Caval interruption has been, historically, the first "treatment" of venous thromboembolic disease. Following ligation, plication, then clips, and finally intracaval filters constituted the successive improvements of this procedure, which can now be considered reasonably safe. However, reliable and clinically relevant data regarding long-term safety are lacking; recent data suggest that caval filters might increase the risk of recurrent deep venous thrombosis. As almost no controlled trials are available, indications for caval interruption are based on fragile grounds: contraindications to and failure of anticoagulant treatment in patients with recent proximal deep venous thrombosis remain the only two widely accepted indications. Although the PREPIC study, first prospective controlled trial on caval filters, confirmed their efficacy for preventing pulmonary embolism, the addition of caval filters to preventive or curative anticoagulant treatment in high-risk patients is still a matter of debate, because "very high-risk" settings despite anticoagulant treatment remain poorly defined. Finally, the risk-benefit ratio of caval interruption in addition to medical thrombolysis, or as an alternative to preventive or curative anticoagulant treatment appears unfavorable. The relevance of debatable indications, the precise identification of "very high-risk" patients, and the determination of the "best" filter should be assessed in specific prospective clinical trials.