We read with interest the recent article by Prandoni and colleagues [1], which described the incidence of recurrent deep venous thrombosis after a first-time episode of thrombosis. The authors' data could be analyzed to answer a very important clinical question: Is the rate of recurrent deep venous thrombosis in younger patients (<55 years of age) with idiopathic deep venous thrombosis (that is, not following temporary exposure to a risk factor such as surgery, trauma, pregnancy, or significant immobilization) and no known procoagulant condition the same as that in younger persons with idiopathic deep venous thrombosis who have a procoagulant disorder (deficiency of protein C or S, deficiency of antithrombin III, or lupus anticoagulant)? For example, we recently saw a resident-in-training with iliofemoral deep venous thrombosis who had no risk factors for the condition and no family history. Work-up for all procoagulant conditions, including factor V Leiden, was negative. Should this person be managed any differently than a patient in whom, for example, the presence of factor V Leiden is demonstrated?

Because the authors used a stepwise Cox proportional-hazards analysis, the comparison group varied according to the terms that entered the model; thus, it is impossible to specifically compare patients who have idiopathic deep venous thrombosis and no procoagulant conditions to patients who have idiopathic deep venous thrombosis and a procoagulant condition. Use of variables such as "positive family history" and "use of oral contraceptives" should be avoided because they really represent subgroups of patients with idiopathic deep venous thrombosis, many of whom have a procoagulant condition [2]. By restricting an analysis to younger patients who lack significant "temporary" risk factors [3] and using the time-dependent covariate that defines anticoagulation, the authors should be able to compare the hazard of recurrent deep venous thrombosis during and after oral anticoagulation. New information gathered by the authors on the presence or absence of factor V Leiden could be used in this analysis. Much larger clinical studies will be necessary to determine whether one specific procoagulant disorder is associated with a higher incidence of recurrent deep venous thrombosis compared with another disorder.

Finally, if the authors have access to the hospital records of each patient, we suggest that they determine whether subtherapeutic heparin treatment during the first 24 hours of anticoagulation was associated with a higher rate of recurrent deep venous thrombosis, an issue that remains controversial.