Monthly incidence (bars) and 95% confidence intervals (error brackets) of recurrent venous thromboembolism after treatment with vitamin K (phytonadione) antagonists for different durations in relation to the time since index event.
van Dongen CJJ, Vink R, Hutten BA, Büller HR, Prins MH. The Incidence of Recurrent Venous Thromboembolism After Treatment With Vitamin K Antagonists in Relation to Time Since First EventA Meta-analysis. Arch Intern Med. 2003;163(11):1285-1293. doi:10.1001/archinte.163.11.1285
After a first episode of venous thromboembolism, patients are treated with vitamin K (phytonadione) antagonists. There are indications that the risk of recurrence after treatment with vitamin K antagonists decreases relative to the time since the first event. The aim of the present meta-analysis is to describe the risk of recurrent venous thromboembolism after treatment with vitamin K antagonist in relation to the time since the index event.
Computerized searches in MEDLINE and EMBASE databases; reference checks of pertinent articles; personal communication with colleagues to find randomized clinical trials and cohort studies in which patients with venous thromboembolism were treated with vitamin K antagonists. Per treatment arm, 2 reviewers independently extracted data on the number of recurrent events and the duration of follow-up per time period of 3 months.
A total of 135 potentially eligible studies were identified. Of these, 18 studies could be included, comprising 25 treatment arms that could be analyzed. Treatment arms were divided into 3 groups based on treatment duration (short, medium, and long). For all 3 groups, the monthly incidence immediately after discontinuation of treatment was high and declined rapidly thereafter. The monthly incidence after 9 months seemed independent of the treatment duration.
There is a diminishing risk of recurrent venous thromboembolism over time and a stabilization after 9 months independent of the duration of the initial treatment with vitamin K antagonists. These findings have important implications for decision making about the optimal duration of treatment with vitamin K antagonists.
PATIENTS WITH venous thromboembolism are initially treated with low-molecular-weight heparin or unfractionated heparin for 5 to 10 days.1,2 Within 48 hours, secondary prophylactic therapy with vitamin K (phytonadione) antagonists is started and continued for 3 to 6 months for most patients. However, there is much debate about the length of this treatment period. In certain subgroups of patients who are thought to be at an increased risk of recurrent events, there is a trend toward a prolonged duration.3 And indeed, continuation of treatment with vitamin K antagonists after 3 months results in a decrease in the risk of recurrent thromboembolic events.4
A retrospective analysis by Van den Belt et al5 suggests that the annual incidence of recurrent venous thromboembolism in patients with a continuous risk factor (familial thrombophilia) is high during the first years and decreases thereafter. However, there is no consensus about the existence and degree of such an effect.6 Clearly, the presence of a diminishing risk over time will affect treatment decisions. If the risk of venous thromboembolism decreases relative to the time since the first event, the benefit of prolongation of vitamin K antagonist treatment would also decrease over time. Since the risk of bleeding related to vitamin K antagonist treatment remains constant, the duration of treatment with vitamin K antagonists after a first thromboembolic event would be highly dependent on the decreasing risk of recurrent thromboembolic events.7
The aim of the present meta-analysis is to gain insight into the natural course of events following a first episode of symptomatic venous thromboembolism, in particular after cessation of treatment with vitamin K antagonists. Therefore, the risk of recurrent thromboembolism is quantified, after a period of treatment with vitamin K antagonists, in relation to time since the first event and the duration of treatment. For this purpose, we conducted a systematic review of published reports on cohorts of patients with documented venous thromboembolism.
A computerized search in MEDLINE (January 1975–July 2001) and EMBASE (January1988–July 2001) was performed to find all studies that addressed the timing of recurrent venous thromboembolism in patients treated with vitamin K antagonists after a first thromboembolic event. The following keywords were used: venous thromboembolism, deep vein thrombosis, pulmonary embolism, oral anticoagulants, vitamin K antagonist, warfarin, dicoumarol, acenocoumarol, phenprocoumon, recurrent, and recurrence. In addition, articles were identified by checking references of pertinent articles and through personal communication with colleagues.
The potential eligible studies were evaluated independently by 2 reviewers and assessed for eligibility based on predetermined criteria. Disagreements were resolved by consensus. If no consensus was reached, the opinion of a third reviewer was decisive. Only randomized clinical trials or cohort studies were included in which the patients were diagnosed with deep vein thrombosis and/or pulmonary embolism and received treatment with vitamin K antagonists. Subsequently, studies were excluded if (1) no data were available about the timing of the recurrent thromboembolic events after cessation of vitamin K antagonist treatment; (2) more than 10% of patients were lost to follow-up; (3) the study was a duplicate report or the cohort was the same as in another included study; (4) other anticoagulant or antiplatelet drugs were used; and/or (5) the diagnosis of venous thromboembolism was objectively confirmed in fewer than 70% of the patients. The diagnosis of the initial and the recurrent episode of venous thromboembolism were considered to be objectively confirmed if the suspected deep vein thrombosis was based on venography or compression ultrasound, if the suspected pulmonary embolism was confirmed by pulmonary angiography or a high-probability ventilation/perfusion lung scan, or when an associated deep vein thrombosis was proven by either venography or compression ultrasound.
