Unprovoked deep vein thrombosis should be treated with long-term anticoagulation – no


Trevor Baglin, Department of Haematology, Addenbrooke’s NHS Trust, Cambridge CB2 0QQ, UK.
Tel.: +44 0 1223 216748; fax: 44 0 1223 217017; e-mail: trevor.baglin@addenbrookes.nhs.uk

Following a first episode of deep vein thrombosis (DVT) continued treatment with oral anticoagulant therapy will prevent most episodes of recurrence. However, the optimal duration of treatment for each patient is uncertain, because the risk of recurrent thrombosis after stopping treatment and the risk of bleeding associated with anticoagulation are not easily predicted on an individual basis. Consequently, patients are usually treated for a finite period of no more than 6 months [1–3]. Nevertheless, it is now well recognized that when treatment is stopped the incidence of recurrence varies according to the clinical risk factors associated with the initial event; for example, in the first Cambridge prospective cohort study of patients with venous thrombosis (DVT or PE) the cumulative recurrence at 2 years was 19% in patients with a first unprovoked event, 8% in patients with an identifiable non-surgical provoking trigger factor at the time of the first event, and 0% in patients with venous thrombosis occurring within 6 weeks of a surgical procedure [4]. This high rate of recurrence after a first episode of unprovoked venous thrombosis [4–9] is an important observation as the benefit of continued anticoagulation may outweigh the risk after such an event. However, in my opinion, risk stratification with a view to long-term anticoagulant therapy based solely on whether the first episode of venous thrombosis is provoked or unprovoked is too limited and the issue of duration of therapy should be addressed in the wider context of individual risk assessment. Clinical outcome data are available from recently published and ongoing observational and management studies, which predict a move to individual patient-focussed management decisions, as opposed to the traditional disease-focussed approach whereby groups of patients receive a generic treatment regimen, which at the moment is typically a finite period of anticoagulation after a first episode of venous thrombosis.

In relation to the risk of recurrent thrombosis we need to consider:

  • Is the risk of recurrence the same in all patients after an unprovoked event?
  • Is the distinction between unprovoked and provoked only one of several clinical factors that predict individual risk of recurrence?
  • What is the likely consequence of recurrence?
  • Should the results of investigations be taken into account, for example D-dimer measurement?
  • How is unprovoked thrombosis defined in clinical practise and how are patients to be reliably stratified?

In relation to long-term anticoagulation we need to consider:

  • What is long-term anticoagulation?
  • What is the risk of bleeding associated with long-term anticoagulation?

There are also issues such as cost and convenience of long-term treatment and how patients perceive relative risks of thrombosis and bleeding and how their own preferences are taken into account. Finally, there is the issue of whether a decision regarding duration of anticoagulation is being made after a first episode of unprovoked DVT or after a recurrence. For the purpose of this debate I will confine my comments to the issue of duration of anticoagulation after a first episode. At the present time I think many clinicians would favour long-term anticoagulant therapy after recurrent episodes of unprovoked venous thrombosis [10].

Recurrent thrombosis

The risk and consequence of recurrence is not the same in all patients after a first episode of unprovoked DVT. Most patients who suffer an unprovoked DVT do not suffer a recurrence of venous thrombosis and so there is a need to try and stratify patients within this group and target therapy at those at highest risk. The type of the event (provoked or unprovoked) is only one clinical factor in the risk profile. For example, in the second Cambridge prospective cohort study the influence of sex, whether or not the first event was unprovoked or not, and age was evaluated in an interactive model and the relationship between the clinical risk profile and D-dimer level following cessation of anticoagulation was examined (awaiting publication). The modelling showed that the relationship between D-dimer and likelihood of recurrent thrombosis was not a direct interaction and that the D-dimer level was largely a biochemical reflection of the clinical risk profile. In the model, sex was the primary determinant of risk of recurrent thrombosis, with an unprovoked first event emerging as an additional risk factor, which itself was more likely in younger males. Males with an unprovoked first episode of venous thrombosis and a positive D-dimer had a recurrence rate of 33/100 patient-years (95% CI 21–48) whereas females with an unprovoked event and a positive D-dimer had a recurrence rate of only 5/100 patient-years (95% CI 2–11).

The location and size of the initial thrombosis also indicate risk of recurrence. Calf DVT is associated with a lower risk of recurrence than proximal DVT [7,9,11]. DVT and pulmonary embolism (PE) are commonly considered to be different parts of the spectrum of a single disease process but patients presenting with DVT are more likely to suffer a recurrence of DVT than PE, whilst patients presenting with PE are more likely to suffer a recurrence of PE [12,13]. This is causally related to the consequence of recurrence. Case-fatality is significantly lower in patients with DVT without symptomatic PE than in patients with symptomatic PE [12–14]. It is also important to distinguish death due to recurrent PE from total case-fatality when considering the benefit of long-term anticoagulation. In a prospective cohort study the 5-year cumulative incidence of fatal PE after a first episode of DVT was 2.6% [11]. In a meta-analysis the annual rate of fatal PE after discontinuation of anticoagulant treatment after PE was 1.5% (95% CI 0.9–2.2) but 0 per 265 years (95% CI 0–3.6) after DVT. The development of thromboembolic pulmonary hypertension is almost 20 times more likely after recurrent PE as compared with after a single episode of PE [15] and so prevention of recurrent PE, rather than DVT, is particularly important in this regard.

