Intensity of Anticoagulation in Atrial Fibrillation

Abstract & Commentary

Synopsis: Anticoagulation that results in an INR of 2.0 or more reduces both the frequency and severity of stroke among patients with nonvalvular atrial fibrillation. Lower intensity anticoagulation was significantly less effective.

Source: Hylek EM, et al. N Engl J Med. 2003;349:1019-1026.

Hylek and colleagues report on the effects of varying intensities of oral anticoagulation on the frequency and severity of strokes among patients with atrial fibrillation. The data were obtained from a longitudinal database of adult patients with nonvalvular atrial fibrillation enrolled in the Kaiser Permanente of Northern California health care system. The cohort included 13,559 patients. Patients who suffered an ischemic stroke were identified by review of hospitalization and billing claims data. The use of warfarin and aspirin at the time of stroke was determined by a review of emergency room data or hospital admission notes. The INR value was recorded at presentation or, if admission data were not available, from a recent clinic visit. Prior anticoagulation data were obtained from pharmacy and laboratory records. Stroke was classified using a modified Rankin scale. Mortality data (30 day) were obtained from health plan records. The independent effect of antithrombotic therapy on 30-day mortality was assessed using a Cox proportional hazard model.

During an 18-month period, 618 patients with atrial fibrillation and ischemic stroke were identified, but 22 were excluded from analysis because of incomplete data or because they had an intracerebral hemorrhage due to thrombolytic or heparin therapy after their initial presentation.

Of the remaining 596 patients, 188 (32%) were taking warfarin, 160 (27%) were taking aspirin, and 248 (42%) were taking neither drug. Among those on warfarin, the median INR was 1.7, and 62% of the values were less than 2.0. Historical data from these patients obtained prior to their presentation showed a median INR value of 2.2.

Stroke severity was strongly correlated with 30-day mortality. In turn, anticoagulation intensity had a significant influence on the severity of stroke. Among patients taking warfarin, 15% of those with an INR below 2.0 either died or were discharged with a severe stroke as compared with 5% of those with an INR of 2.0 or greater. Patients with an INR of less than 1.5 had an outcome similar to those with an INR of 1.5-1.9. Patients who were not taking anticoagulants had worse outcomes, with 22% either dead or discharged with a severe stroke. Among patients taking aspirin, 13% either died or were discharged with a severe stroke. After adjustment for baseline variables, the medication group remained a significant predictor of outcome. Compared to patients with an INR of 2.0 or greater, patients not taking any anticoagulant had a relative hazard for death of 4.9. Patients taking aspirin had a hazard ratio of 2.5. Patients with an INR below 2.0 had a hazard ratio of 3.4.

Incidence rates of ischemic stroke in the entire cohort were also calculated for patients taking warfarin. If the INR was < 1.5, the rate was 7.7 per 100 person-years. For INR values between 1.5 and 1.9, the stroke rate was 1.9 per 100 person-years. For values from 2.0 to 3.9, stroke rates were below 0.9 per 100 person-years. Higher stroke and intracerebral hemorrhage rates were noted when INR values were 4.0 or greater.

Hylek et al conclude that anticoagulation that results in an INR of 2.0 or more reduces both the frequency and severity of stroke among patients with nonvalvular atrial fibrillation. Lower intensity anticoagulation was significantly less effective.

Comment by John P. DiMarco, MD, PhD

This paper provides important observational data concerning anticoagulation in patients with atrial fibrillation. Current guidelines recommend maintaining an INR between 2.0 and 3.0 in patients with atrial fibrillation and one or more risk factors for stroke, but it suggests use of a lower INR target (1.5-1.9) in patients older than 75 years of age because of an increased risk for bleeding.1 The data in this paper do not support this suggestion. This lower- target INR will provide some protection against stroke but does not mitigate the severity of stroke should one occur.

Management of warfarin anticoagulation is challenging. Up to one-third of patients screened may have strong contraindications to anticoagulation at presentation. Bleeding, both major and minor, is frequent during long-term anticoagulation. Even in clinical trials, only 60-65% of INR values are within the therapeutic range. Although the concept of using a lower target INR to reduce bleeding risk might seem attractive, this paper points out significant limitations to this approach. The safety margin for patients with a lowered target is too small for this approach to be used unless bleeding during therapy necessitates it.

Unfortunately, this paper does not include information on why each anticoagulation approach was chosen for individual patients. Many of those taking either no anticoagulant or aspirin might have had firm contraindications to warfarin therapy. This paper highlights the difficult risk:benefit analysis required of clinicians managing patients with atrial fibrillation. Clearly, if anticoagulation is prescribed, efforts to ensure adequate INR values must be taken.

Dr. DiMarkco, Professor of Medicine, Division of Cardiology, University of Virginia, Charlottesville, is on the Editorial Board of Clinical Cardiology Alert.


1. Fuster V, et al. Circulation. 2001;104:2118-2150.