The optimal course of oral anticoagulant therapy is determined according to the risk of recurrent venous thromboembolism after stopping therapy and the risk of anticoagulant-related bleeding. Clinical risk factors appear to be important in predicting the risk of recurrence whereas the influence of biochemical and morphological tests is uncertain. The risk of recurrent venous thromboembolism is low when the initial episode was provoked by a reversible major risk factor (surgery): 3 months of anticoagulation is sufficient. Conversely, the risk is high when venous thromboembolism was unprovoked or associated with persistent risk factor (cancer): 6 months or more prolonged anticoagulation is necessary. After this first estimation, the duration of anticoagulation may be modulated according to the presence or absence of certain additional risk factors (major thrombophilia, chronic pulmonary hypertension, massive pulmonary embolism): 6 months if pulmonary embolism was provoked and 12 to 24 months if pulmonary embolism was unprovoked. If the risk of anticoagulant-related bleeding is high, the duration of anticoagulation should be shortened (3 months if pulmonary embolism was provoked and 3 to 6 months if it was unprovoked). Lastly, if pulmonary embolism occurred in association with cancer, anticoagulation should be conducted for 6 months or more if the cancer is active or treatment is on going. Despite an increasing knowledge of the risk factors for recurrent venous thromboembolism, a number of issues remain unresolved. Randomised trials comparing different durations of anticoagulation are needed.