Home > Neurology > IHC 2021 > Cardioembolic Comorbidities > AI-enabled ECG algorithm predicts atrial fibrillation risk in migraine

AI-enabled ECG algorithm predicts atrial fibrillation risk in migraine

Presented by
Dr Nikita Chhabra, Mayo Clinic, NY, USA
Conference
IHC 2021

For patients without a confirmed diagnosis of atrial fibrillation (AF), the probability of AF, as predicted by an artificial intelligence (AI)-enabled ECG algorithm, is significantly higher in patients with migraine with aura than in those without aura, even after adjusting for age and sex [1].

Migraine with aura is associated with an approximately 2-fold increased risk of ischaemic stroke [2-4]. “However, the mechanism behind this relationship is largely unknown,” said Dr Nikita Chhabra (Mayo Clinic, NY, USA). Migraine with aura is associated with an increased risk of cardioembolic stroke when compared with migraine without aura, but not with risk of lacunar or non-lacunar thrombotic stroke [5]. Interestingly, higher incidence of AF has been demonstrated in patients with migraine with aura than in those without aura in longitudinal cohort studies [6].

These findings may imply that AF-associated cardioembolism plays a key role in the interaction between migraine and stroke. However, investigating this hypothesis has many challenges. Firstly, AF is difficult to detect given its paroxysmal nature. Furthermore, long-term cardiac monitoring is expensive, time-consuming, low-yield, and often not justified in patients with few vascular risk factors [7]. Prior to developing clinical AF, structural changes occur in the atria that predispose patients to future risk of atrial arrhythmias [8]. These changes might be reflected on ECGs, but could be too subtle to be detected by human eyes.

The cardiology team at Mayo Clinic developed a new way of predicting subclinical AF: an AI-enabled ECG algorithm that can predict the probability of paroxysmal AF based on a single sinus rhythm ECG [9]. In ECGs that were interpreted as normal sinus rhythm by cardiologists, this algorithm was able to identify and predict the probability of paroxysmal AF with a sensitivity of 79.0% and a specificity of 79.4%. This is relevant because few patients with migraine have undergone long-term cardiac monitoring, but many have had routine ECGs at some point during their treatment course for migraine or other medical conditions. The AI-enabled ECG algorithm could be a powerful tool that is readily available to help elucidate the association between migraine, AF, and stroke.

After excluding patients with a previous confirmed AF diagnosis, 676 migraine patients with aura and 1,124 patients without aura were analysed in the current study [1]. The migraine with aura group was significantly older than the migraine without aura group (50.2 vs 46.6 years; P<0.001). After adjustment for age and sex, patients with aura were found to have a higher mean probability of AF than those without aura (7.6% vs 5.9%; P=0.003). Interestingly, the difference of AF probability between migraine patients with and without aura was significant in men (P=0.043), but not in women (P=0.079). Since autonomic dysfunction has a role in the pathophysiology of both migraine and AF, it begs the possibility that migraine with aura development is attributable to cardioembolic stroke arising from AF.

These results are consistent with results observed from longitudinal cohorts and support the theory that AF-mediated cardioembolism plays a key role in the association between migraine and stroke, especially in patients with migraine with aura.

  1. Chhabra N, et al. An AI-enabled ECG Algorithm Predicts Higher Subclinical Atrial Fibrillation Risk in Patients with Migraine with Aura Compared to Migraine without Aura. AL02, IHC 2021, 8–12 September.
  2. Øie LR, et al. J Neurol Neurosurg Psychiatry. 2020;91(6):593–604.
  3. Spector JT, et al. Am J Med. 2010;123(7):612–24.
  4. Etminan M, et al. BMJ. 2005;330(7482):63.
  5. Androulakis XM, et al. Neurology. 2016;87(24):2527–32.
  6. Sen S, et al. Neurology. 2018;91(24):e2202–10.
  7. Seet RCS, et al. Circulation. 2011;124(4):477–86.
  8. Kottkamp H. Eur Heart J. 2013;34(35):2731–8.
  9. Attia ZI, et al. Lancet. 2019;394(10201):861–7.

Copyright ©2021 Medicom Medical Publishers



Posted on