Correspondence: Dr S Middeldorp, Academic Medical Center, Department of Vascular Medicine, F4-276, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands. E-mail: firstname.lastname@example.org
Venous thromboembolism (VTE) is effectively treated with anticoagulant therapy. After an initial treatment phase, extended treatment is effective to prevent recurrence after a first event but this is at the expense of a continued risk of bleeding. Ideally, patients at a high risk of recurrence and low risk of bleeding continue anticoagulant therapy, and for those at low risk of recurrence the duration of treatment can be limited. Identifying these patients, however, is difficult. Duration of treatment after a first VTE provoked by a transient risk factor should be limited to 3 months. Although guidelines suggest extended treatment for all patients after unprovoked VTE unless bleeding risk is high, we emphasize that the long-term risks of recurrent VTE off anticoagulation are uncertain whereas the risk of bleeding associated with anticoagulant therapy increases with age. In the absence of evidence of replaced mortality or improved quality of life with extended anticoagulant treatment, we suggest a limited duration for most patients after a first VTE. Extended treatment can be considered, based mainly on patient preference.
Venous thromboembolism (VTE) is a frequently occurring, potentially fatal disease that affects approximately 2–3 per 1000 men and women annually (Anderson et al, 1991; Naess et al, 2007). It leads to post-thrombotic syndrome (PTS) in approximately a quarter to half of patients who present with deep vein thrombosis (DVT) (Prandoni et al, 1996). VTE comprises deep vein thrombosis (DVT) and pulmonary embolism (PE), which are considered manifestations of the same disease; they have the same risk factors and are treated with a similar anticoagulant regimen (Hull et al, 1983; Doyle et al, 1987; Huisman et al, 1989; Kruit et al, 1991; Kearon et al, 2012).
After discontinuation of anticoagulant treatment, VTE recurs in up to 30% of patients and this risk is highly dependent on circumstances at the time of first VTE (Prandoni et al, 1996; Baglin et al, 2003). Vitamin K antagonists (VKAs) reduce the risk of recurrent VTE by more than 80%, but also induce major bleeding in a substantial number of patients (Hutten & Prins, 2006). Hence, assessing the optimum duration of anticoagulant therapy after an episode of VTE remains challenging. The recently updated 9th edition of the evidence-based guidelines of the American College of Chest Physicians (ACCP) strongly recommends the use of a 3-month course of anticoagulation in patients with VTE provoked by a surgical risk factor, whereas the recommendations for duration of treatment for patients with non-surgical risk factors and unprovoked VTE are weak and based on moderate-quality evidence (Kearon et al, 2012). Here, we describe our view on the arguments that can be used to justify a particular duration of anticoagulation in patients with VTE, with a focus on patients for whom evidence is not straightforward. These include determinants of the risk of recurrent VTE, as well as determinants of the risk of bleeding while using anticoagulant therapy. Ultimately, adequately estimated absolute risks of either of these unwanted health effects (e.g. recurrent VTE and anticoagulant-induced bleeding) should guide the clinician and the patient in their decision to continue extended anticoagulation.
Risk of recurrence after a provoked VTE
The most important determinant of the risk of recurrence of VTE is whether the first VTE was provoked by an eliciting clinical factor and whether this risk factor was transient or persists over time.
Recurrence risk after VTE provoked by a transient risk factor
Transient risk factors strongly increase the risk of developing VTE, but this subsides when the risk factor is no longer present. Surgery is associated with a 22-fold increased risk of first VTE, but the risk of recurrence is very low (Heit et al, 2000a). Two years after discontinuation of anticoagulant therapy, the annual rate of recurrent VTE in patients in whom the first VTE was provoked by a surgical risk factor was 0·7 per patient year [95% confidence interval (CI) 0·0–1·5] in a meta-analysis of 15 randomized controlled trials and prospective cohort studies (Iorio et al, 2010). The incidence of recurrence of VTE associated with non-surgical transient risk factors including trauma, fractures, temporary immobilization, pregnancy or oestrogen use, was 4·2 per patient year (95% CI 2·8–5·6%), which was lower than the 7·8% per patient year (95% CI 6·5–8·2%) after an episode of unprovoked VTE (Iorio et al, 2010).
