We evaluated a multiethnic cohort of subjects with nonrheumatic atrial fibrillation hospitalized over a 6-year period to determine the racial and ethnic differences in the risk of intracranial hemorrhage (ICH) and the effect of warfarin treatment on ICH risk. Treatment with warfarin was associated with a 2-fold greater risk of ICH in whites, a 4- to 5-fold greater risk in both blacks and Hispanics, and a 15-fold greater risk in Asians. After adjusting for established stroke risk factors and warfarin use, Asians were 4 times as likely as whites to have ICH, whereas blacks and Hispanics were twice as likely.
Atrial fibrillation is a common arrhythmia that increases the risk of stroke by 5 times.1 Approximately 15% of strokes occur in patients with atrial fibrillation. Prospective studies of predominantly white subjects have shown that warfarin (Coumadin) therapy reduces the risk of stroke by about 64%.2 The annual excess risk of intracranial hemorrhage (ICH) secondary to warfarin treatment is estimated to be 0.2%.3 In the general population, stroke incidence among Hispanics and blacks is twice as high as among whites.4,5 However, the rate of ICH as a proportion of all strokes is 15% to 20% in whites but 20% to 30% in nonwhites.4,6,7 It is unknown whether the risk of ICH outweighs the reduction of ischemic stroke with warfarin treatment in nonwhites with atrial fibrillation. The goal of this study was to assess differences in the risk of warfarin-related ICH among whites, blacks, Hispanics, and Asians, and to estimate the hazard ratio of ICH in nonwhites compared with whites after adjusting for warfarin use and established stroke risk factors.
Patients and methods
We identified all Kaiser Permanente Southern California patients who had been hospitalized for the first time for atrial fibrillation or atrial flutter (International Classification of Diseases, 9th Revision, Clinical Modification codes 427.31 and 427.32) from January 1995 to December 2000 through a search of hospital discharge records. Because the risk of stroke among patients with atrial flutter is similar to that of atrial fibrillation, patients with atrial flutter were included in the analysis.8
We excluded patients who had a previous stroke or transient ischemic attack, congenital heart disease, heart transplantation, brain tumor, mitral valve replacement prior to stroke or transient ischemic attack, mitral valvuloplasty, and rheumatic heart disease. We identified patients with diabetes mellitus, hypertension, and heart failure, which are risk factors for stroke.9 Only whites, blacks, Hispanics, and Asians were included in the study, as these 4 groups encompassed 97% of all patients.
Each patient was followed from the day he or she was admitted to the hospital for atrial fibrillation until a censoring event (disenrollment from Kaiser Permanente Southern California, carotid endarterectomy within 2 weeks of a stroke, mitral valve replacement, ischemic stroke, non-ICH stroke death, or the end of the study) or an outcome event (fatal or nonfatal ICH) occurred.
We initially identified all strokes by electronic searches of our databases. The chart for each patient identified as having a stroke was reviewed to verify that the stroke was primary and nontraumatic. Computed tomography brain imaging was required to diagnose ICH. Stroke events that did not qualify were censored.
We retrieved all patients' warfarin prescription records and all INR values. Once a patient filled a warfarin prescription, he or she was considered to have taken the medication for 100 days. When there was a break between prescriptions of >100 days, the patients' INR values were used to determine whether they had been taking warfarin. If the INR was ≥ 2, patients were considered to have been taking the medication. Therapeutic INR values were considered valid for 60 days if there were no subsequent INR values or if the supply of warfarin was expected to be finished. With this procedure, an individual patient may have contributed to the follow-up times of both warfarin-treated and nonwarfarin-treated groups. We referred to the percentage of follow-up time that subjects were taking warfarin as percent-time receiving warfarin.
We used Poisson log-linear regression analysis to compute crude event rates, which were reported as 100 person-years of follow-up. The effect of race/ethnicity on the risk of stroke was analyzed using Cox proportional hazard models, after adjusting for percent-time receiving warfarin, diabetes, heart failure, hypertension, sex, and age.
Of the 22,196 patients who were hospitalized with atrial fibrillation for the first time, 18,867 qualified for entry into the study. Stroke factors were more prevalent among Asians and Hispanics, and whites were the least likely group to have diabetes. Heart failure and hypertension were more common among blacks. The median age was 74 years, and whites were 6 years older than nonwhites. For more than 80% of the follow-up period, approximately 25% of patients were taking warfarin. Approximately 40% to 45% of patients were not prescribed warfarin at any time during the study. No between-group differences existed in the percent-time on warfarin.
We identified 185 patients with ICH over the course of the study, of whom 173 qualified for analysis. The cohort follow-up totaled 63,204 person-years, with 0.27 ICH events per 100 person-years. ICH events (total person-year follow-up) according to race/ethnicity were as follows: white, 112 (49,324); black, 19 (5004); Hispanic, 23 (6206); and Asian, 19 (2669).
