We examined data from a large cohort of subjects with atrial fibrillation to determine whether men and women had different risks for atrial fibrillation-related thromboembolism. Women with atrial fibrillation who were not taking warfarin had a 60% higher rate of stroke than did men, independent of other clinical risk factors for stroke. When taking warfarin, both women and men had significant reductions in stroke risk without differences by sex in the rates of bleeding complications. These findings support the addition of female sex to traditional stroke risk stratification schemes for assisting in antithrombotic decision making for patients with atrial fibrillation.
The most common and clinically important cardiac arrhythmia is atrial fibrillation, which is also a potent risk factor for systemic thromboembolism and ischemic stroke.1 Currently, the most effective therapy for preventing atrial fibrillation-related thromboembolism is adjusted-dose anticoagulation therapy with oral vitamin K antagonists, such as warfarin (Coumadin).2 However, warfarin carries a significant risk of hemorrhage; therefore, optimal treatment of patients with atrial fibrillation depends on appropriate risk stratification of stroke and hemorrhage risk.
There are a number of well-known risk stratification methods available to help clinicians estimate a patient’s risk for atrial fibrillation-related stroke. Unfortunately, these strategies sometimes provide contradictory evidence regarding whether men and women have different risks. The Atrial Fibrillation Investigators (AFI)2 and the CHADS23 risk schemes do not show that women with atrial fibrillation are at higher risk for stroke; however, the Stroke Prevention in Atrial Fibrillation (SPAF)4 and Framingham5 risk schemes do. It is also unclear from these schemes whether the risk is disproportionately high for older women, particularly those over the age of 75 years.4
Because accurate risk stratification is important when recommending a potentially harmful drug such as warfarin and because some investigators have reported higher rates of hemorrhage in women,6 clarifying whether women have higher rates of ischemic stroke is crucial. We assessed subjects with nonvalvular atrial fibrillation to determine stroke and hemorrhage rates in women and men while adjusting for other known risk factors for thromboembolism and stroke.
Subjects and methods
International Classification of Diseases, 9th Revision, Clinical Modification
Subjects enrolled in the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study were all patients in the Kaiser Permanente of Northern California network and had nonvalvular atrial fibrillation. The details regarding the assembling of subjects were reported previously.7,8 Subjects with codes for atrial fibrillation were identified by examining electrocardiographic outpatient and inpatient databases. Because our focus was on nonvalvular atrial fibrillation, we excluded subjects with thyrotoxicosis, transient postoperative atrial fibrillation, valvular repair or replacement, and mitral stenosis. Subjects were followed up for a median of 2.4 years, and follow-up was censored at the time of disenrollment from the cohort, death, or an outcome event.
Clinical characteristics of subjects were collected using automated health plan databases. We obtained information on age, sex, and clinical conditions that increased the risk of stroke, specifically prior ischemic stroke, hypertension, heart failure, coronary artery disease, and diabetes mellitus.2 We also collected data on risk factors for bleeding, including prior hemorrhage, cirrhosis, dementia, prior mechanical fall, and anticoagulation intensity (ie, international normalized ratio level). We determined patient warfarin use from pharmacy and laboratory databases using a previously validated algorithm.7 For women, we also collected information on the use of estrogen replacement therapy because it has been linked to higher stroke rates in women.9
We identified incident thromboembolic and hemorrhagic outcome events by searching hospitalization and billing claims databases. Thromboembolic events (ischemic stroke or systemic embolism) were validated by at least 2 physicians. In addition, we searched for major hemorrhagic events, defined as intracranial or major extracranial (hemorrhage into a critical anatomic location, requiring ≥2 units of transfused blood, or fatal). Finally, we determined 30-day mortality after the thromboembolic event by searching state death files, health plan databases, and medical records.
We used chi-square tests to evaluate the clinical characteristics of men and women during periods when they were not taking warfarin. We subsequently compared the rates of hemorrhage and thromboembolism using log-linear models with generalized estimating equations, and we used multivariable log-linear regression to control for clinical risk factors for stroke. We also categorized men and women by their CHADS2 Risk Index.3 To assess whether older women were at disproportionate risk for stroke, we assessed the interaction term between subject age and sex on the risk of thromboembolism. Finally, we compared the rates of major hemorrhage between men and women who were taking warfarin, controlling for risk factors for hemorrhage.
A total of 5795 women and 7764 men were included in our study. As shown in
, women were older than men and were more likely to have a history of hypertension or stroke. Women were less likely to have a diagnosis of diabetes or coronary artery disease, however. Less than a quarter of the total person-time for women was in the setting of estrogen exposure.