Pregnancy, postpartum period and oestrogen use
Pregnancy and the postpartum period are strong and transient risk factors for VTE (Pomp et al, 2008). The cumulative incidence of recurrence after stopping anticoagulant therapy was much lower after VTE related to pregnancy, postpartum or oestrogen use, compared to unprovoked VTE (20% and 53% after total follow-up of 10 years) in a large prospective study (Prandoni et al, 2007). Likewise, in a study with a mean follow-up of 3 years, the cumulative incidence of recurrent VTE after a pregnancy-associated first VTE was 5·8% compared to 10·4% in women with an unprovoked VTE [adjusted hazard ratio (HR) 0·6, 95% CI 0·4–0·9] (White et al, 2008). The absolute risk of recurrence is estimated to be between 5 and 10% during a subsequent pregnancy, which is the rationale for antepartum pharmaceutical thrombosis prophylaxis for these women (Pabinger et al, 2002, 2005; De Stefano et al, 2006; Bates et al, 2012).
Whether VTE occurring during oestrogen use [combined oral contraceptives (COC), or hormonal replacement therapy (HRT)] should be considered provoked is controversial, particularly when these preparations have been used for a long time. However, the risk of a first VTE in women using COC remains increased four- to eight-fold regardless of the number of years of use (van Hylckama Vlieg et al, 2009). Oral HRT increases the risk of a first VTE 2- to threefold, whereas VTE risk associated with transdermal HRT is lower [odds ratio (OR) 1·2, 95% CI 0·9–1·7] (Canonico et al, 2008; Farmer et al, 1997; Lidegaard et al, 2009; van Hylckama Vlieg & Middeldorp, 2011). Obviously, oestrogen therapy can be discontinued and, as such, is a transient risk factor. Although the recurrence risk after oestrogen-provoked VTE is clearly higher than after VTE provoked by surgery, most studies suggest that the risk of recurrent VTE is lower in women with an oestrogen-provoked VTE when compared to women with an unprovoked VTE, with relative risks (RRs) ranging from 0·3 to 1·2 (Badaracco & Vessey, 1974; Baglin et al, 2003; Kyrle et al, 2004; Christiansen et al, 2005; Cushman et al, 2006; Le Gal et al, 2010; Douketis et al, 2011).
Recurrence risk after VTE provoked by cancer
The most common persistent clinical risk factor for VTE is cancer. The cumulative incidence of recurrence 12 months after the first VTE was 20·7% in 181 cancer patients with VTE, of whom 44% were treated for 6 months and 56% were treated for longer than 6 months (Prandoni et al, 2002a). Several other observational studies suggest that the risk of recurrence in cancer patients after stopping anticoagulant therapy is unacceptably high (Prandoni et al, 1996; Heit et al, 2000b; Murin et al, 2002; Palareti et al, 2002). Furthermore, in a randomized trial in cancer patients with VTE comparing treatment with VKAs with low molecular weight heparin (LMWH), 80 of 672 (12%) patients experienced a recurrence within 6 months after diagnosis while receiving anticoagulant therapy (Lee et al, 2003).
Other risk factors
Although obesity, autoimmune disorders and active inflammatory diseases, such as Crohn's disease or ulcerative colitis, are associated with an increased risk of first VTE, their influence on the risk of recurrent VTE is less well studied (Bernstein et al, 2001; Miehsler et al, 2004; Pomp et al, 2007; Zoller et al, 2012). In 116 patients with a first VTE and inflammatory bowel disease (IBD) the risk of recurrence 5 years after discontinuation of anticoagulation therapy was 2·5 times higher (95% CI 1·4–4·2) than in a cohort of 1255 patients with a first unprovoked VTE (Novacek et al, 2010). Neurological disease with extremity paresis is an independent predictor of recurrent VTE (HR 2·02, 95% CI 1·24–3·27) (Heit et al, 2000b).
