Clinical Question
What is the best way to predict the risk of bleeding in patients taking warfarin (Coumadin)?
Evidence Summary
When considering anticoagulation therapy in patients with atrial fibrillation or venous thromboembolism (VTE), physicians and patients must balance the benefits of anticoagulation with the risk of bleeding, particularly major bleeding complications. For example, in patients who are at high risk of bleeding, physicians may wish to consider aspirin instead of warfarin, especially if the risk of stroke or recurrent VTE is relatively low.
The Outpatient Bleeding Risk Index (OBRI; Table 11–3) is one of several models that have been developed to predict the risk of bleeding with warfarin (Table 21–5 ). The OBRI, which assigns one point for each of four variables, was developed in a population of 556 patients with VTE who were discharged on warfarin therapy.1 It was validated in a similar group of 264 patients.1 The OBRI was a good predictor of bleeding risk in the initial study and in two subsequent validation studies,2,3 but a more recent and larger study found it less helpful in identifying patients at high risk of major bleeding.5 However, this study did not include data for patients with renal impairment, and because all patients were 65 years or older (two of the four variables), the score could not be fully calculated.
Table 1. Outpatient Bleeding Risk Index
| Variable | Points |
|---|---|
| 65 years or older | 1 |
| History of stroke | 1 |
| History of gastrointestinal bleeding | 1 |
| Recent myocardial infarction, severe anemia, diabetes mellitus, or renal impairment* | 1 |
| Total points: _____ |
*—Anemia is defined as hematocrit level < 30%, and renal impairment is defined as serum creatinine level > 1.5 mg per dL (133 μmol per L).
†—Only two patients in this risk group.
Adapted with permission from Aspinall SL, DeSanzo BE, Trilli LE, Good CB. Bleeding Risk Index in an anticoagulation clinic. Assessment by indication and implications for care. J Gen Intern Med. 2005;20(11):1009, with additional information from references 1 and 2.
Table 2. Decision Rules to Predict Bleeding Risk in Patients Taking Warfarin (Coumadin)
| Study | Participants and indications | Demographics | Prediction of major bleeding by risk group | ||
|---|---|---|---|---|---|
| Beyth, 1998 (n = 820)1 | Patients discharged on warfarin therapy (VTE, 47%; cardiac surgery, 18%; other, 35%) | Mean age = 60 years; 53% women | One-year bleeding risk using OBRI: | ||
| Low risk: 3% | |||||
| Intermediate risk: 12% | |||||
| High risk: 48% | |||||
| Kuijer, 1999 (n = 1,021)4 | Patients discharged on warfarin therapy after diagnosis of VTE | Mean age = 61 years; 51% men | 90-day bleeding risk: | ||
| Low risk: 1% | |||||
| High risk: 7% | |||||
| Wells, 2003 (n = 222)2 | Pulmonary embolism or deep venous thrombosis; started on low-molecular-weight heparin as outpatients and then switched to warfarin | Mean age = 58 years; 43% women | Bleeding risk per person-year using OBRI: | ||
| Low risk: 0% | |||||
| Intermediate risk: 4.3% | |||||
| High risk: not applicable* | |||||
| Aspinall, 2005 (n = 1,269)3 | Patients treated with warfarin at a Veterans Affairs anticoagulation clinic | Mean age = 68 years; 92% men | Bleeding risk per person-year using OBRI: | ||
| Low risk: 0.8% | |||||
| Intermediate risk: 2.5% | |||||
| High risk: 10.6% | |||||
| Shireman, 2006 (n = 26,345)5 | Registry of patients hospitalized with atrial fibrillation and discharged on warfarin therapy | All 65 years or older (88% 70 years or older); 53% women | 90-day bleeding risk | ||
| Shireman rule: | |||||
| Low risk: 0.9% | |||||
| Intermediate risk: 2.0% | |||||
| High risk: 5.4% | |||||
| Intermediate- plus high-risk: 2.3% | |||||
| OBRI†: | |||||
| Intermediate risk: 1.0% | |||||
| High risk: 2.5% | |||||
| Kuijer rule: | |||||
| Intermediate risk: 1.5% | |||||
| High risk: 1.8% | |||||
OBRI = Outpatient Bleeding Risk Index; VTE = venous thromboembolism.
*— Only two patients in high-risk group.
†— Because all patients were 65 years or older, all were at intermediate or high risk. No data were available for patients with renal impairment, one of the four variables in the OBRI.
Another model that uses only three variables (60 years or older, female sex, and presence of malignancy) predicts 90-day bleeding risks of 1 percent in low-risk patients and 7 percent in high-risk patients.4 However, this rule was not validated in a larger, more recent study,5 and does not include important risk factors such as anemia, history of bleeding, or use of antiplatelet agents. Therefore, it cannot be recommended for clinical use.
Most recently, Shireman and colleagues developed a new prediction model using 19,875 patients hospitalized with atrial fibrillation and discharged on warfarin therapy.5 The multivariate model was validated in 6,470 patients.5 It has eight clinical variables and identifies groups at low, intermediate, and high risk of major bleeding within 90 days of hospital discharge (Table 3).5 The Shireman model has good face validity, and because it was developed and validated in a large group of patients, it can distinguish between recent and remote bleeding, and can account for concurrent use of antiplate-let agents. However, as with the OBRI, only a small percentage of patients are identified as high risk (3.4 percent). When the intermediate- and high-risk groups are combined, two groups are created (low- and intermediate/high-risk) with bleeding risks of 0.9 and 2.3 percent, respectively. These results are remarkably similar to the 90-day risks found in the same study when it evaluated the OBRI (1.0 percent for intermediate risk, and 2.5 percent for high risk). Thus, these rules provide similar results. The difference between 1 and 2.5 percent over 90 days seems small, but becomes more significant over time as large numbers of patients (especially those with atrial fibrillation) are on anticoagulation therapy for many years.
Table 3. Shireman Rule to Predict Bleeding Risk in Older Patients Taking Warfarin (Coumadin)
| Variable | Points |
|---|---|
| Anemia | 86 |
| Alcohol or drug abuse | 71 |
| Recent bleeding | 62 |
| Remote bleeding | 58 |
| 70 years or older | 49 |
| Female sex | 32 |
| Antiplatelet use (e.g., aspirin, clopidogrel [Plavix]) | 32 |
| Diabetes mellitus | 27 |
| Total points: | ______ |
| Score | Risk of major bleeding at 90 days |
| Low risk: < 108 | 0.9% (35 / 3,889) |
| Intermediate risk: 108 to 218 | 2.0% (48 / 2,400) |
| High risk: > 218 | 5.4% (12 / 222) |
| Intermediate/high risk combined | 2.3% (60 / 2,622) |
note: The original score was presented as an equation with factors between 0 and 1.0. Factors have been multiplied by 100 to create whole numbers and an additive score. The denominator for major bleeding was calculated from the percentage and numerator, because it was not reported in the original article.
Information from reference 5.
There are potentially important differences between the two models. Although the Shireman model was developed and validated in a large population, it is somewhat complex to calculate and is limited to patients 65 years or older with atrial fibrillation.5 The OBRI is simpler and has been validated in patients with atrial fibrillation or VTE, and in younger patients.1–3 Either rule can be used confidently—in combination with predictors of the risk of stroke or recurrent VTE—to help make decisions about the treatment strategy that best balances potential benefits and harms.