Throughout the 15,494 person-years during which subjects were not taking warfarin, 394 thromboembolic events occurred, of which 93.7% were ischemic strokes. The rates of thromboembolism were higher in women than they were in men. The annualized rate of thromboembolism was 3.5% for women and 1.8% for men (unadjusted rate ratio [RR] = 1.9; 95% confidence interval [CI], 1.6-2.4). As shown in
, women also had consistently higher rates of stroke across all risk factor categories and categories of the CHADS2 Risk Index. After controlling for risk factors for stroke (estrogen replacement therapy, diabetes, coronary artery disease, congestive heart failure, hypertension, previous stroke, and age), female sex continued to be an independent risk factor for thromboembolism (adjusted RR = 1.6; 95% CI, 1.3-1.9). The same trend was shown in analyses restricted to ischemic stroke (adjusted RR = 1.5; 95% CI, 1.2-1.8). Death at 30 days did not significantly differ between women and men (23.7% in women and 23.4% in men; = .94). As shown in the
, an independent risk for stroke was observed both in older as well as younger women. The interaction term between age and sex was not significant ( = .38), indicating that the risk of thromboembolism was present in both younger and older women. Estrogen replacement therapy was not significantly associated with an increased risk of thromboembolism (adjusted RR = 1.0; 95% CI, 0.7-1.4).
Warfarin therapy appeared to be equally efficacious in men and women. The crude annualized rate of thromboembolism for subjects taking warfarin was 1.5% for women and 1.2% for men. After adjusting for risk factors for stroke, warfarin therapy correlated with a 60% (95% CI, 50%-70%) reduction in the rate of thromboembolism in women and a 40% (95% CI, 20%-50%) reduction in men.
The rates of warfarin-associated hemorrhage did not differ significantly between women and men. The crude annualized rate of major hemorrhage in subjects taking warfarin was 1.0% for women and 1.1% for men. After adjusting for risk factors for hemorrhage, the adjusted RR of major hemorrhage in women compared with men was 0.8 (95% CI, 0.6-1.1). Women were also at no higher risk for developing warfarin-associated intracranial hemorrhage than were men.
Our study showed that women with atrial fibrillation have a 60% higher rate of stroke and systemic embolism than do men. This risk is independent of age and other risk factors for stroke. Mortality after stroke was no different between men and women, and treatment with warfarin appeared to be equally effective by sex and was associated with similar rates of bleeding complications.
Our study supports the addition of patient sex to existing criteria in stroke risk stratification schemes for atrial fibrillation. Although some studies have not found female sex to be an independent risk factor, these studies were limited by smaller sample sizes and fewer person-years of observation of female subjects. Our findings are similar to those noted in the SPAF study,4 which reported a relative risk associated with female sex of 1.6, and the Framingham study,5 which reported a relative risk of 1.9.
Although some investigators have suggested a higher rate of bleeding complications in women with atrial fibrillation,6 our study showed similar hemorrhage rates in women and men. Of greatest importance, we found no difference in the rates of intracranial hemorrhage, the most devastating complication of warfarin treatment.10 In our study, the rates of major hemorrhage among patients taking warfarin were lower than the rates of ischemic stroke in patients not taking warfarin, supporting the net benefit of warfarin therapy in patients with atrial fibrillation. Because women had a higher baseline rate of stroke when they were not taking warfarin, they derived a greater absolute benefit from warfarin therapy than did men.
The exact mechanism explaining the increased risk of stroke observed in women is not known. We controlled for estrogen replacement therapy, which has been linked to increased stroke risk in other studies. We also controlled for other known risk factors for stroke, such as diagnosed hypertension and a history of prior stroke, although we did not have data on specific blood pressure measurements or left ventricular function. Increased stroke risk may be related to higher levels of prothrombotic factors in women,11,12 but this area needs further exploration.
Among the limitations of our study is that we lacked data regarding possible differences in blood pressure measurements and systolic function, although these factors were also not included in the AFI and CHADS2 risk schemes. We also lacked comprehensive data about potential differences in aspirin use between men and women. To address this issue, we conducted a chart review of 232 subjects who were not receiving warfarin and found that differences in aspirin use between men and women were small and unlikely to affect our original estimates. In addition, our cohort of patients generally had lower rates of stroke than those reported in other settings, which may reflect an overall healthier group of patients.
The results of our study showed that the risk of atrial fibrillation-related stroke and peripheral embolism is higher for women than it is for men. This risk is present in both younger and older age groups. Women appear to derive as much benefit from warfarin as do men and may even benefit more so. Women and men have similar rates of warfarin-associated bleeding complications, including similar rates of intracranial hemorrhage. These results support the addition of female sex to traditional stroke risk stratification schemes for assisting in decision making regarding antithrombotic therapy for patients with atrial fibrillation.
This work was supported by Public Health Services research grant AG15478 from the National Institute on Aging and the Eliot B. and Edith C. Shoolman Fund of Massachusetts General Hospital.