Tracking atrial fibrillation burden trends over time may enhance stroke prediction accuracy, with one model achieving a 0.726 area under the receiver operating characteristic curve, which outperformed traditional risk scores, according to a recent study. 

Researchers evaluated the predictive value of atrial fibrillation (AF) burden trends for ischemic stroke risk stratification, suggesting potential improvements over conventional methods. The retrospective cohort study included 5,013 patients (mean age 69.2 years, 50% male) with insertable cardiac monitors (ICMs) used for AF management, suspected AF, or cryptogenic stroke, with data from over 2.4 million cumulative days of follow-up. During this period, 869 patients experienced an ischemic stroke.

The study, published in Circulation: Arrhythmia and Electrophysiology, applied machine learning models to assess temporal AF burden trends, which showed incremental prognostic value for stroke risk prediction. Among predictive factors, prior stroke or transient ischemic attack (TIA) was identified as the strongest predictor (variable importance score 13.13), followed by the absence of prior AF (7.35) and AF burden trends during follow-up (2.59). Temporal AF burden trends, such as the 21-day simple moving average offset against cumulative averages, demonstrated significant predictive power, especially in cases of cryptogenic stroke (94% accuracy) and AF management (11%).

The model that combined baseline clinical characteristics with AF burden trends achieved an area under the receiver operating characteristic curve of 0.726, exceeding the predictive accuracy of the CHA2DS2-VASc score and daily AF burden alone. The study found that the optimal temporal window for AF burden in predicting stroke risk varied by clinical indication: AF management prioritized shorter windows (1-5 days), while cryptogenic stroke required a longer assessment period (21 days). The findings suggest that incorporating AF burden trends alongside standard clinical data may enhance stroke risk stratification and support targeted anticoagulation therapy. Prospective trials are recommended to validate these findings.

Full disclosures can be found in the published study.