Nonvalvular atrial fibrillation is a common cardiac arrhythmia that strongly increases a person’s risk of ischemic stroke.1 Warfarin (Coumadin) therapy can dramatically reduce the risk of stroke associated with atrial fibrillation, but it increases the risk of life-threatening bleeding complications.2 Intracranial hemorrhage is a particularly devastating complication of warfarin therapy. Despite their high associated morbidity, intracranial hemorrhages are quite uncommon, with a rate of about 0.4% per year in patients receiving warfarin therapy.2 In contrast, the risk of ischemic stroke in patients with atrial fibrillation who are not taking warfarin ranges between 1% and 12% per year, depending on clinical risk factors, such as previous stroke and older age.2,3
Despite the well-documented benefits of warfarin, many physicians are hesitant to prescribe it for elderly individuals or tend to use a lower target international normalized ratio (INR) intensity, presumably to reduce the risk of life-threatening hemorrhages.4 In fact, recent guidelines have suggested a target INR range of 1.6 to 2.5 for the primary prevention of stroke in some atrial fibrillation patients over the age of 75 years.5 It is unclear, however, whether this strategy reduces the risk of hemorrhage. To address these issues, we designed a case-control study to examine the association between age, anticoagulation intensity, and the risk of intracranial hemorrhage in patients taking warfarin for nonvalvular atrial fibrillation.
We searched the Massachusetts General Hospital billing system for case patients admitted with intracranial hemorrhage between the years 1993 and 2002. Case patients were eligible if they were taking warfarin for atrial fibrillation and had no other explanation for intracranial hemorrhage (such as severe head trauma or cerebral aneurysm). We then matched each case to six anticoagulated control patients without intracranial hemorrhage, randomly selecting controls from the hospital’s large anticoagulation clinic. We reviewed medical records for clinical information, including patient age and INR at the time of hemorrhage. Multivariable conditional logistic regression was used to assess the relative odds of intracranial hemorrhage at varying ages and INR intensities, controlling for potential clinical confounders (such as cerebrovascular disease and hypertension) as well as aspirin use.
We identified 170 case patients who developed intracranial hemorrhage while taking warfarin for atrial fibrillation and matched them to 1,020 control patients from the anticoagulation clinic. INR data were available for 145 case patients and for all control patients. Only patients with INR data were used in the multivariable analysis.
There were 79 primary intracerebral hemorrhages, 71 subdural hemorrhages, and 20 hemorrhages of other types. A history of minor head trauma was more common in those with subdural compared with in-
tracerebral hemorrhage (48% versus 8.1%, P < .001). The inpatient mortality rate was quite high, particularly among those with primary intracerebral hemorrhages (43%). Although 91% of patients were living at home before the hemorrhage event, only 14% were able to be discharged directly home after the hospitalization.
Comparing cases with controls on bivariate analysis, case patients were older and had higher median INRs (Table). Aspirin use was similar between cases and controls (20% versus 19%). In multivariable models, we included terms for age, INR, sex, race, and the following clinical variables:
a history of cerebrovascular disease (prior ischemic stroke or carotid ar-tery disease), hypertension, congestive heart failure, coronary artery disease, diabetes mellitus, cancer, and aspirin use.
After multivariable adjustment, we found that the risk of intracranial hemorrhage rose significantly at an age of 85 years or older compared with a reference age of 70 to 74 years, with an adjusted odds ratio (OR) of 2.5 and 95% confidence interval (CI) of 1.3—4.7; (Figure 1). After adjusting for clinical risk factors, including age, we found that the risk of intracranial hemorrhage rose significantly at an INR range of 3.5 to 3.9 (adjusted OR, 4.6 [CI, 2.3–9.4]) and rose markedly at INRs of 4.0 and higher (adjusted OR, 8.8 [CI, 4.6–17.0]; Figure 2), compared with a reference INR range of 2.0 to 3.0. Lower INRs (< 2.0) were not associated with a lower risk of intracranial hemorrhage compared with INRs of 2.0 to 3.0 (adjusted OR, 1.3 [CI, 0.8–2.2]). A history of cerebrovascular disease was also significantly associated with an increased risk of intracranial hemorrhage (adjusted OR, 2.2 [CI, 1.4–3.4]).
This is the largest study to date to examine risk factors for intracranial hemorrhage in patients taking warfarin for atrial fibrillation. We found that the risk of intracranial hem-orrhage rose significantly at age 85 years or older and at INRs of 3.5
and higher. However, INRs below 2.0 did not offer additional protection against intracranial hemorrhage compared with INRs in the 2.0 to 3.0 range. Thus, our study does not support the use of lower INR target ranges for patients with nonvalvular atrial fibrillation, even in elderly patients. Previous studies have shown that the risk of ischemic stroke in
atrial fibrillation rises significantly
at INRs of less than 2.03,6; lower
INR targets will place patients at higher risk for ischemic stroke while not reducing the risk of intracranial hemorrhage.
Although older age appears to be a risk factor for intracranial hemorrhage, it is also a risk factor for atrial fibrillation—associated ischemic stroke.7 In general, because the incidence rate of ischemic stroke in those not receiving warfarin far exceeds the rate of intracranial hemorrhage in those taking warfarin,8 the net benefits of warfar-in continue to outweigh the risks in elderly patients with atrial fibrilla-tion. Care must be taken to maintain the INR intensity within a therapeutic range, however, to maximize the benefits while minimizing the risks. Although increasing age and INR are associated with a greater risk of this complication, the majority (62%) of intracranial hemorrhages in our study occurred at INRs of below 3.0. Similar to other studies, we found that a history of cerebrovascular disease (such as previous stroke) was a risk factor for future intracranial hemorrhage.9 However, although aspirin use has been linked to a higher risk of hemorrhagic stroke in previous studies,10 we did not find such an association in our study.
Aside from older age, previous stroke, and possibly hypertension, there are few validated clinical predictors of intracranial hemorrhage.9,11 Consequently, individualized risk assessment often represents guesswork. Because of the increasing number of elderly patients taking warfarin for atrial fibrillation,12 their increased risk of warfarin-associated intracranial hemorrhage necessitates better predictors of this condition. Identification of additional factors, including through biochemical or radiographic means,13,14 would make individualized decisions about anticoagulation in atrial fibrillation much more rational and effective.
Our study has several limitations. Because many case patients were not followed in the anticoagulation clinic, differences in outpatient anticoagulation management and patient characteristics may have biased some of the risk estimates. To address this, we performed a nested case-control study, which showed results similar to the full case analysis. We did not have data on duration of warfarin use, a factor that has been linked to hemorrhage in previous studies. However, restricting the analyses to patients taking warfarin for at least 1 year did not substantially change our results. Finally, our study was not designed to assess the risk of intracranial hemorrhage in patients not taking warfarin, because both case and control patients were already on this medication.
Intracranial hemorrhage is a rare but devastating complication of warfarin therapy. Our study shows that an age of 85 years or older and an INR of 3.5 or higher are associated with an increased risk of intracranial hemorrhage. However, subtherapeutic INRs (< 2.0) do not offer additional protection from hemorrhage compared with the standard therapeutic INR range. Because the risk of ischemic stroke rises with increasing age and INRs below 2.0, our study supports using an INR range of 2.0 to 3.0 (target 2.5) for older as well as younger patients with nonvalvular atrial fibrillation.