Migraine Aura mimicking Stroke in Code Stroke. Analysis of diagnostic prediction models

Background: Differences between migraine with aura (MA) mimicking stroke and ischemic strokes were analyzed to create a predictive diagnostic model.Methods: Prospective cohort of code strokes attended between January 2005 and June 2020 in a tertiary hospital evaluated by a vascular neurologist. Data recorded included: Place of activation, demographics, vascular risk factors, vitals, initial blood test (glucose, blood count, electrolytes, coagulation times, D-dimer, fibrinogen and C-reactive protein), initial NIHSS, door-to-onset time, and door-to-needle time. After the vascular study, diagnosis of ischemic stroke or MA was registered. Multivariate logistic regression analyses were performed to create a predictive model. Discrimination and calibration performance were assessed. A secondary aim was to describe detailed clinical and radiological data of MA cases in order to identify new predictive factors that could be used in future studies.Results: Of 3140 code strokes attended, 2424 (77.2%) were ischemic strokes and 34 (1.1%) were MA. Migraine cases were younger, more frequently females and with low prevalence of vascular risk factors. Initial NIHSS was lower in MA cases, but no differences were seen in fibrinolysis rate (30%). Blood test showed lower levels of glucose, D-dimer, and fibrinogen in MA cases. Multivariate model showed and independent association for ischemic stroke with age [OR, (95%CI):1.09, (1.07-1.12, p65, NIHSS>6, and fibrinogen>400mg/dL) showed a good global discrimination capacity (AUC = 0.897) but a good calibration only for extreme predicted probabilities (>98% in ischemic strokes or < 10% in MA). We did not find any differential clinical or radiological factor that could be exclusive of MA.Conclusions: A predictive model including age >65, men sex, NIHSS >6 and fibrinogen >400 mg/dL showed a good discrimination, but low calibration performance to distinguish between MA and ischemic stroke. Future studies should involve migraine biomarkers in order to improve clinical decision making.