Stroke Risk with Warfarin Interruption

Abstract & Commentary

. Professor of Medicine, Chief of Clinical Cardiology, University of California, San Francisco.

This article originally appeared in the October 2012 issue of Clinical Cardiology Alert. It was peer reviewed by Ethan Weiss, MD, Assistant Professor of Medicine, Division of Cardiology and CVRI, University of California, San Francisco. Dr. Crawford reports no financial relationships relevant to this field of study, and Dr. Weiss is a scientific advisory board member for Bionovo.

Sources: Raunso J, et al. Increased short-term risk of thrombo-embolism or death after interruption of warfarin treatment in patients with atrial fibrillation. Eur Heart J 2012;33:1886-1892. Hohnloser S, et al. The hazards of interrupting anticoagulation therapy in atrial fibrillation. Eur Heart J 2012; 33:1864-1866.

The risk of interrupting prophylactic warfarin for stroke prevention in atrial fibrillation (AF) patients is unclear. Thus, these investigators from Denmark evaluated their national health registry and found 102,591 patients > age 30 with a first-time hospitalization for AF between 1997 and 2008. Valvular AF patients were excluded. Follow-up was started 7 days after hospitalization to ensure achievement of steady-state warfarin dosing. Warfarin therapy was subsequently determined by their national pharmacy database and warfarin usage was estimated based on the supply of medications dispensed. The primary outcome was the combined endpoint of all-cause mortality or hospitalization for thromboembolism. The mean follow-up was 3.5 years. During warfarin therapy, the primary endpoint occurred in 6.9/100 patient-years. At least one treatment interruption occurred in 72% of the patients and these patients had lower CHADS2 scores compared to the no interruption group (1.34 vs 1.56, P < 0.001). The median duration of interrupted therapy was 36 days. Among the 16,738 primary events, 49% occurred during the treatment interruption, for a rate of 14.2/100 patient-years. More events occurred during the first 90 days of interruption (31.6/100 patient-years) and leveled off after 180 days. The hazard ratio for treatment interruption was 2.9 (95% CI 2.8-3.0). Also, the hazard ratio was similar if death was excluded as an endpoint. The authors concluded that interruption of warfarin therapy in non-valvular AF patients increased the short-term risk of death or thromboembolism, especially during the first 90 days of treatment interruption.


Because of their national health systems, many European countries have very large patient databases that dwarf those at some of our single hospital systems or even multicenter trial databases. Although limited by their retrospective observational nature and the unique structure of national databases, their sheer size makes these analyses important. This study from Denmark has two interesting findings. First, the incidence of warfarin therapy interruption in AF patients is high, about three-quarters of patients, and the median length is relatively long, 36 days. Randomized AF therapy trials have noted interruptions at a frequency of 15-30%. Clearly, real-world experience is very different from trials. Second, the risk of thromboembolism or death rises three-fold in the first 90 days of interrupted therapy, then tapers off. Thus, real world patients on warfarin stroke prophylaxis for AF are often at considerable risk because of therapy interruptions.

Interestingly, there were no subgroup differences in the incidence of the primary event whether stratified by age, sex, duration of therapy, or CHADS2 score. Also, excluding death as an endpoint did not appreciably alter the results, suggesting that the important events were thromboembolism. This raises the question of the etiology of the increase in events. One possibility is that warfarin is preventing strokes and after its withdrawal the stroke rate returns to its natural state (the so-called catch-up phenomenon). Another is that there is a warfarin withdrawal phenomenon that actually increases the rate of thromboembolism over what it would naturally be. Although there are some experimental data showing that coagulation factors transiently rise above normal levels after warfarin withdrawal, there are no clinical mechanistic data to support this theory. Finally, it is possible that whatever occasioned the interruption in therapy was the cause of the event, such as surgery (confounder). Unfortunately, this study is not able to sort out these potential mechanisms.

Other studies have shown that warfarin can be interrupted for clinical reasons such as surgery or due to patient decisions unrelated to their health conditions. This study does not specify the reasons and we have no clinical data such as INR values. Also, we do not know if the interruptions were transient or the patients stopped therapy. Although the median duration of interruption was 36 days, the 75th percentile was 207 days. Specific guidelines exist for medical issues such as high INR values, episodes of major bleeding, and surgery, which should minimize the risks of thromboembolism, so one could conclude that the majority of the interruptions that lead to events were in the patient decision category. If so, it behooves us to emphasize the importance of continuous therapy to our patients. Perhaps the use of the newer oral anticoagulants, which do not require INR-based management, will improve patient compliance.