Intended for healthcare professionals


Secondary prevention of venous thromboembolism

BMJ 2013; 347 doi: (Published 09 September 2013) Cite this as: BMJ 2013;347:f5440
  1. Sam Schulman, professor of medicine12,
  2. James Douketis, professor of medicine2
  1. 1Thrombosis Service, HHS-General Hospital, Hamilton, ON, Canada L8L 2X2
  2. 2Department of Medicine and Thrombosis and Atherosclerosis Research Institute, McMaster University, Hamilton, Canada
  1. schulms{at}

Emerging anticoagulation treatments not more effective but safer than warfarin

The seemingly endless debate on the optimal duration of anticoagulation after unprovoked venous thromboembolism has been accorded another dimension. With the addition of new oral anticoagulants (NOACs) and aspirin being evaluated for extended duration prophylaxis after three to six months of anticoagulant therapy for venous thromboembolism, there are now alternatives to treatment with vitamin K antagonists. But this has come at the cost of increased uncertainty among physicians who are less familiar with the recent trials of different agents and the interpretation of these results. Adding to the murkiness is the paucity of head to head comparisons between the NOACs and comparisons of each compound to vitamin K antagonists, administered to achieve an international normalised ratio (INR) of 2.0 to 3.0, which has been the reference standard. So far, only one such trial has been performed—dabigatran versus warfarin.1 To clarify the relative benefits and harms of emerging treatments for extended prophylaxis against disease recurrence in patients with unprovoked venous thromboembolism, Castellucci and colleagues (doi:10.1136/bmj.f5133) present a network meta-analysis of 13 randomised trials, totalling 12 000 patients.2

Studies in atrial fibrillation have shown that NOACs are at least as efficacious (dabigatran 110 mg twice daily, rivaroxaban 20 mg daily, apixaban 5 mg twice daily)3 4 5 or more efficacious (dabigatran 150 mg twice daily) as warfarin for ischaemic stroke prevention.3 However, warfarin remains the best treatment for preventing recurrent venous thromboembolism after the initial three to six months of treatment, which seems to be supported by Castellucci and colleagues’ findings.

On closer inspection of the findings, warfarin may have had a systemic advantage over the comparator treatments. Firstly, none of the patients in the included studies was naive to treatment with vitamin K antagonist, having received, on average, seven to 12 months of previous treatment in the two largest studies.1 6 This would favour warfarin treated patients, because the risk of major bleeding is highest in the initial few months of treatment,7 in part related to the high variability of INR results.7 Such patients, in whom warfarin treatment can be challenging, would be excluded from anticoagulation treatment of extended duration. This exclusion would result in attrition bias, whereby the remaining patients with more stable INR results are more likely to have favourable outcomes than with alternative treatments. A sensitivity analysis that adjusted for the duration of previous anticoagulation treatment would have been informative.

Secondly, the meta-analysis assessed only patients who participated in randomised trials who tend to be highly motivated, especially when enrolled in double blind trials with sham INRs. Such patients treated with vitamin K antagonists may have spent more time in the therapeutic range, again favouring this treatment arm. A more generalisable comparison of risks and benefits of long term treatment with vitamin K antagonists, NOACs, and acetylsalicylic acid—as the authors acknowledged—would involve “real world” data of unselected patients with venous thromboembolism.

Despite the limitations of pooled data when derived from randomised trials, the advantage provided by this meticulously conducted meta-analysis is the reliability and precision of comparisons based on the available data. This is important for discussions with patients on whether to continue or stop anticoagulation treatment after unprovoked venous thromboembolism. Such decisions typically hinge on comparing a patient’s estimated risk for recurrence (and fatal events) if treatment is stopped with the estimated risk for serious (and fatal) bleeding if treatment is continued, coupled with the patient’s values and preferences. The linked meta-analysis shows that fatal events are infrequent (0.34% for fatal venous thromboembolism, 0.04% for fatal bleeding) and not affected by stopping or continuing treatment with a vitamin K antagonist, NOAC, or acetylsalicylic acid. In other meta-analyses of patients with unprovoked venous thromboembolism, rates of fatal bleeding with continued anticoagulation were 0.2-0.6% and rates of fatal pulmonary embolism after stopping treatment were 0.3-0.5%.8 9 10 Overall, the risk for fatal events does not seem to be a pivotal factor when deciding whether to extend antithrombotic therapy and the choice of treatment if extended treatment is selected.

Another key factor in decisions on extended anticoagulant therapy and the type of anticoagulant used is physicians’ fear of “causing a bleed” if anticoagulation is continued.11 In the linked meta-analysis, acetylsalicylic acid was safer (although with less efficacy) than anticoagulants, and NOACs seemed safer than vitamin K antagonists of standard intensity when compared with placebo (increase of 0.2-0.6 v 1.3 major bleeds per 100 patient years). The exception was rivaroxaban treatment, with 5.7 more major bleeds per 100 patient years; however, the imprecision in this estimate is considerable because of the need to adjust for zero events in one group, which generated a disproportionally high odds ratio of 20.79 versus placebo.

In the only study on rivaroxaban for extended treatment, there were four major bleeding episodes among 602 patients versus no episodes among 594 patients on placebo.12 The similarly designed study with dabigatran versus placebo recorded two versus zero major bleeding episodes, respectively.1 However, with the addition of data from the warfarin controlled study, dabigatran ends up with an odds ratio of 2.79, or 0.6 more events per 100 patient years. Furthermore, in a recent meta-analysis on the NOACs in the initial treatment of venous thromboembolism, rivaroxaban had the lowest risk of bleeding.13 These contradictory results are unlikely to reflect the truth, which is probably somewhere in between. The warfarin controlled study with dabigatran1—where the hazard ratio for major bleeding on dabigatran was 0.52 (95% confidence interval 0.27 to 1.02)—supports the impression that the new anticoagulants in the long run are safer than vitamin K antagonists.

So what do we recommend to our patients? With a similar efficacy for NOACs as with vitamin K antagonists,2 13 extended treatment with anticoagulation treatment has a low risk of major bleeding, as well as increased convenience in the absence of the need for laboratory monitoring and dose adjustments. As a result, more physicians and patients will be likely to accept extended anticoagulation, provided that the risk for recurrent venous thromboembolism was deemed sufficiently high (for example, >5% per year) as determined by individualised clinical factors and patient preferences. The rationale for extended anticoagulation would focus less on preventing apparently infrequent fatal venous thromboembolism, but rather on mitigating the risk for the more common post-thrombotic syndrome (20-50% incidence) and pulmonary thromboembolic hypertension (1-4% incidence). The less efficacious option of acetylsalicylic acid may be reserved as a convenient and less costly option when physicians or their patients decide against long term anticoagulation. Real world data are still needed on the risk of bleeding and evaluations of adherence to anticoagulants in people with venous thromboembolism. Although such data seem promising for patients with atrial fibrillation,14 15 16 the adherence might decrease if the threat of a disabling stroke is replaced by the risk of a recurrent and almost invariably non-fatal event of venous thromboembolism.


Cite this as: BMJ 2013;347:f5440


  • Research, doi:10.1136/bmj.f5133
  • Competing interests: SS has received honoraria for study related work from Boehringer Ingelheim and Bayer Healthcare. JD has received honorariums from Boehringer Ingelheim, Bayer Healthcare, and BMS-Pfizer for consulting or advisory boards.

  • Provenance and peer review: Commissioned; not externally peer reviewed.


View Abstract