Risk of recurrence after unprovoked VTE
In approximately half of all patients, no risk factor can be identified and the episode is classified as unprovoked (Naess et al, 2007). As mentioned in the previous paragraphs, the risk of recurrence is clearly higher after unprovoked VTE compared to provoked VTE. The risk of recurrence after unprovoked VTE is around 10 per 100 patient years in the first 6–12 months after cessation of anticoagulants and decreases to around 5 per 100 patient years thereafter (Douketis et al, 2011; Prandoni et al, 2007; van Dongen et al, 2003). Patient characteristics and diagnostic findings may help to identify those patients at a particularly high or low risk of recurrence after a first unprovoked VTE. However, none of the discussed risk factors for recurrent VTE have been established enough to guide treatment duration (Kearon et al, 2012).
Men have a higher risk of recurrent VTE than women. In a meta-analysis that included 5416 patients, the 1·6-fold increased risk (95% CI 1·2–2·0) appeared to be irrespective of whether the initial event was provoked or unprovoked and hence, could not be explained by a lower risk of recurrence in women after a first pregnancy- or oestrogen use-related VTE (McRae et al, 2006; Douketis et al, 2011). In contrast, in a study of thrombophilic families, the lower age of women at the time of their first VTE due to oestrogen use or pregnancy resulted in a much longer interval between a first episode and recurrence in women (Lijfering et al, 2009). The 1·6 (95% CI 1·3–2·0) RR of recurrence for men decreased to 1·2 (95% CI 0·8–1·7) after exclusion of provoked events from the analysis.
Location of venous thrombosis
Patients with isolated distal DVT have a lower risk of recurrence compared to those with proximal DVT (4·7 recurrent episodes per 100 patient years; HR 0·49, 95% CI 0·34–0·71) (Boutitie et al, 2011). The annual incidence of recurrent VTE after DVT located in an upper extremity is estimated to be 2–4% (Martinelli et al, 2004; Flinterman et al, 2008).
In patients who presented with PE, the risk of recurrent VTE did not clearly differ from those with proximal DVT; HR 0·85 (95% CI, 0·66–1·10) (Baglin et al, 2010). However, the risk of recurrence as PE is threefold higher for patients presenting with PE compared to patients presenting with proximal DVT (95% CI 1·9–5·1) (Baglin et al, 2010). This finding is potentially relevant because the case-fatality was higher for recurrent PE than for recurrent DVT in an older study (26·4% vs. 5·1%) (Douketis et al, 1998). However, the overall case-fatality rate for recurrent VTE appears to have decreased in the last decades (3·6%)(Carrier et al, 2010). This may reflect an increased awareness of doctors and patients regarding the risk of recurrent VTE, which may lead to earlier diagnostic testing and treatment that would reduce mortality. It is therefore uncertain if the presentation (PE or DVT) of the first VTE is relevant.
Inherited thrombophilia and antiphospholipid syndrome
Inherited thrombophilias increase the risk of a first episode of VTE by 3- to 10-fold (Middeldorp, 2011a). However, the risk of recurrent VTE is only slightly increased compared to patients without thrombophilic defects (RR 1·4–1·5) (Ho et al, 2006; Middeldorp, 2011a). The risk of recurrent VTE in patients with antiphospholipid syndrome appears much higher than for inherited thrombophilic defects, with estimates ranging from 10% to 70% per year. However, these estimates are based on studies with methodological limitations and incorrect definitions for the syndrome (Schulman et al, 1998; Kearon et al, 1999; Lim et al, 2006).
