee of Allergan plc and owns stock in the company. M. Reed is Managing Director of Vedanta Research, which has received research funding from Allergan, Amgen, Promius, Eli Lilly, GlaxoSmithKline, and Merck & Co., Inc., via grants to the National Headache Foundation. Vedanta Research has received funding directly from Allergan for work on the CaMEO Study. P. Goadsby reports personal fees from Alder Biopharmaceuticals, Autonomic Technologies Inc., Biohaven Pharmaceuticals Inc., Dr. Reddy's Laboratories, Electrocore LLC, Novartis, Scion, Teva Pharmaceuticals, Trigemina Inc., MedicoLegal work, Massachusetts Medical Society, Up-to-Date, Oxford University Press, and Wolters Kluwer outside the submitted work; grants and personal fees from Amgen, Eli Lilly and Company, and eNeura Inc.; and has a patent for magnetic stimulation for headache assigned to eNeura without fee. Go to Neurology.org/N for full disclosures.

Acknowledgment

Writing and editorial support was provided by Lisa Feder, PhD, of Peloton Advantage, LLC, an OPEN Health company (Parsippany, NJ), and Gerard P. Johnson, PhD, of CHC Group, LLC (North Wales, PA), an ICON plc company, and was funded by Allergan plc, Dublin, Ireland.

Appendix Author contributions

Table6

Footnotes

  • Go to Neurology.org/N for full disclosures. Funding information and disclosures deemed relevant by the authors, if any, are provided at the end of the article.

  • The Article Processing Charge was funded by the Allergan plc.

  • Received January 10, 2019.
  • Accepted in final form June 26, 2019.

This is an open-access article distributed under the terms of the Creative Commons Attribution-Non Commercial-No Derivatives License 4.0 (CCBY-NC-ND), where it is permissible to download and share the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal.

