Stroke-free duration and stroke risk in patients with atrial fibrillation: simulation using a Bayesian inference

ABSTRACT

Background: Prior stroke or transient ischemic attack (TIA) is a major predictor of stroke in patients with atrial fibrillation (AF). It may be regarded as a Bayesian predictor by which some underlying causes can be inferred from the result. Similarly, stroke-free duration may be a predictor of freedom from stroke, but this subject has remained unaddressed. Objective: To evaluate the above hypothesis by simulating how stroke-free duration can affect estimated stroke risk in patients with AF using a Bayesian inference. Methods: The development of stroke was modeled as a yearly Bernoulli trial. A Bayesian inference method was applied to the model with correction for the risk increase by aging. The model was composed of two risk parameters prior stroke (or TIA) and aging. The parameters were treated as random variables. Their distributions were determined to agree with the clinical data reported in the meta-analysis by Stroke Risk in Atrial Fibrillation Working Group. In the simulation, we noted the stroke risk after 1,..,10 stroke-free years for initial risk 1,..,8 %/year. Results: Increasing stroke-free duration reduced the estimated stroke risk, offsetting the risk increase by aging. In patients whose stroke (including TIA) risk was 3%/year or less, the estimated risk remained almost constant with aging if they were stroke-free. In patients whose risk was 4%/year or higher, the risk decreased if they were stroke-free. Conclusion: Our analysis predicted that stroke-free duration in patients with AF is an important predictor of freedom from stroke. Increasing stroke-free duration will reduce the estimated stroke-risk, which may influence our decision to start anticoagulation. Stroke-free duration should be considered when estimating stroke risk in patients with AF.