Indefinite anticoagulation after a first episode of unprovoked venous thromboembolism: yes

Authors


Clive Kearon, Hamilton Health Sciences, Henderson Division, 711 Concession Street, Hamilton, ON L8V 1C3, Canada.
Tel.: +1 905 383 2252; fax: +1 905 574 7625; e-mail: kearonc@mcmaster.ca

Risk of recurrent venous thromboembolism after a first episode

Among consecutive patients with a first episode of venous thromboembolism (VTE) who stop anticoagulants after at least three months of treatment, about 30% have another episode within eight or ten years, which corresponds to an average risk of acute VTE of about 3% or 4% each year [1,2]. Among persons of a similar age (i.e. mean of 60 years) who have not had VTE, the risk of VTE is about 0.1% each year [3]. Therefore, after treatment of a first episode of VTE, patients have, on average, a risk of new episodes of VTE that is about 40 times higher than for those who have never had VTE. However, patients with a first episode of VTE are a heterogeneous group and the risk of having a second episode of VTE after stopping anticoagulant therapy differs markedly among patients [4]. There is strong evidence that the circumstance under which VTE has occurred is the most important predictor of the risk of recurrence after anticoagulant therapy is stopped [4,5]. If VTE was provoked by a reversible (i.e. temporary) risk factor, such as recent major surgery, patients have a much lower risk of recurrence compared to patients who had an unprovoked (also termed ‘idiopathic’) episode of VTE, or patients who had VTE that was associated with a non-reversible risk factor, such as cancer [4,5].

Although the practise has never been evaluated in a controlled trial, because patients with active cancer have a high risk of recurrence if treatment is stopped, it is widely accepted that patients with VTE and cancer should remain on long-term anticoagulant therapy [6–8]. It is also widely accepted that patients who have VTE provoked by a major reversible risk factor, such as recent surgery, have a low risk of recurrent VTE and that they do not need to be treated for longer than three months [6–9]. For patients with a first episode of unprovoked VTE, a group which has a risk of recurrence that is higher than for VTE provoked by a reversible risk factor and lower than for VTE associated with cancer [4,10], there is controversy about whether anticoagulant therapy should be continued indefinitely. Based on the results of recently completed controlled trials, I argue that most patients with a first episode of unprovoked proximal deep vein thrombosis (DVT) or pulmonary embolism (PE) would benefit from long-term therapy, provided anticoagulant therapy is looked after well.

Trials that have evaluated long-term anticoagulant therapy for unprovoked VTE

Three trials have compared long-term anticoagulant therapy with stopping therapy in patients with unprovoked VTE who have completed at least three months of initial treatment: Long-term Anticoagulation for First Idiopathic venous Thromboembolism (LAFIT) [11], Prevention of Recurrent Venous Thromboembolism (PREVENT) [12] and PROLONG [13] (see Table 1). Although the design of these three studies was similar, there were important differences among the studies. The LAFIT study enrolled patients with a first episode of proximal DVT or PE (patients with isolated distal DVT were not eligible) [11]. The PREVENT study enrolled patients with one (62%) or more (38%) episodes of VTE, including patients who had isolated distal DVT (proportion unknown) [12]. The PROLONG study enrolled patients with a first episode of proximal DVT or PE who had a positive D-dimer result one month after stopping anticoagulant therapy, a subgroup of patients who have a higher than average risk of recurrence [10,14,15]. Long-term anticoagulant therapy was targeted to an International Normalized Ratio (INR) of 2.5 (range INR 2.0–3.0) in the LAFIT and PROLONG studies, and to an INR of 1.75 (range INR 1.5–2.0) in the PREVENT study. The control group who stopped therapy received placebo and sham INR monitoring in the LAFIT and PREVENT studies, whereas an open-label design was used in the PROLONG study.

Table 1.   Randomized trials that have evaluated long-term anticoagulant therapy for venous thromboembolism
Acronym first author yearInterventionBlindingPatients analyzedLength of follow-upRecurrent DVT or PEMajor bleedingTotal mortalityComments
  1. DVT, deep vein thrombosis; INR, International Normalized Ratio; PE, pulmonary embolism; RR, relative risk; VKA, vitamin K antagonist therapy; VTE, venous thromboembolism. Parentheses after relative risk indicate 95% CIs for these estimates.