References

  1. 1.↵
    1. Louter MA,
    2. Fernandez-Morales J,
    3. de Vries B, et al
    . Candidate-gene association study searching for genetic factors involved in migraine chronification. Cephalalgia 2015;35:500–507.
    OpenUrlCrossRefPubMed
  2. 2.↵
    1. Dahlof CG,
    2. Johansson M,
    3. Casserstedt S,
    4. Motallebzadeh T
    . The course of frequent episodic migraine in a large headache clinic population: a 12-year retrospective follow-up study. Headache 2009;49:1144–1152.
    OpenUrlCrossRefPubMed
  3. 3.↵
    1. Bigal ME,
    2. Lipton RB
    . Migraine chronification. Curr Neurol Neurosci Rep 2011;11:139–148.
    OpenUrlCrossRefPubMed
  4. 4.↵
    1. Buse DC,
    2. Greisman JD,
    3. Baigi K,
    4. Lipton RB
    . Migraine progression: a systematic review. Headache 2019;59:306–338.
    OpenUrl
  5. 5.↵
    1. Chong CD,
    2. Gaw N,
    3. Fu Y,
    4. Li J,
    5. Wu T,
    6. Schwedt TJ
    . Migraine classification using magnetic resonance imaging resting-state functional connectivity data. Cephalalgia 2017;37:828–844.
    OpenUrl
  6. 6.↵
    1. Schwedt TJ,
    2. Si B,
    3. Li J,
    4. Wu T,
    5. Chong CD
    . Migraine subclassification via a data-driven automated approach using multimodality factor mixture modeling of brain structure measurements. Headache 2017;57:1051–1064.
    OpenUrl
  7. 7.↵
    1. Lipton RB,
    2. Fanning KM,
    3. Buse DC, et al
    . Identifying natural subgroups of migraine based on comorbidity and concomitant condition profiles: results of the Chronic Migraine Epidemiology and Outcomes (CaMEO) Study. Headache 2018;58:933–947.
    OpenUrl
  8. 8.↵
    1. Peng YH,
    2. Chen KF,
    3. Liao WC, et al
    . Association of migraine with asthma risk: a retrospective population-based cohort study. Clin Respir J 2018;12:1030–1037.
    OpenUrl
  9. 9.↵
    1. Martin VT,
    2. Fanning KM,
    3. Serrano D, et al
    . Chronic rhinitis and its association with headache frequency and disability in persons with migraine: results of the American Migraine Prevalence and Prevention (AMPP) Study. Cephalalgia 2014;34:336–348.
    OpenUrlCrossRefPubMed
  10. 10.↵
    1. Chen YC,
    2. Tang CH,
    3. Ng K,
    4. Wang SJ
    . Comorbidity profiles of chronic migraine sufferers in a national database in Taiwan. J Headache Pain 2012;13:311–319.
    OpenUrlCrossRefPubMed
  11. 11.↵
    1. Plesh O,
    2. Adams SH,
    3. Gansky SA
    . Self-reported comorbid pains in severe headaches or migraines in a US national sample. Headache 2012;52:946–956.
    OpenUrlCrossRefPubMed
  12. 12.↵
    1. Martin VT,
    2. Fanning KM,
    3. Serrano D,
    4. Buse DC,
    5. Reed ML,
    6. Lipton RB
    . Asthma is a risk factor for new onset chronic migraine: results from the American Migraine Prevalence and Prevention Study. Headache 2016;56:118–131.
    OpenUrl
  13. 13.↵
    1. Ashina S,
    2. Serrano D,
    3. Lipton RB, et al
    . Depression and risk of transformation of episodic to chronic migraine. J Headache Pain 2012;13:615–624.
    OpenUrlCrossRefPubMed
  14. 14.↵
    1. Scher AI,
    2. Buse DC,
    3. Fanning KM, et al
    . Comorbid pain and migraine chronicity: the Chronic Migraine Epidemiology and Outcomes Study. Neurology 2017;89:461–468.
    OpenUrlAbstract/FREE Full Text
  15. 15.↵
    1. Tietjen GE,
    2. Herial NA,
    3. Hardgrove J,
    4. Utley C,
    5. White L
    . Migraine comorbidity constellations. Headache 2007;47:857–865.
    OpenUrlCrossRefPubMed
  16. 16.↵
    1. Manack Adams A,
    2. Serrano D,
    3. Buse DC, et al
    . The impact of chronic migraine: the Chronic Migraine Epidemiology and Outcomes (CaMEO) Study: methods and baseline results. Cephalalgia 2015;35:563–578.
    OpenUrlCrossRefPubMed
  17. 17.↵
    Headache Classification Committee of the International Headache Society. The International Classification of Headache Disorders, 3rd edition. Cephalalgia 2018;38:1–211.
    OpenUrlCrossRefPubMed
  18. 18.↵
    1. Lipton RB,
    2. Stewart WF,
    3. Diamond S,
    4. Diamond ML,
    5. Reed M
    . Prevalence and burden of migraine in the United States: data from the American Migraine Study II. Headache 2001;41:646–657.
    OpenUrlCrossRefPubMed
  19. 19.↵
    1. Stewart WF,
    2. Lipton RB,
    3. Celentano DD,
    4. Reed ML
    . Prevalence of migraine headache in the United States: relation to age, income, race, and other sociodemographic factors. JAMA 1992;267:64–69.
    OpenUrlCrossRefPubMed
  20. 20.↵
    1. Silberstein S,
    2. Loder E,
    3. Diamond S,
    4. Reed ML,
    5. Bigal ME,
    6. Lipton RB
    . Probable migraine in the United States: results of the American Migraine Prevalence and Prevention (AMPP) study. Cephalalgia 2007;27:220–229.
    OpenUrlCrossRefPubMed
  21. 21.↵
    1. Silberstein SD,
    2. Lipton RB,
    3. Sliwinski M
    . Classification of daily and near-daily headaches: field trial of revised IHS criteria. Neurology 1996;47:871–875.
    OpenUrlAbstract/FREE Full Text
  22. 22.↵
    1. Liebenstein M,
    2. Bigal M,
    3. Sheftell F,
    4. Tepper S,
    5. Rapoport A,
    6. Lipton R
    . Validation of the Chronic Daily Headache Questionnaire (CDH-Q), abstract F25. 49th Annual Scientific Meeting of the American Headache Society; 2007 June 7–11; Chicago, IL: 760–761.