All included articles were evaluated by 2 reviewers independently. Disagreements were resolved by consensus. If no consensus was reached, the opinion of a third reviewer was decisive.
Per study, treatment arms were defined by treatment duration. Time elapsed since the index event was categorized into 5 periods (1-3 months, 3-6 months, 6-9 months, 9-12 months, and 12-18 months). Per treatment arm, we extracted the number of recurrent events after treatment with vitamin K antagonists, the number of patients, and the follow-up time. In addition, data on the intensity of treatment with vitamin K antagonist were extracted. Also, information about the number, mean age, sex, and comorbidity of patients included in the studies was recorded.
For each treatment arm separately, a monthly incidence was calculated of recurrent venous thromboembolism per time period after discontinuation of treatment with vitamin K antagonists. This was done by dividing the number of recurrent events by the number of patient-years in that period.
The treatment arms were categorized into 3 groups based on the duration of treatment with vitamin K antagonists. The different durations of treatment were defined as short (4-6 weeks), medium (3 months), and long (4-6 months). For each group, a pooled monthly incidence rate per period of 3 months was calculated by weighing the separate monthly incidence rates by the duration of follow-up for that particular period. Finally, the 95% confidence intervals (CIs) around these estimates were calculated. In addition, an identical analysis was done on the studies in which patients with cancer were excluded. The advisability of combining the trial results was addressed with a statistical test of heterogeneity by means of a Pearson χ2 test (P>.20).
A total of 135 potentially eligible studies were identified. Of these, 17 did not meet the criteria for inclusion. Of the remaining 118 studies,1,2,8- 123 100 had to be excluded. The reasons for exclusion were absence of data on timing of the recurrent events (83 studies),1,2,8- 88 reports on the same cohort (9 studies),89- 97 loss to follow-up exceeding 10% (4 studies),97- 101 lack of confirmation of diagnosis by objective diagnostic methods (3 studies),102- 104 and concomitant use of antiplatelet or other anticoagulant drug (1 study).105 Therefore, a total of 18 studies were available for data extraction,106- 123 some of which used prothrombin time, or Thrombotest,124 to estimate international normalized ratio. Characteristics of these studies are summarized in Table 1.
Of the 18 studies, 16 (89%) were randomized clinical trials,106- 121 one was a combination of a randomized clinical trial and a prospective cohort study,122 and one was a prospective cohort study.123 In 6 studies, the cohort could be subdivided in 2 or more treatment arms based on differences in the duration of treatment with vitamin K antagonists. Hence, the 18 studies resulted in a total of 25 treatment arms that could be evaluated in the meta-analysis, with a total of 3186 patients at risk. In most studies, initial treatment of venous thromboembolism consisted of unfractionated heparin, while low-molecular-weight heparin as initial treatment was used in 3 more recent studies. In 12 studies, deep vein thrombosis was the only indication for inclusion. In the other 6, patients with pulmonary embolism were also included. The composition of patient groups varied considerably per study and is detailed in Table 1.
In general, the target international normalized ratio ranged from 2.0 to 3.0. In 5 studies, an upper limit of more than 3.0 was used. Only 2 of the included studies reported on the adequacy of anticoagulation. In 5 of the 25 treatment arms (710 patients), the duration of treatment with vitamin K antagonists was short; in 14 treatment arms (1718 patients) the duration of treatment was of medium length; while in 6 treatment arms (758 patients) the duration of vitamin K antagonist treatment was long.
For the groups defined by the length of the treatment period (short, medium, and long), the data in relation to time since the first event and after discontinuation of treatment are summarized in Table 2, Table 3, and Table 4. The number of events, the number of patient-years and the monthly incidences of recurrent venous thromboembolism are given for 3-month intervals. The overall incidences and their 95% CIs are also presented in Figure 1.
In all 3 groups the incidence of recurrent venous thromboembolism declined in time since the index event. In the period of 9 to 18 months after the first event, the monthly incidence of recurrent venous thromboembolism was of comparable magnitude for all 3 treatment durations (0.50%-0.71% in the period 9-12 months after the event and 0.42%-0.59% in the period 12-18 months after the event).
The incidence of recurrent venous thromboembolism was highest in the group of patients treated for a long duration in the period 3 to 6 months after the index event (3.96%; 95% CI, 0.47%-14.3%). However, this calculation is based on a small number of observations and is not significantly different from the incidence for that period in the groups for which the duration of treatment was short (1.23%; 95% CI, 0.79%-1.83%) and those for which it was of medium length (1.19%; 95% CI, 0.91%-1.53%). From 3 studies, the monthly incidence in the period from 18 to 21 months could be calculated; it was 0.39%.106,120,123
In 5 studies, only patients without cancer were included.106,109,111,121,122 A summary analysis was only possible for the group of patients who were treated with vitamin K antagonists for a medium time period (3 months). A total of 672 patients could be included in this analysis. The monthly incidence of recurrent venous thromboembolism was 1.59% in the period between 3 and 6 months and declined to 0.58% in the period between 6 and 12 months and to 0.34% in the period between 12 and 18 months after the index event.