d-dimer levels after discontinuation of anticoagulant therapy
Increased d-dimer levels might reflect an ongoing hypercoagulable state. In a systematic review, patients with a normal d-dimer level, measured after discontinuation of anticoagulant therapy, after a first episode of unprovoked VTE had a 3·5% annual risk of recurrence, compared to 8·9% in patients with elevated post-treatment d-dimer levels (Verhovsek et al, 2008). A management study was performed, in which patients with a normal d-dimer, measured 1 month after cessation of anticoagulation, refrained from further use, whereas patients with an elevated d-dimer level were randomized between discontinuation and resumption of anticoagulant treatment (Palareti et al, 2006). After an average follow-up of 1·4 years, patients with a high d-dimer level who did not use anticoagulant therapy had a risk of recurrent VTE of 15%, which was reduced to 2·9% by resumption of anticoagulant therapy. Patients with a normal d-dimer level had a 6·2% risk of recurrence. A randomized controlled trial to investigate the effect of d-dimer testing with subsequent prolonged anticoagulant treatment on recurrent VTE has not been performed.
Residual thrombosis and post-thrombotic syndrome
Persistence of non-compressibility of the vein after an episode of DVT, i.e. residual thrombosis, is present in approximately one-quarter of patients after 3 years, but is not associated with an increase in risk of recurrent VTE in patients with unprovoked DVT following discontinuation of anticoagulant therapy (OR 1·24, 95% CI 0·9–1·7) (Prandoni et al, 2002b; Carrier et al, 2011).
PTS develops in approximately 25% of patients with symptomatic DVT within 2 years, despite adequate treatment of DVT and use of compression stockings (Prandoni et al, 1996, 2004; Brandjes et al, 1997). The cumulative probability of recurrent VTE was 7·4% among patients with PTS (clinically evaluated at or after 18 months after diagnosis of the index DVT), compared to 1·6% among patients without PTS (RR 2·6, 95% CI 1·2–5·9) (Stain et al, 2005). In a prospective study that aimed to identify patients with unprovoked VTE at low risk of recurrence, the presence of at least one leg sign (hyperpigmentation, oedema or redness of either leg) combined with another risk factor was associated with an increased risk of recurrence in women (Rodger et al, 2008).
Risk of recurrence after recurrent VTE
In patients with a second episode of VTE, the subsequent risk of recurrence after discontinuation of anticoagulant therapy is not well studied. It is estimated to be 50% higher than the risk of recurrence after a first episode (Murin et al, 2002; Schulman et al, 2003). Although it is plausible to assume that this risk increase is lower after two provoked episodes of VTE compared to two unprovoked episodes, there are no data to support this.
Risk of bleeding during anticoagulant therapy
Use of VKAs increases the risk of bleeding (Ageno et al, 2012). In a meta-analysis including 4374 patient-years of VKA use for VTE, the rate of major bleeding was 2·1% during the initial 3 months of anticoagulation. After the first 3 months, this rate was 2·7 per 100 patient-years (Linkins et al, 2003). The case-fatality rate of major bleeding for all patients was 13·4%, and 9·1% if the analysis was limited to patients receiving anticoagulant therapy for more than 3 months (Linkins et al, 2003).
Identification of patients at increased risk of bleeding is important in the decision to extend the duration of anticoagulant therapy. Several patient characteristics, anticoagulant intensity, poor anticoagulant control and duration of therapy are determinants of bleeding (Ageno et al, 2012). Ageing has consistently been shown to increase the risk of bleeding (Hutten et al, 1999). In a cohort of patients treated with VKAs for various indications, every 10-year age increment inflated the risk of major bleeding by approximately 1·5-fold (van der Meer et al, 1996). Concomitant use of antiplatelet agents, such as aspirin and clopidogrel, increases the risk of major bleeding by 1·6% per year with a RR of intracranial bleeding of 2·4 (95% CI, 1·2–4·8) compared to VKA monotherapy (Hart et al, 1999; Ageno et al, 2012). A history of bleeding and comorbidities, such as cancer, hypertension, cerebrovascular disease such as prior stroke, heart disease, diabetes, renal insufficiency, alcoholism and liver disease, also are associated with increased risk of bleeding during treatment with VKAs (Ageno et al, 2012).