LAFIT (Kearon 1999) [11] VKA stopped (placebo)
VKA (INR 2.0–3.0) for two more years
Allocation: yes
Patients: yes
Caregivers: yes
Adjudications: yes
83/83
79/79
10 months (mean) (max 2 years)17/83 (20%)
1/79 (1%) RR 0.1 (0.0, 0.5)
0/83
3/79 (4%) RR 7.4 (0.4, 140)
3/83 (4%)
1/79 (1%) RR 0.3 (0.0, 3.3)
Population: first unprovoked proximal DVT or PE (5% had previous provoked VTE). The recurrent VTE in the VKA patient was after stopping VKA.
Prevention of Recurrent Venous Thromboembolism (Ridker 2003) [12]VKA stopped or not restarted (placebo)
VKA INR 1.5–2.0
Allocation: yes
Patients: yes
Caregivers: yes
Adjudications: yes
253/253
255/255
2.1 years (mean) (max 4.3 years)37/253 (15%)
14/255 (5%) RR 0.4 (0.2, 0.7)
2/253 (1%)
5/255 (2%) RR 2.5 (0.5, 13)
8/253 (3%)
4/255 (2%) RR 0.5 (0.1, 1.6)
Population: unprovoked DVT (distal or proximal) or PE (first episode in 38%). eight recurrent VTE in the VKA group after stopping VKAs.
PROLONG (Palereti 2006) [13]Remain off (stop) VKA
Restart Indefinite VKA (INR 2.0–3.0) (not blinded)
Allocation: yes
Patients: no
Caregivers: no
Adjudications: yes
103/105
120/122
1.4 years (mean) (max 1.5 years)18/103 (17%)
2/120 (2%) RR 0.1 (0.0, 0.4)
0/103
1/120 (1%) RR 2.6 (0.1, 63)
1/103 (1%)
1/120 (1%) RR 0.9 (0.1, 14)
Population: first unprovoked proximal DVT or PE, treated for ≥3 months. VKA stopped and D-dimer positive one month later.
Eight control patients restarted VKA, some after superficial phlebitis. One recurrent VTE in VKA group after VKA stopped.
DURAC 2 (Schulman 1997) [17]VKA (INR 2.0–2.85) for six months
VKA (INR 2.0–2.85) Indefinitely
Allocation: yes
Patients: no
Caregivers: no
Adjudications:
VTE, yes;
Other, unlikely
111/111
116/116
4 years23/111 (2%)
3/116 (3%) RR 0.1 (0.0, 0.4)
3/111 (3%)
10/116 (9%) RR 3.2 (0.9, 11)
16/111 (14%)
10/116 (9%) RR 0.6 (0.3, 1.3)
Second VTE: DVT (distal or proximal) or PE. All recurrent VTE in the indefinite VKA group were after stopping VKAs. Bleeding during the first six months of VKA in one of six-months group and six of indefinite group (only asked about bleeding while on VKAs).
ELATE (Kearon 2003) [19]VKA INR 1.5–1.9
VKA INR 2.0–3.0 (blinded)
Allocation: yes
Patients: yes
Caregivers: yes
Adjudications: yes
369/369
369/369
2.4 years (mean)16/369 (4%)
6/369 (2%) RR 0.4 (0.1, 0.9)
9/369 (2%)
8/369 (2%) RR 0.9 (0.3, 2.3)
16/369 (4%)
8/369 (2%) RR 0.5 (0.2, 1.2)
Population: unprovoked proximal DVT or PE (first episode in 31%). Treated for ≥ three months. VKA (INR 2.0–3.0) (mean 12 months) five recurrent VTE in INR 1.5–1.9 and three in the INR 2.0-3.0 group after stopping VKAs.

As shown in Table 1, the findings of the three studies were consistent and unequivocal. Allocation to long-term, standard-intensity, anticoagulation reduced recurrent VTE by about 90% (intention-to-treat analysis), and in those who remained on anticoagulant therapy (inclusive of times when anticoagulation was ‘subtherapeutic’), the reduction in recurrence was close to 95% [11,13]. The frequency of recurrent VTE in the control groups of these three studies was high (Table 1). Based on these findings and those of other prospective studies [1,2,5,16], a reasonable estimate is that the risk of recurrence is about 10% and about 30% one and five years, respectively, after stopping anticoagulant therapy in patients with a first unprovoked proximal DVT or PE. There was an increase in bleeding with long-term therapy, with these and other controlled studies suggesting that remaining on anticoagulants is associated with about a doubling of the frequency of major bleeding [i.e. 1.1% vs. 0.6% per patient year; relative risk of 1.8; 95% CI: 0.7–4.5] [16]. With respect to balancing the clinical importance of the reduction in VTE and the increase in bleeding with long-term anticoagulant therapy, the benefit to remaining on anticoagulant therapy was so overwhelming that the Data Safety and Monitoring Committees stopped the LAFIT and PREVENT studies early (the PROLONG study did not include an interim analysis) [11–13].

