To improve the early detection of persons with headache at increased risk for incident chronic migraine (CM).

Detection of persons with headache at risk for developing CM is essential to the development of preventive interventions.  Latent transition models provide a promising approach for identifying groups at high risk for CM (Collins & Lanza, 2009; Ip, 2013).

Using longitudinal data from the 2005 and 2007 American Migraine Prevalence and Prevention Study surveys, we fit latent transition models to identify natural subgroups of people with headache. Headache-related domains upon which transitions were modeled included migraine-defining symptoms, allodynia (Allodynia Symptom Checklist, ASC-12) and headache frequency plus disability (Migraine Disability Assessment Scale). The model assigned subjects to transition classes, reflecting changes in these domains occurring between 2005 and 2007.  Models were estimated in R version 3.5.1.  We assessed the incidence of CM in 2009 for each latent transition class.

In the 3948 subjects contributing data in 2005, 2007, and 2009, 404 (10%) populated the estimated transition classes believed to be at risk for disease progression. The 2009 incident CM rate among subjects in these classes – 16.1% - was greater than the total-sample 2009 incident CM rate of 2.6%.

Latent transition classes predicted the onset of CM in persons with headache. The 2-year early detection CM incidence rate was greater than the previously reported annual rate and cumulative quarterly rate within a year. This suggests that more subjects may be at risk for incident CM than is currently expected, and that these subjects can be detected. Additional validation should be done on an independent sample to confirm that particular latent transition classes can indeed predict incident migraine.