Ischemic Stroke in Nonvalvular Atrial Fibrillation at Warfarin Initiation: Assessment via a Large Insurance Database.

Abstract

Stroke risk may increase shortly after warfarin initiation in nonvalvular atrial fibrillation patients. Because of the brief period and limited number of events, large samples are needed to study this effect. We compared 30-day rates of ischemic stroke between nonvalvular atrial fibrillation patients initiating warfarin to nonwarfarin comparators using an insurance claims database. We identified nonvalvular atrial fibrillation patients via 1 inpatient/2 outpatient diagnosis claims from the MarketScan database, January 1, 2009, to December 31, 2010. We studied patients initiating warfarin using prescription claims and 1:1 matched 22 669 initiators to comparators based on age, sex, diagnosis date, and warfarin propensity score. Follow-up began on initiation date; patients were censored at discontinuation/switch of therapy, disenrollment, or end of the study. The median follow-up was 415 days. Cox regression was used to study differences in ischemic stroke risks between warfarin initiators and comparators while controlling for important prognostic factors. Warfarin initiators were generally similar to comparators in clinical features but had higher CHADS₂ scores (1.26 versus 1.19). The first 30-day ischemic stroke rate was higher among warfarin initiators than comparators (1.47%/y (27/1836) versus 0.98%/y (18/1837); <i>P</i>=0.18) but lower subsequently (0.81%/y [134/16 543] versus 1.09%/y [406/37 248]; <i>P</i>=0.002). Multivariable regression confirmed a significant interaction between follow-up and warfarin use with the adjusted hazard ratios (95% confidence intervals) for warfarin/comparator as 1.46 (0.80-2.65) in the first 30 days and 0.70 (0.57-0.85) afterward. Warfarin effect was qualitatively different in the first 30 days after initiation than subsequently. This is consistent with a modest increase in stroke risk occurring briefly after starting warfarin.