Back

Evaluating Migraine Trigger Surprisal: Associations with Migraine Activity

Turner, D. P.; Patel, T.; Caplis, E.; Houle, T. T.

2025-04-25 neurology
10.1101/2025.04.23.25325821 medRxiv
Show abstract

BackgroundIdentifying migraine triggers is a common goal for most individuals with migraine but remains challenging due to the vast number of possible trigger candidates and their fluctuating nature. The unexpectedness of daily experiences, quantified through information-theoretic surprisal, may integrate many sources of variation and predict future migraine attacks. ObjectiveThe objective of this study was to evaluate the association between surprisal and future headache attacks. MethodsIn this prospective daily diary study, N = 109 individuals with migraine were enrolled, with N = 104 completing twice-daily electronic entries for up to 28 days, yielding 5,176 total diaries. Diary items captured exposure to potential migraine triggers across behavioral, emotional, and environmental domains. For each entry, total surprisal scores were calculated using within-person empirical probability distributions to reflect how atypical each days experiences were. ResultsParticipants experienced a headache on 1,518/5,145 (29.5%) of days with complete diary information. Higher surprisal significantly predicted increased migraine risk within 12 hours (OR = 1.86 [95%CI: 1.12-3.08], p = 0.016) and 24 hours (OR = 2.15 [95%CI: 1.44-3.20], p < 0.001), with stronger effects observed at the longer interval. Notably, the association between current surprisal and migraine onset was moderated by recent surprisal history and exhibited nonlinear properties at 12 hours. Random effects revealed substantial between-person variability in surprisal sensitivity, and individuals with higher baseline headache risk showed attenuated associations. ConclusionsSurprisal offers a novel, individualized measure of trigger unpredictability that is associated with short-term migraine risk. Incorporating surprisal into digital tools may improve personalized prediction and prevention strategies, moving beyond static trigger lists to a dynamic, context-aware model of migraine self-management.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.