Two other randomized trials have evaluated long-term anticoagulation in patients with VTE. Although these studies did not address whether patients with a first episode of unprovoked VTE should stop or continue therapy, their results complement the findings of the LAFIT, PREVENT and PROLONG studies. The Duration of Anticoagulation (DURAC 2) study showed that indefinite anticoagulant therapy, rather than six months of treatment, markedly reduced the risk of recurrence in patients with a second episode of VTE (see Table 1) [17]. Long-term anticoagulant therapy was associated with an increase in bleeding. However, of the 10 episodes of major bleeding that occurred during four years of follow-up in the 116 patients in the long-term treatment arm, only four occurred during the extended phase of treatment (i.e. after the first six months of therapy) [17,18]. The Extended Low-intensity Anticoagulation for Thromboembolism (ELATE) study, which randomized patients with one or more episodes of unprovoked VTE who had already been treated with standard-intensity therapy for at least three months (mean 12.2 months) to long-term anticoagulant therapy with either a target INR of 1.5–1.9 or an INR 2.0–3.0, showed that standard-intensity therapy was more effective than low-intensity therapy and that the rate of major bleeding was low, and similar, in the two groups (see Table 1) [19].

So, if randomized trials have shown unequivocal benefit to continuing anticoagulant therapy in patients with unprovoked VTE, why are some physicians opposed to this practise and are their criticisms of long-term anticoagulant therapy likely to be valid? Some of the most common arguments against long-term anticoagulation for unprovoked VTE are considered here.

The duration of follow-up was relatively short in the randomized trials; with indefinite anticoagulant therapy the increase in bleeding will outweigh the benefits of reduced VTE.

Clearly, the longer patients remain on anticoagulant therapy, the higher their cumulative risk of bleeding. However, if the risk of major bleeding is 1–2% per year during the extended phase (i.e. after the first three months) of anticoagulant therapy (see Table 1, and see meta-analysis by Ost et al. [16]), and if about half of these bleeds are attributable to the anticoagulant therapy (bleeding also occurs in patients who are not anticoagulated) [16], the increase in major bleeding as a result of anticoagulant therapy is about 0.5–1% per year. Therefore, even after 10 or more years of follow-up, the proportion of patients who will have a major bleed as a result of anticoagulant therapy will be small. In contrast, if anticoagulant therapy is stopped, about 30% of patients are expected to have a recurrent episode of VTE within five years, and about 50% within 10 years [1,2,5,11,12,16]. Patients who have a non-fatal recurrence, in addition to risking development of chronic complications such as post-thrombotic syndrome, will need to be restarted on anticoagulant therapy. In the PREVENT study, which used a composite of recurrent VTE, major bleeding and death, as the primary outcome measure, the cumulative frequency of these events was much lower in those who remained on anticoagulant therapy and the longer the follow-up, the greater this advantage (Fig. 1) [12]. In the ELATE study, for the same composite outcome that assessed both efficacy and bleeding, there was a benefit to standard-intensity over low-intensity therapy that increased with the duration of follow-up (Fig. 1) [19,20]. Taken together, the results of the PREVENT and ELATE studies indicate that the benefits of long-term anticoagulant therapy do not diminish with an extended duration of therapy; instead, they just continue to increase.

Figure 1.

 Cumulative risk of the composite outcome of recurrent venous thromboembolism, major bleeding and death in the Prevention of Recurrent Venous Thromboembolism (PREVENT) and Extended Low-intensity Anticoagulation for Thromboembolism (ELATE) studies. Hazard ratio 0.52 (95% CI: 0.31–0.87) for low-intensity warfarin vs. placebo in the PREVENT study. Hazard ratio 0.52 (95% CI: 0.31–0.95) for standard-intensity vs. low-intensity warfarin in the ELATE study. (The figure is adapted from Ridker et al. [12] and Kearon et al. [19,20].)

The consequences of major bleeds are more severe than the consequences of a recurrent episode of VTE.

About 10% of major bleeding episodes during long-term anticoagulant therapy are fatal [21]. After a DVT, about 5% of recurrent episodes of VTE are fatal, whereas after a PE a higher proportion of recurrent episodes of VTE are expected to be fatal as more of the recurrences will also be PEs [1,4,22,23]. Therefore, while it is acknowledged that the consequences of major bleeding tend to be worse than the consequences of a recurrent episode of VTE (particularly after a preceding DVT rather than a PE), the much greater increase in the number of recurrent episodes of VTE if anticoagulant therapy is stopped (e.g. about 27 extra episodes in 100 patients over five years) far outweigh the consequences associated with the relatively modest increase in major bleeding if anticoagulant therapy is continued (e.g. about two to three extra episodes in 100 patients over five years).

Long-term anticoagulant therapy has not been shown to reduce mortality in patients with unprovoked VTE.

