IngentaConnect The absence of `minor' risk factors for recurrent venous thromboe...

Abstract:

Summary.

Background: The decision to stop anticoagulation after an episode of venous thromboembolism (VTE) depends on the risk of recurrence. A number of recently discussed risk factors that are not routinely used in clinical practice might be useful for the identification of patients with a low risk of recurrence. Objectives: The aim of this study was to determine the negative predictive value and likelihood ratio of the absence of each of these factors in patients with VTE. Patients/methods: We performed a systematic review of the literature, calculated negative predictive values and likelihood ratios, and summarized the available evidence. Results: A negative D-dimer result, non-elevated factor VIII level and non-elevated thrombin generation after discontinuation of anticoagulant treatment, female sex and distal location might be useful in determining a low risk of recurrence in patients with VTE. Each of these factors individually is associated with a negative predictive value for recurrence-free survival in the next 2-3 years of about 90% (with an a priori risk of about 85%), with negative likelihood ratios between 0.8 and 0.5. Some of the risk factors for first VTE event (FV Leiden and prothrombin mutation) are not useful at all for prediction of non-recurrence. Conclusions: None of the risk factors that we identified is strong enough on its own to guide treatment. Combinations of risk factors might be more useful, but additional data on the independence of these factors are required before a prediction rule can be designed.

Keywords: anticoagulation; likelihood ratio; negative predictive value; recurrence; venous thromboembolism

Document Type: Research article

DOI: 10.1111/j.1538-7836.2009.03557.x

Affiliations: 1: Thrombosis Service, McMaster Clinic, HHS-General Hospital, Hamilton, ON, Canada

The absence of `minor' risk factors for recurrent venous thromboembolism: a systematic review of negative predictive values and negative likelihood ratios