Identifying Natural Subgroups of Migraine Based on Comorbidity and Concomitant Condition Profiles: Results of the Chronic Migraine Epidemiology and Outcomes (CaMEO) Study

Objective: To identify natural subgroups of people with migraine based on profiles of comorbidities and concomitant conditions, hereafter referred to as comorbidities. Background: Migraine is a heterogeneous disease. Identifying natural subgroups (endophenotypes) may facilitate biological and genetic characterization and the development of personalized treatment. Methods: The Chronic Migraine Epidemiology and Outcomes Study is a prospective web-based survey study designed to characterize the course of migraine and related comorbidities in a systematic US sample of people with migraine. Respondents were asked if they ever had a specific comorbidity and, if present, whether the comorbidity was confirmed/diagnosed by a “doctor”; 62 comorbidities were available for analysis. Latent class analysis (LCA) modeling determined the optimal number of classes and a parsimonious set of comorbidities. Results: Of the 12,810 respondents with migraine, 11,837 reported ≥1 comorbidity and were included in this analysis. After statistical analysis and clinical judgment reduced the number of comorbidities, we selected an 8-class model based on 22 comorbidities. Each class had a distinct pattern summarized as follows: Class 1, Most Comorbidities; Class 2, Respiratory/Psychiatric; Class 3, Respiratory/Pain; Class 4, Respiratory; Class 5, Psychiatric; Class 6, Cardiovascular; Class 7, Pain; Class 8, Fewest Comorbidities. The distribution of individuals across models was variable, with one-third of respondents in Class 8 (Fewest Comorbidities) and <10% in Class 1 (Most Comorbidities). Demographic and headache characteristics, not used in assigning class membership, varied across classes. For example, comparing Class 1 (Most Comorbidities) and Class 8 (Fewest Comorbidities), Class 1 had a greater proportion of individuals with severe disability (Migraine Disability Assessment grade IV; 48.1% vs 22.3% of overall individuals) and higher rates of allodynia (67.6% vs 47.0%), medication overuse (36.4% vs 15.0%), chronic migraine (23.1% vs 9.1%), and aura (40.1% vs 28.8%). Conclusions: LCA modeling identified 8 natural subgroups of persons with migraine based on comorbidity profiles. These classes show differences in demographic and headache features not used to form the classes. Subsequent research will assess prognostic and biologic differences among the classes.