None of the trials that evaluated long-term anticoagulant therapy in patients with unprovoked VTE was designed (i.e. large enough) to detect a reduction in fatal PE or all cause mortality. However, as shown in Table 1 and Fig. 2, the LAFIT, PREVENT and ELATE studies all show fewer deaths in patients who were randomized to long-term anticoagulant vs. no anticoagulant therapy (LAFIT and PREVENT), or to standard-intensity vs. low-intensity therapy (ELATE) [11,12,19]. A similar finding was observed in DURAC 2 (Table 1, Fig. 2) [17]. Combining the results of these four studies [11,12,17,19] and the PROLONG study (no deaths in either group) [13] yields a relative risk for death of 0.53 (95% CI: 0.33–0.87) in favor of the more aggressive anticoagulant strategies. Therefore, although individual trials were not designed to compare rates of death, their combined results are consistent with a reduction in all cause mortality in patients with unprovoked VTE who receive long-term anticoagulant therapy (Fig. 2).

Figure 2.

 All cause mortality in patients with venous thromboembolism allocated to long-term anticoagulation vs. no anticoagulation, or to standard-intensity vs. low-intensity anticoagulation.

Better predictors of individual risk of recurrence are required before patients with a first unprovoked VTE should be treated with long-term therapy.

Patients with isolated distal (i.e. calf) DVT have about half the risk of recurrent VTE as patients with proximal DVT or PE [1,4,24,25] and most physicians, including myself, agree that long-term anticoagulation is not justified in patients with a first unprovoked VTE that is an isolated calf DVT. As previously noted, the risk of recurrence after a first unprovoked proximal DVT or PE is estimated to be about 30% after five years, and 50% after 10 years. If it were possible to identify the 50% of patients who would not have a second episode of VTE, anticoagulants could safely be withdrawn in these patients. Current evidence suggests that female sex [26], low D-dimer levels one month after stopping therapy [10,13], absence of residual DVT on ultrasound [27], and absence of hereditary and acquired thrombophilia [28] may be able to identify patients with a lower risk of recurrence. Among these factors, as pioneered by Palareti et al. [10,13], a low D-dimer level one month after withdrawal of anticoagulants currently appears to have the greatest potential as a way to identify patients with a first unprovoked proximal DVT or PE who have low enough a risk of recurrence that long-term anticoagulant therapy is not indicated [10,13–15]. However, at this time, I believe that no factor is well enough validated as a negative predictor of recurrence in patients with unprovoked proximal DVT or PE that the presence of that factor would justify routinely withdrawing anticoagulant therapy. In order to justify stopping anticoagulant therapy in subgroups of patients with unprovoked VTE, the following studies would need to be performed in the target population: (i) randomized trials showing that long-term anticoagulation is not beneficial; or (ii) rigorous cohort studies showing that the risk of recurrent VTE after stopping anticoagulant therapy was less than a predefined, generally acceptable, rate of recurrence.

Applying the results of the randomized trails to clinical practise

To realize the benefit of long-term anticoagulant therapy in clinical practise, patients with a first episode of unprovoked proximal DVT or PE who are selected for extended therapy should be comparable to the patients who were included in the LAFIT [11], PREVENT [12] and ELATE [19] studies, and need to have access to good quality anticoagulant management. Achieving good quality long-term control of anticoagulant therapy requires commitment from patients and health care providers. Having completed three months of initial anticoagulation, it may be evident that, because of poor compliance or other factors that cannot be corrected, individual patients may be poor candidates for long-term therapy. In addition, some health care providers may not be in a position to provide good quality anticoagulant care and may not have the option of referring patients to another health care provider who can take on that role (e.g. an anticoagulant service). Even if good quality anticoagulant management is accessible, patients may not be candidates for long-term therapy because of a high risk of bleeding and, as the risk of bleeding increases substantially in patients who are older than 75 [29–31], I am reluctant to recommend long-term therapy to such patients. However, provided good quality anticoagulation can be achieved and there is not a high risk of bleeding, patients with a first unprovoked proximal DVT or PE should be offered the option of long-term therapy. I tell patients that if they stop anticoagulant therapy they have about a10% risk of having another episode of thrombosis within the first year, and a 30% risk within five years. If they remain on therapy, their risk of thrombosis will be extremely low, but their risk of having a bleed that is severe enough that they would have to return to the hospital is about 1% (<65 years) or 2% (65–75 years) each year. Those patients who find anticoagulant therapy a burden and who are less fearful of another episode of thrombosis may decide to stop treatment. However, a majority of patients without contraindications elect to remain on therapy and have an annual review to ensure that they are still suitable for such therapy. Patients who have had a PE and males receive a little more encouragement to remain on long-term therapy, but having had proximal DVT or being female are not considered grounds for withholding this highly effective intervention. If patients ask me what I would do if I were in their situation, I tell them with an easy mind that I would remain on treatment.

Disclosure of Conflict of Interests

The author states that he has no conflict of interest.

Ancillary