Predicting Atrial Fibrillation after Ischemic Stroke: Clinical, Genetics, and Electrocardiogram Modelling.

Introduction: Detection of atrial fibrillation (AF) is challenging in patients after ischaemic stroke due to its paroxysmal nature. We aimed to determine the utility of a combined clinical, electrocardiographic, and genetic variable model to predict AF in a post-stroke population.

Materials and methods: We performed a cohort study at a single comprehensive stroke centre from November 09, 2009, to October 31, 2017. All patients recruited were diagnosed with acute ischaemic stroke or transient ischaemic attacks. Electrocardiographic variables including p-wave terminal force (PWTF), corrected QT interval (QTc), and genetic variables including single nucleotide polymorphisms (SNPs) at the 4q25 (rs2200733) were evaluated. Clinical, electrocardiographic and genetic variables of patients without AF and those who developed AF were compared. Multiple logistic regression analysis and receiver operating characteristics were performed to identify parameters and determine their ability to predict the occurrence of AF.

Results: Out of 709 patients (median age of 59 years, inter-quartile range 52-67) recruited, sixty (8.5%) were found to develop AF on follow-up. Age (odds ratio [OR]): 3.49, 95% confidence interval [CI]: 2.03-5.98, p < 0.0001), hypertension (OR: 2.76, 95% CI: 1.36-5.63, p = 0.0052), and valvular heart disease (OR: 8.49, 95% CI: 2.62-27.6, p < 0.004) were the strongest predictors of AF, with an area under receiver operating value of 0.76 (95% CI: 0.70-0.82), and 0.82 (95% CI: 0.77-0.87) when electrocardiographic variables (PWTF and QTc) were added. SNP did not improve prediction modelling.

Conclusion: We demonstrated that a model combining clinical and electrocardiographic variables provided robust prediction of AF in our post-stroke population. Role of SNP in prediction of AF was limited.