Prediction of the Risk of Bleeding During Anticoagulation Treatment for Venous Thromboembolism

Abstract & Commentary

Synopsis: Overall, this study emphasizes the importance of concurrent malignancy as a major risk factor for bleeding while on dicumarol.

Source: Kuijer PM, et al. Arch Intern Med 1999;159:457-460.

The use of anticoagulation is increasingly common in clinical practice including atrial fibrillation, prosthetic heart valves, venous thromboembolism (VTE) and cardioembolic cerebrovascular events. The major adverse side effect of concern with anticoagulation is bleeding, which can range from mild to life threatening and varies from 10 to 17 per 100 patient-years for all bleeding complications and from 2 to 5 per 100 patient-years for major bleeding complications. Kuijer and associates attempted to construct a prediction rule based on easily identifiable variables that would be useful in clinical practice.

The study was conducted in two phases. In the first phase, Kuijer et al constructed a bleeding risk prediction score based on a careful literature review. Three variables were chosen: age, sex, and presence of malignancy. These variables were repeatedly associated in previous studies with an increased bleeding risk and are readily obtainable in clinical practice.

The score developed ranged from 0 to 5.1; 1.6 points for age older than 60; 1.3 points for female gender; and 2.2 for the presence of malignancy. The optimal cutoff points for the prediction of all bleeding complications as well as for major bleeding complications was determined using receiver operator curve (ROC) analysis. A cohort of 241 patients receiving anticoagulation for a clinical trial was used to construct the ROC. High-risk patients were defined as those with a score of 3 or more points, intermediate risk as a score of 1-3, and low risk as a score of 0. In the test group, a high-risk score was associated with a 26% incidence of all bleeding complications and a 14% incidence of major bleeding complications. In the test group, the high-risk category accounted for 60% of all bleeding complications and 78% of all major bleeding complications.

The score developed in the test phase was validated in a cohort of 780 patients. Using their bleeding risk score, those patients categorized as "high risk" (score > 3) had an incidence of 17% for all bleeding complications and 7% for major bleeding complications. Patients categorized as low risk (score = 0) had an incidence of 4% for all bleeding complications and 1% for major bleeding complications. Approximately 20% of all patients were categorized as low risk. The high-risk category accounted for 37% of all bleeding complications and 53% of all major bleeding complications.


Because of broadening indications for anticoagulation, physicians face the dilemma of whether to initiate dicumarol in patients who have risks for bleeding complications. In such circumstances, the decision involves a careful assessment of risk vs. benefit. While major bleeding complications range from 0% to 15%,1 reliable data are often lacking. In addition, many studies were done before the current international normalized ratio system replaced the prothrombin time. Independent predictors of bleeding include age older than 65 years, history of stroke, history of gastrointestinal bleeding, atrial fibrillation, and other serious comorbid conditions. Comorbid illnesses found to increase the risk of bleeding include heart disease, renal insufficiency, liver disease, malignancy, and use of NSAIDS and diuretics.2 Three or more comorbid conditions were found to be an independent risk factor for bleeding in a multicenter study.3

The simple score reported in this survey is able to identify patients with approximately a 10% incidence of major bleeding episodes. However, the scoring system devised has several limitations. Malignancy is required for inclusion in the high-risk group. This is evidenced by the fact that high risk is defined as a score greater than 3 while the sum of the other two risk factors is 2.9. Also, this score was developed in patients with VTE and may not necessarily be expandable to other indications.

Overall, this study emphasizes the importance of concurrent malignancy as a major risk factor for bleeding while on dicumarol. (Dr. Ost is Assistant Professor of Medicine, NYU School of Medicine, Director of Interventional Pulmonology, Division of Pulmonary and Critical Care Medicine, Northshore University Hospital, Manhasset, NY.)


    1. Palareti G, et al. Lancet 1996;348:423-428.

    2. Landefeld CS, et al. Am J Med 1993;95:315-328.

    3. Stephan D, et al. Ann Intern Med 1993;118:511-520.

The most important risk factor in prediction of bleeding while on dicumarol is:

    a. sex.
    b. presence of malignancy.
    c. hypertension.
    d. age.