In clinical trials for treatment and secondary prevention of VTE and for stroke prevention in atrial fibrillation, the novel oral anticoagulants dabigatran, rivaroxaban or apixaban with VKA showed a reduced risk of major bleeding, especially intracranial haemorrhage, compared to VKA (Connolly et al, 2009; Schulman et al, 2009; Bauersachs et al, 2010; Patel et al, 2011). However, the patient characteristics associated with increased bleeding are likely to be similar for these new drugs and the bleeding risks of long-term treatment in daily clinical practice, especially for patients with VTE, have not yet been established.
Weighing the risks and benefits of anticoagulant treatment for VTE
In the next paragraphs, we discuss the evidence on the duration of anticoagulant VTE in various patient groups and make recommendations, which are also summarized in Table 1.
Table 1. Recommended duration of anticoagulant therapy
Limited duration of treatment after a first episode of VTE provoked by a transient risk factor
Various treatment durations after a first episode of VTE were compared in an individual patient data meta-analysis of seven contemporary studies (Boutitie et al, 2011). The risk of recurrent VTE was higher in patients who had been treated for 1 or 1·5 months compared to patients who were treated for a longer period (HR 1·52, 95% CI 1·14–2·02). The risk of recurrent VTE was not significantly higher in the 3-month group, as compared to longer durations of treatment (6, 12 or 27 months), indicating that there is no benefit of extended treatment in terms of recurrence, once it is stopped (Middeldorp, 2011b).
If a first episode of VTE is provoked by a transient risk factor, including non-surgical and hormonal risk factors, the risk of recurrent VTE is clearly lower than the bleeding risks associated with long-term anticoagulation. Hence a limited duration of anticoagulant therapy is strongly recommended (Middeldorp, 2011b; Baglin et al, 2012; Kearon et al, 2012).
Given a low absolute risk of recurrence, a limited duration of anticoagulation therapy is also recommended after isolated distal DVT of the leg or DVT located in an upper extremity, regardless of whether the first episode was provoked or unprovoked.
Extended duration of treatment after VTE in patients with cancer
Because of the high risk of recurrent VTE, extended treatment is recommended for patients with VTE and cancer (Kearon et al, 2012). In the first 6 months after diagnosis, LMWH is recommended, based on the superior efficacy of LMWH as compared to VKAs (Lee et al, 2003; Kearon et al, 2012). As follow-up studies of a beneficial effect after 6 months of one regimen over the other are yet to be awaited, ACCP guidelines suggest that treatment should continue with the same anticoagulant chosen for the initial period (Kearon et al, 2012). In clinical practice, patients have a preference for either injections or VKAs with monitoring of the International Normalized Ratio, and the decision to switch depends heavily on this. After successful treatment of cancer, the risk of recurrent VTE probably decreases and we therefore prefer to discontinue anticoagulant treatment 6 months after assumed cure.
Duration of treatment after recurrent VTE
After a second or subsequent episode of idiopathic VTE, we recommend extended treatment in accordance with current guidelines (Kearon et al, 2012). In case of two episodes of VTE provoked by a transient risk factor, data to guide management is largely absent. We prefer to individually counsel the patient, taking into account both patient preferences and the specific circumstances of the VTE episodes.
Duration of treatment after unprovoked VTE: how long should extended duration last?
Indefinite duration of VKA at various intensities has been compared to durations of 3 or 6 months (Schulman et al, 1997; Kearon et al, 1999; Ridker et al, 2003; Farraj, 2004; Palareti et al, 2006). Results indicate an approximate 90% reduction of recurrent VTE in those patients randomized to indefinite treatment. Extended treatment with the oral factor Xa inhibitor rivaroxaban or the thrombin inhibitor dabigatran etexilate after 6 or 12 months of treatment for acute VTE resulted in an 82% and 92% RR reduction of recurrence compared to placebo (Bauersachs et al, 2010; Schulman et al, 2011).
