Statistical Testing of Association Between Menstruation and Migraine

ObjectiveTo repair and refine a previously proposed method for statistical analysis of association between migraine and menstruation. BackgroundMenstrually related migraine (MRM) affects about 20% of female migraineurs in the general population. The exact pathophysiological link from menstruation to migraine is hypothesized to be through fluctuations in female reproductive hormones, but the exact mechanisms remain unknown. Therefore, the main diagnostic criterion today is concurrency of migraine attacks with menstruation. Methods aiming to exclude spurious associations are wanted, so that further research into these mechanisms can be performed on a population with a true association. MethodsThe statistical method is based on a simple two-parameter null model of MRM (which allows for simulation modeling), and Fisher's exact test (with mid-p correction) applied to standard 2x2 contingency tables derived from the patients' headache diaries. Our method is a corrected version of a previously published flawed framework. To our best knowledge, no other published methods for establishing a menstruation-migraine association by statistical means exist today. ResultsThe probabilistic methodology shows good performance when subjected to receiver operator characteristic curve analysis. Quick reference cutoff values for the clinical setting were tabulated for assessing association given a patient's headache history. ConclusionsIn this paper, we correct a proposed method for establishing association between menstruation and migraine by statistical methods. We conclude that the proposed standard of 3-cycle observations prior to setting an MRM diagnosis should be extended with at least one perimenstrual window to obtain sufficient information for statistical processing.