Warfarin-associated intracranial hemorrhage risk in ethnic/racial groups: Is the devil in the details?

One of the great advances in medical practice arose from studies on atrial fibrillation in the 1990s. These studies demonstrated that patients with chronic atrial fibrillation have a very high risk of devastating stroke and, more importantly, that this risk was reduced by approximately 60% in those who could safely take therapeutic warfarin (Coumadin) compared with only 20% in those who took aspirin.¹ The studies led to guidelines produced by the American College of Chest Physicians, which delineate criteria for the use of warfarin versus aspirin for atrial fibrillation alone or complicated by other risk factors, especially hypertension, heart failure, diabetes, and age over 75 years.² The most feared complication of long-term warfarin use is severe bleeding, especially intracranial hemorrhage (ICH). This risk is estimated from previous trials to be about 0.3% per year, approximately double that of aspirin or no antithrombotic use.^1,3 Recent studies have shown that there has been a dramatic increase in the incidence of warfarin-associated ICH since the widespread use of warfarin in the last 10 years⁴ and that higher risks of ICH exist, particularly in elderly and hypertensive patients.⁵

The study by Shen and colleagues suggests a new and alarming variable to this risk, that of race and ethnicity. This was a retrospective study of 18,867 patients with atrial fibrillation who were members of the Kaiser Permanente health plan in southern California.⁶ In the study, the risk of ICH was compared between groups receiving and not receiving warfarin, and the patients were followed for a median of 3.3 years. The cohort was classified into 4 racial/ethnic groups: white (78.5%), black (8.0%), Hispanic (9.5%), and Asian (3.9%). There were 173 cases of nontraumatic ICH not related to hemorrhagic conversion of an ischemic stroke. The hazard ratio for warfarin-associated ICH was similarly high for the black and Hispanic groups, at 2.04 and 2.06, respectively, using whites as the referent group. For Asians, it was startlingly high, at 4.06, with a 15-fold likelihood of ICH in Asian patients with atrial fibrillation receiving warfarin versus those not receiving warfarin, compared with 4 times the risk in blacks and Hispanics, and 2 times the risk in whites.

There is some support for these findings in previous studies. Among blacks, Hispanics, and Asians, the risk of ICH alone is approximately twice that of whites.^7,8 A study in Hong Kong noted a very high risk of ICH in patients with international normalized ratio (INR) values between 2 and 3,⁹ and an increased risk was found in a study of Japanese patients compared with Westerners.¹⁰ There are numerous studies showing variability in optimum warfarin dosage based on race and ethnicity, particularly noting that it takes lower doses to produce the desired INR range of 2 to 3 in Asian patients.¹¹ Much of this variability may be explained by frequencies of genotypes for enzymes of vitamin K metabolism, particularly VKORC1 and CYP2C9, which vary with racial and ethnic populations.¹² In the study by Shen and colleagues, however, there were no increases in ICH in the nonwhite groups not receiving warfarin, nor was there clearly an increase in supratherapeutic INR levels seen with ICH in the nonwhite groups.

There are other unexpected details in this study. The proportion of white patients younger than age 65 with atrial fibrillation was 18% of that group, consistent with US and Taiwan population studies.^13,14 It was double (37%) this frequency for each of the nonwhite groups in the study, however, suggesting that the nonwhite patients may have had unreported cardiac or other complications associated with their initial admission and/or more concomitant use of aspirin. Hypertension was measured as a dichotomous variable and was presumed consistent over the period of observation, rather than recording actual blood pressures at the time of ICH. This applies as well to INR and blood glucose levels; INR was instead analyzed using thresholds. Finally, because few ICH events were recorded in the nonwhite groups not receiving warfarin, the confidence intervals (CIs) for their marked increase in rate ratios were very broad. This applies especially to the Asian group, whose 15-fold ICH rate increase had a CI of 3.4 to 64.3. With these discrepancies, one should view the findings of the study with appropriate suspicion, and yet the magnitudes of the findings, particularly for the Asian group, remain compelling and not easily bedeviled. Can such a large devil hide in this study’s details?

I believe the most important conclusion one can derive from the study by Shen and colleagues is that caution must be exercised when prescribing warfarin for anyone and perhaps more so if the patient has a black, Hispanic, or Asian racial/ethnic background. This caution must be weighed against the grim specter of atrial fibrillation-related ischemic stroke, especially for those in a very high-risk group with multiple risk cofactors. Absent a confirmatory study demonstrating unacceptably high ICH rates with the use of this medication in these racial/ethnic groups, I would not withhold such powerful therapy from an otherwise intact patient without contraindications. I would instead urge ever more caution in monitoring the devils we know with warfarin risk: supratherapeutic INR, uncontrolled hypertension and diabetes, and advanced age.