  23. 23.↵
    1. Lipton RB,
    2. Manack Adams A,
    3. Buse DC,
    4. Fanning KM,
    5. Reed ML
    . A comparison of the Chronic Migraine Epidemiology and Outcomes (CaMEO) Study and American Migraine Prevalence and Prevention (AMPP) study: demographics and headache-related disability. Headache 2016;56:1280–1289.
    OpenUrl
  24. 24.↵
    1. Lipton RB,
    2. Bigal ME,
    3. Ashina S, et al
    . Cutaneous allodynia in the migraine population. Ann Neurol 2008;63:148–158.
    OpenUrlCrossRefPubMed
  25. 25.↵
    1. Buse DC,
    2. Loder EW,
    3. Gorman JA, et al
    . Sex differences in the prevalence, symptoms, and associated features of migraine, probable migraine and other severe headache: results of the American Migraine Prevalence and Prevention (AMPP) Study. Headache 2013;53:1278–1299.
    OpenUrlCrossRefPubMed
  26. 26.↵
    1. Stewart WF,
    2. Lipton RB,
    3. Dowson AJ,
    4. Sawyer J
    . Development and testing of the Migraine Disability Assessment (MIDAS) Questionnaire to assess headache-related disability. Neurology 2001;56:S20–S28.
    OpenUrlAbstract/FREE Full Text
  27. 27.↵
    1. Tietjen GE,
    2. Brandes JL,
    3. Peterlin BL, et al
    . Allodynia in migraine: association with comorbid pain conditions. Headache 2009;49:1333–1344.
    OpenUrlCrossRefPubMed
  28. 28.↵
    1. Chen N,
    2. Zhang J,
    3. Wang P,
    4. Guo J,
    5. Zhou M,
    6. He L
    . Functional alterations of pain processing pathway in migraine patients with cutaneous allodynia. Pain Med 2015;16:1211–1220.
    OpenUrl
  29. 29.↵
    1. Katsarava Z,
    2. Schneeweiss S,
    3. Kurth T, et al
    . Incidence and predictors for chronicity of headache in patients with episodic migraine. Neurology 2004;62:788–790.
    OpenUrlAbstract/FREE Full Text
  30. 30.↵
    1. Bigal ME,
    2. Serrano D,
    3. Buse D,
    4. Scher A,
    5. Stewart WF,
    6. Lipton RB
    . Acute migraine medications and evolution from episodic to chronic migraine: a longitudinal population-based study. Headache 2008;48:1157–1168.
    OpenUrlCrossRefPubMed
  31. 31.↵
    1. Zwart JA,
    2. Dyb G,
    3. Hagen K,
    4. Svebak S,
    5. Stovner LJ,
    6. Holmen J
    . Analgesic overuse among subjects with headache, neck, and low-back pain. Neurology 2004;62:1540–1544.
    OpenUrlAbstract/FREE Full Text
  32. 32.↵
    1. Bigal ME,
    2. Liberman JN,
    3. Lipton RB
    . Obesity and migraine: a population study. Neurology 2006;66:545–550.
    OpenUrlAbstract/FREE Full Text
  33. 33.↵
    1. Gelaye B,
    2. Sacco S,
    3. Brown WJ,
    4. Nitchie HL,
    5. Ornello R,
    6. Peterlin BL
    . Body composition status and the risk of migraine: a meta-analysis. Neurology 2017;88:1795–1804.
    OpenUrlAbstract/FREE Full Text
  34. 34.↵
    1. Lucke T,
    2. Herrera R,
    3. Wacker M, et al
    . Systematic analysis of self-reported comorbidities in large cohort studies: a novel stepwise approach by evaluation of medication. PLoS One 2016;11:e0163408.
    OpenUrl
  35. 35.↵
    1. Probyn K,
    2. Bowers H,
    3. Caldwell F, et al
    . Prognostic factors for chronic headache: a systematic review. Neurology 2017;89:291–301.
    OpenUrlAbstract/FREE Full Text
  36. 36.↵
    1. Lipton RB,
    2. Serrano D,
    3. Pavlovic JM, et al
    . Improving the classification of migraine subtypes: an empirical approach based on factor mixture models in the American Migraine Prevalence and Prevention (AMPP) Study. Headache 2014;54:830–849.
    OpenUrl
  37. 37.↵
    1. Serrano D,
    2. Lipton RB,
    3. Scher AI, et al
    . Fluctuations in episodic and chronic migraine status over the course of 1 year: implications for diagnosis, treatment and clinical trial design. J Headache Pain 2017;18:101.
    OpenUrl
View Abstract

Disputes & Debates: Rapid online correspondence

Comment

NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.

  • Stay timely. Submit only on articles published within the last 8 weeks.
  • Do not be redundant. Read any comments already posted on the article prior to submission.
  • 200 words maximum.
  • 5 references maximum. Reference 1 must be the article on which you are commenting.
  • 5 authors maximum. Exception: replies can include all original authors of the article.
  • Submitted comments are subject to editing and editor review prior to posting.

More guidelines and information on Disputes & Debates

Compose Comment

More information about text formats

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Author Information
NOTE: The first author must also be the corresponding author of the comment.
First or given name, e.g. 'Peter'.
Your last, or family, name, e.g. 'MacMoody'.
Your email address, e.g. higgs-boson@gmail.com
Your role and/or occupation, e.g. 'Orthopedic Surgeon'.
Your organization or institution (if applicable), e.g. 'Royal Free Hospital'.
Publishing Agreement
NOTE: All authors, besides the first/corresponding author, must complete a separate Disputes & Debates Submission Form and provide via email to the editorial office before comments can be posted.
CAPTCHA
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.

Vertical Tabs

You May Also be Interested in

Back to top
Advertisement

Related Articles

  • No related articles found.

Topics Discussed

Alert Me