Oral anticoagulants in non-valvular atrial fibrillation for primary stroke prevention (no prior stroke)

25 for prevented stroke

Benefits in NNT

1 in 25 were helped (preventing stroke*)
1 in 42 were helped (preventing death from any cause)
96% saw no benefit
4% were helped by preventing 1 stroke*
2.4% were helped by avoiding death

Harms in NNT

1 in 25 were harmed (having bleeding)
1 in 384 were harmed (intracranial hemorrhage)
4% were harmed by having bleeding
0.26% were harmed by intracranial hemorrhage
View As:

Efficacy Endpoints

All ischemic strokes, All cause mortality (within 30 days from stroke)

Harm Endpoints

All intracranial hemorrhages, Major extracranial hemorrhage


Non-valvular atrial fibrillation (AF) is the most common cause of cardio-embolic stroke and a powerful risk factor for strokes. The overall risk ranges from 2.5% to 4% per year. Oral anticoagulants (OAC) reduce this risk, but also increase the risk of hemorrhagic stroke and major bleeding.

The review compared warfarin with placebo for prevention of stroke in patients with atrial fibrillation but no history of strokes (or transient ischemic attacks). Five high quality trials with 2313 participant were included, most examining warfarin versus placebo. The warfarin group achieved a mean INR between 2.0 and 2.6. The overall mean age was 69, 74% were men, and mean follow up was 1.5 years per participant. Warfarin use as compared to placebo was associated with significant decreased risk in ischemic strokes (4.05% absolute risk reduction [ARR]). It also reduced the risk of combined ischemic and hemorrhagic strokes (3.79% ARR) and all cause mortality (2.38% ARR). However, warfarin increased the risk of intracranial hemorrhage (0.26%), and slightly increased the risk of major extracranial hemorrhage (0.093%). Overall, warfarin offered advantages over placebo during the 1.5 year study period.

The rate of harms is often inconsistent in randomized trials designed to track benefits, and in the studies included in this review the subjects were carefully selected based on a low risk of bleeding. Moreover, the total number of subjects was small, leaving the studies under-powered for bleeding outcomes. Therefore we have used data from a systematic review of observational studies that suggests the true bleeding rate for patients taking warfarin to be just under 3% for the first 6 months of use, with a 0.6% fatal bleeding rate.[Dahri, Newman] This risk is heavily weighted toward the first 6 months and tapers after, thus we have estimated a total bleeding rate of 4% per year, and 0.6% fatal. After the first year the risks are less well known and poorly documented, therefore we have reported only the first year of hemorrhage rates, and leave all further estimating to physician judgment, as data suggest that this is as accurate as any method for predicting hemorrhage.1


Follow up was limited to about 1.5 years and it is unclear whether the protective effect of oral anticoagulants extends beyond this period, or if the bleeding rate continues to climb. This may be further confounded by the relatively young age of participants in these trials, as it is unclear if the results would be the same for the older population. There was some heterogeneity of the results in respect to efficacy and safety profiles for different age groups. Additionally, it is possible that the low rates of major bleeding were, at least partially, the result of careful patient selection and INR monitoring of a drug with a narrow therapeutic window, which is not always possible in general practice. Finally, clinicians and researchers were not blinded in any of the studies due to difficulty in masking oral anticoagulation.

*The term ‘stroke’ refers to ischemic strokes, those caused by blockage of a blood vessel, while we use the term ‘intracranial hemorrhage’ to indicate bleeding in the brain. The latter condition is often called a stroke as well, however we prefer to separate the two for a more nuanced understanding of harms and benefits.


Kirill Shishlov, MD MPH


October 21, 2013