The development of post-thrombotic syndrome (PTS) is no more likely after an unprovoked DVT than after a provoked event [16–18]. However, development of PTS more than doubles the risk of recurrent DVT [17] and so the presence or absence of PTS after an episode of DVT should be taken into consideration.

The distinction between unprovoked and provoked can be problematic in practise. The development of DVT during plaster cast immobilization is acceptably classified as provoked. Is a thrombosis in a healthy female taking a combined oral contraceptive pill provoked or unprovoked? In the Cambridge studies this was classified as provoked [4](and Cambridge 2 ref). When translating the results of clinical studies into routine clinical decision-making, knowledge of such detail is important. Furthermore, it is not always easy to classify patients with any certainty on the basis of whether or not a first event was provoked; for example, is a DVT after unaccustomed exercise provoked or coincidental (unprovoked)? Is a DVT within 7 days of a 4-hour air-flight provoked or coincidental (unprovoked)? It would therefore seem sensible to consider the complete risk profile and moderate management decisions in accordance with the overall apparent risk rather than simply whether the first event appeared unprovoked or provoked.

Long-term anticoagulation

In patients with venous thrombosis the risk of recurrence is only reduced for the duration of anticoagulation and the clinical benefit associated with extending the period of anticoagulation is not maintained after treatment is stopped [19]. Therefore, the decision to be made is between stopping treatment after 6 months or continuing indefinitely. This would be either for life or until a point in time at which the perceived risk of bleeding becomes so high as to make continued treatment disadvantageous; unfortunately, identifying a high risk of bleeding before hemorrhage occurs can be difficult. Duration of anticoagulant therapy is a major determinant of risk of bleeding [20] but data on risk in relation to prolonged anticoagulation is limited. For very long-term therapy beyond 4 years data are non-existent. A meta-analysis of anticoagulant therapy-associated bleeding events in prospective controlled trials or cohort studies of patients with venous thrombosis identified rates of major bleeding of 7 per 100 patient-years, with fatal bleeding in 1.3/100 patient-years, and of intracranial bleeding in 1.15/100 patient-years. However, study periods were usually for only 3 months and bleeding rates were lower after the initial 3 months of therapy [21]. In the Italian Study on Complications of Oral Anticoagulant Therapy (ISCOAT) the mean follow-up was less than 1 year [22]. In a study that evaluated a predictive bleeding index the incidence of bleeding was still accumulating at 4 years [23]. Whilst it is reasonably assumed that bleeding risk may decline after the first few years [24], it is likely to rise again as patients age, particularly in the over 65 years age group [23],and many patients are over 65 when they first present with DVT [25].

Current state of play

It is becoming apparent that risk stratification for recurrence after an episode of venous thrombosis will have clinical utility. However, the complex interaction of risk factors still makes individual risk assessment imprecise. So far studies that have identified and examined risk factors have not been sufficiently large to provide confident risk estimates in subgroups of patients based on interdependent multiple risk factors. The Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis (MEGA) study may be sufficiently large to be informative in this regard [26].

Despite a plethora of randomized trials and observational studies of venous thrombosis there is still a need for further studies that evaluate the benefit and risk of continued anticoagulation in high-risk patients. Such studies should not be prevented by a premature perception that it is now possible to identify those patients who should remain on life-long anticoagulation after a first episode of venous thrombosis. Accepting this caution it is possible to envisage how clinical decision-making may be refined in the light of such studies. Factors that would appear to favour a finite period of anticoagulation after an unprovoked DVT include female sex, calf vein thrombus, absence of symptomatic PE, absence of PTS, negative D-dimer, age over 65 at time of first event, and a high anticoagulant therapy-associated bleeding risk. Factors that would appear to favour long-term anticoagulation include male sex, additional symptomatic PE (especially in patients with reduced cardiorespiratory reserve), severe PTS, a positive D-dimer after completion of 6 months anticoagulation, age under 65 years at time of first event, and a low anticoagulant therapy-associated bleeding risk. For example, in practise one might recommend a finite period of anticoagulation after an unprovoked DVT in a 67-year-old female who presented without associated symptomatic PE and who had a negative D dimer and no evidence of PTS on completion of 6 months anticoagulant therapy. However, long-term anticoagulation might be recommended after an unprovoked DVT in a 47-year-old male who presented with associated symptomatic PE and who had a positive D dimer and moderate to severe PTS after 6 months anticoagulant therapy. In other words, it would not be the fact that the first DVT was unprovoked that would indicate a need for long-term anticoagulation but rather this clinical factor should prompt a careful review of the total risk profile. In my opinion the question ‘Should a patient be treated with long-term anticoagulation after a first episode of unprovoked DVT?’ is the wrong question. The question to which the answer is yes is ‘Should a patient be considered for long-term anticoagulation after a first episode of unprovoked DVT?’.

Disclose of Conflict of Interests

The author states that he has no conflict of interest.