Laurent Pinede, Department of Internal Medicine, Edouard Herriot Hospital, 69 437 Lyons cedex 03, France (fax: +33 ???: e-mail: ???).

Sir, Thank you for your comments.

  • 1 Of course, the risk of haemorrhagic iatrogenic complications increases with the duration of therapy and we have developed this point in paragraph 2, page 554, in our article [1]. The risk of bleeding is higher during the first months of therapy, and decreases after the first three months but, naturally, the cumulative frequency of bleeding events is higher when oral anticoagulant therapy is prolonged [2]. However, the frequency of major or minor bleeding reported in the trials during the past decade is only one-quarter of those reported earlier, probably as a result of better management of anticoagulants in clinical practice [3].

  • 2 The problem of the optimal duration is clearly to balance two individual risk factors: that of haemorrhage and that of recurrence. Of course, the patients included in controlled trials (and in a meta-analysis of these trials) do not represent precisely the general population (selection bias). However, in routine practice, the clinicians can use results from randomized controlled trials, and may individually tailor the duration of anticoagulants according to the different risk factors of each patient.

  • 3 In addition, new clinical trials are needed, specifically designed for high-risk recurrence patients (idiopathic VTE, protein C or S deficiencies, antiphospholipid antibodies, etc.), who could benefit from a long course of treatment with an assessment of a lower level of prolonged anticoagulation (target INR 1.5–2.0) in order to minimize the risk of bleeding.

  • 4 A meta-analysis based on summary data has some limitations [4]. A meta-analysis on individual data offers more reliable means of assessing and quantifying the treatment effect, with less bias and more appropriate analyses [4]. The pooling of individual patient data makes possible adjustments to baseline patient characteristics. It also allows for a more precise and reliable exploration of the effect model. In the case of a rather complex therapy, such as the prevention of VTE recurrence, several key features of the treatment and of the patient conditions can be introduced in the effect model, allowing an in-depth assessment of the treatment–effect relationship. A meta-analysis of individual data is also needed in the field of optimal duration of anticoagulants in VTE, in order to allow a tailored duration for managing each case. We propose to perform this analysis in an international collaborative project, and the protocol has already been set up.

Received 20 July 2000; accepted 24 August 2000.