The ACCP guidelines suggest extended anticoagulant therapy over a 3-month course for all patients with a first unprovoked episode of venous thromboembolism and a low or moderate risk of bleeding (Kearon et al, 2012). The strength of recommendation is graded 2B, which means that it is weak and is based on moderate-quality evidence, with benefits being closely balanced with risks and burden (Guyatt et al, 2012). Furthermore, regular review (e.g. annually) is proposed to ensure that patients have not developed contraindications to extended therapy, have not changed their preferences, might benefit from improved ways of selecting a patients for extended therapy, and that they are being treated with the anticoagulant regiment that best suits them (Kearon et al, 2012). Hence, extended duration of treatment is not synonymous to indefinite treatment, and it is questionable whether the available evidence is able to guide us on the very long term after an episode of VTE. This is also clear from a recent guidance paper published by the Scientific and Standardization Committee of the International Society on Thrombosis and Haemostasis (Baglin et al, 2012).
It is important to note that follow-up in cohort studies and randomized controlled trials in patients with VTE generally did not exceed 5 years and thus estimates of long-term risk of recurrence are less certain. As detailed above, the bleeding risks of anticoagulant therapy are well known and increase with age. In a systematic review, the rate of recurrent fatal VTE was 0·4% (95% CI 0·3–0·6%) during the initial 3 months of anticoagulation, with a case-fatality rate of 11·3% (95% CI 8·0–15·2%) (Carrier et al, 2010). After discontinuation of anticoagulant therapy, the rate of fatal recurrent VTE was 0·3 per 100 patient-years (95% CI 0·1–0·4%), with a case-fatality rate of 3·6% (95% CI 1·9–5·7%) (Carrier et al, 2010). Hence, if we consider a typical 60-year old patient with a first unprovoked VTE, the annual risk estimate of recurrent VTE after discontinuation of a limited duration (i.e. 3 months) of anticoagulant therapy is reasonably well-known for 2–5 years after his index event. The current recommendation is to extend anticoagulant treatment. However, after 5–10 years and every decade thereafter, estimates of the risk for recurrent VTE become less certain, whereas the bleeding risk clearly increases with age (RR 1·5 for every decade increase (van der Meer et al, 1996). This is conceptually depicted in Fig 1A. Likewise, the risk of dying from a fatal VTE or fatal bleed is depicted in Fig 1B. Given that a benefit of indefinite anticoagulant therapy in terms of mortality has not been shown over the long term, discontinuing anticoagulation therapy is therefore justified.
We propose a 3- to 6-month course of anticoagulant therapy in most patients with unprovoked VTE. A trend towards more recurrences after a 3-month course compared to 6 months (HR 1·39, 95% CI 0·96–2·01) is a valid argument in favour of a 6-month regimen (Boutitie et al, 2011). We consider extended treatment in individual cases in which the pros and cons of extended treatment are discussed with the patient. This approach takes into account personal health evaluations (Locadia et al, 2004) and may, for example, include a severe PTS or limited cardiopulmonary reserve. Like others, we have not implemented indefinite treatment duration in all patients with an unprovoked proximal DVT or PE in our clinical practice (Donadini & Ageno, 2011).
The risk of recurrence after a first VTE is variable, depending on circumstances associated with the first episode. The recurrence rate is low after VTE provoked by transient clinical risk factors and treatment for a limited duration of 3 months is strongly recommended (Kearon et al, 2012). After a first unprovoked VTE the risk of recurrence is higher and ACCP guidelines suggest extended anticoagulant therapy beyond 3 months, unless the bleeding risk is high (Kearon et al, 2012). Individuals at high risk of recurrence or bleeding are not easy to discern from those who will most likely not suffer from recurrence or bleeding. Moreover, follow up durations for recurrent VTE in patients off anticoagulant therapy are limited. While the annual risk of recurrence after 5–10 years is uncertain, the risk of bleeding associated with anticoagulant treatment is likely to increase with age. Extended treatment is not synonymous with indefinite treatment. Given that indefinite treatment has not been shown to benefit patients in terms of mortality, we consider in dubio abstine justified. For most patients with unprovoked VTE we propose 3–6 months of anticoagulation, and only extend this in those with an informed preference to continue.