The ICH crude rate for the whole sample was 0.47 per 100 patient-years while taking warfarin and 0.15 while not taking warfarin, for a rate ratio of 3.12 (Table 1). For all groups, taking warfarin increased the risk of ICH, but major differences in the degree of risk existed among groups. Warfarin increased ICH risk by 2.3-fold in whites, 5-fold in Hispanics and blacks, and 15-fold in Asians. ICH risk was increased in both those younger and older than 75 years of age. Risks were the same among men and women. The risk of ICH increased 3-fold for patients with stroke risk factors while taking warfarin, but ICH was extremely rare for those with no stroke risk factors. The risk of ICH increased 3.3-fold for patients with hypertension while taking warfarin. Patients taking warfarin had an increased risk of ICH whether or not they had heart failure or diabetes.
Table 1. Intracranial hemorrhage (ICH) event rates.
*Risk factors for stroke include age ≥75 years, hypertension, diabetes, and heart failure.
CI indicates confidence interval. (Reprinted from Shen AY-J, Yao JF, Brar SS, et al. Racial/ethnic
differences in the risk of intracranial hemorrhage among patients with atrial fibrillation. J Am Coll Cardiol. 2007;50:309-315, with permission from Elsevier.)
The risk of ICH was 4 times as great in Asians compared with whites after adjusting for age, sex, hypertension, diabetes, heart failure, and percent-time treated with warfarin; the risk was 2 times as great for Hispanics and blacks after adjusting for the same factors (Table 2). Each advancing year in age was associated with a 4% increase in the risk of ICH. Compared with patients who did not take warfarin, the risk of ICH was 1.8 to 2.1 times as great for those with > 41% time on warfarin. No independent associations between ICH and hypertension, diabetes, or heart failure were shown. Kaplan-Meier estimates of event-free survival (Figure) showed a consistent increased risk in nonwhites compared with whites.
Table 2. Adjusted hazard ratio for intracranial hemorrhage.
CI indicates confidence interval. (Reprinted from Shen AY-J, Yao JF, Brar SS, et al. Racial/
ethnic differences in the risk of intracranial hemorrhage among patients with atrial fibrillation. J Am Coll Cardiol. 2007;50(4):309-315, with permission from Elsevier.)
Figure. Adjusted intracranial hemorrhage (ICH) event-free curve. (Reprinted from Shen AY-J, Yao
JF, Brar SS, et al. Racial/ethnic differences in the risk of intracranial hemorrhage among patients
with atrial fibrillation. J Am Coll Cardiol. 2007;50:309-315, with permission from Elsevier).
Warfarin-related ICH was associated with anticoagulation intensity. The median duration of patient-time that INR was ≥ 2 was about 70% for the entire sample. The patient-time spent in the therapeutic range (INR 2-3) overall was 54.5%. Time spent in the therapeutic range was not significantly different among whites, Hispanics, and Asians, but blacks spent more time in the subtherapeutic range (INR < 2) than the other groups. The INR values were >3 for about 9% of patient-time, with no differences among groups. The percentage of ICH patients with INR values > 3.0 or ≥ 4.0 at the time of ICH was similar among groups.
Although more than 25% of US residents are nonwhite, the treatment, prognosis, and epidemiology of atrial fibrillation in nonwhites has not been extensively researched. In the Atrial Fibrillation Investigators pooled analysis,10 which included the 5 pivotal randomized studies showing the benefit of warfarin therapy, nonwhites constituted only 10% of the combined cohort.
Significant racial/ethnic differences exist in the epidemiology of stroke. ICH is more prevalent among nonwhites. The relative risk of ICH was shown to be 2 to 3 for blacks and Hispanics compared with whites in the Northern Manhattan Stroke Study.4 A 1.6-fold greater risk of ICH was shown in northern California Asians compared with whites.11 Among individuals who died of stroke from 1995 to 1998 in the United States, ICH accounted for 18% among whites, 24% among blacks, 35% among Hispanics, and 38% among Asians.12 These studies show that there is a higher baseline risk of ICH among minorities.
The risk of warfarin-related ICH is increased with anticoagulation intensity, hypertension, and advanced age.13 The nonwhites in our study were 6 to 7 years younger than whites, but they had higher ICH rates. Asians and blacks were slightly more likely to be hypertensive, but this finding was unlikely to account for the enormous risk we found. Compared with the other groups, blacks were more likely to be underanticoagulated, but there were otherwise no differences in anticoagulation intensity. The high risk of ICH among nonwhites, therefore, cannot be accounted for by overanticoagulation.
This study cannot elucidate the mechanisms responsible for the observed discrepancy in ICH risk. Certain polymorphisms that pertain to the pharmacokinetics and pharmacodynamics of warfarin are found in greater or lesser proportions in certain racial/ethnic groups.14 To what extent, if any, such polymorphisms influenced our findings is unknown. Some studies have suggested that Chinese patients may require a lower target INR.15 Future research should evaluate whether anticoagulation intensity should be modified according to the patient's racial/ethnic group.
Our study consisted of patients who were hospitalized with atrial fibrillation as a primary diagnosis. The findings of our study may not be generalizable to nonhospitalized patients. Because of limitations of our past medical records system, hypertension was considered a dichotomous variable. We were not able to control for the stage of hypertension or the adequacy of blood pressure control.
We assessed a multiethnic cohort of patients with atrial fibrillation hospitalized over a 6-year period and followed for a median of 3.3 years. All races experienced an increased risk of ICH while taking warfarin. However, nonwhite patients, especially Asians, experienced a substantially higher risk of warfarin-related ICH compared with whites.