The statistical test of heterogeneity was negative (P>.20) for each of the monthly incidences calculated per period, except for patients treated for 3 months in the period 3 to 6 months after the index event (P = .005).
In the present meta-analysis, in which a total of 3186 patients were included, the risk of recurrent thromboembolism in relation to the time since the index event and the duration of vitamin K antagonist treatment was evaluated. It was clearly shown that the monthly incidence of recurrent venous thromboembolism after treatment with vitamin K antagonists decreased over time, while the incidence of recurrent events seemed to stabilize 9 months after the index event, independent of the duration of treatment with vitamin K antagonists. Our findings are consistent with observations in other thrombotic diseases, such as unstable coronary artery disease and myocardial infarction for which the incidence of recurrent episodes also diminishes over time.125,126
Despite the diversity of the patients included in the different studies, all tests for heterogeneity were negative except for the period 3 to 6 months after the index event for the patients who were treated for 3 months. This heterogeneous outcome is mainly due to the study of Kearon et al,106 in which 28% of the included patients had cancer, which resulted in a relatively high number of recurrent venous thromboembolic events. However, the presence of patients with cancer in the other included studies did not account for the rather high monthly incidence of recurrent venous thromboembolism immediately after discontinuation of treatment. An analysis of data from studies in which no patients with cancer were included showed the same trend in recurrent venous thromboembolism.
Ideally, to gain insight into differences in risk of recurrent venous thromboembolism after different durations of treatment, a meta-analysis should be performed that includes randomized clinical trials that are similar with respect to treatment durations and the characteristics of the patients included. Thus, a pooled odds ratio for the comparison between a short and a longer duration could be calculated, which would give a powerful estimate of differences for risk of recurrence between groups. Unfortunately, as far as we know, not many randomized controlled trials have been done that are identical on the treatment duration and the composition of the patients groups. Therefore, we believe that the best method to compare different durations of treatment was to calculate summarized incidences of recurrent venous thromboembolism per time period for the available cohorts.
A relative high monthly incidence (3.96%) of recurrent venous thromboembolic events shortly after the treatment with vitamin K antagonists for a longer period (4 to 6 months) can point to the existence of a "rebound phenomenon," as discussed in several studies.103,127 However, the incidence of recurrent venous thromboembolism for that period was calculated from data of only 2 small studies,113,115 resulting in a large CI, which included the point estimate of the patients treated for 3 months. Thus, it cannot be concluded specifically for patients who have been treated for a duration between 4 and 6 months that the incidence of recurrent events is higher immediately after discontinuation of treatment than for those treated for 3 months only.
In the 3 groups we distinguished, the monthly incidence of recurrent venous thromboembolism was similar and low (0.50%-0.71% per month) 9 months after the index event. It may be argued that there were differences in baseline risk of recurrent venous thromboembolism among the 3 treatment groups. However, in only 2 of the 18 studies, patients were not fully randomized between the distinctive treatment strategies. Therefore, it is unlikely that the baseline risks of recurrences of the 3 groups differed much. Taken together, our analyses indicate that the duration of exposure to vitamin K antagonists does not seem to influence the long-term incidence of recurrent venous thromboembolism. If true, this will profoundly affect our decision making about the optimal duration of treatment with vitamin K antagonists in individual patients with venous thromboembolism.
To what extent do our results give a reliable estimate of the incidence of recurrent venous thromboembolism over time? In the present systematic review, only randomized clinical trials and cohort studies that met predefined quality criteria were included to ensure a valid estimate of the risk of recurrent venous thromboembolic events.128 In addition, more than 70% of the diagnoses of thromboembolic events had to be objectively confirmed; a maximum of 10% of the patients was allowed to be lost to follow-up; and no antiplatelet or other anticoagulant drug was administered to the patients included. Finally, a sufficient number of patients was studied to allow for reasonably narrow CIs. Hence, we believe that the incidences obtained are reliable estimates for the risk of recurrent thromboembolic events after discontinuation of treatment with vitamin K antagonists.
In 12 of the 18 included studies, only patients presenting with deep vein thrombosis were included. It can be argued that the risk of developing recurrent venous thromboembolism is dependent on the type of initial thromboembolic event (pulmonary embolism or deep vein thrombosis). However, it has been shown in a large systematic review129 that the overall risk of a recurrent thromboembolic event in patients presenting with pulmonary embolism does not materially differ from patients presenting with deep vein thrombosis. Therefore, it is not likely that another distribution of the 2 types of initial thromboembolic events will result in other incidence rates.
One limitation of our analyses, in particular for decision making in individual patients, should be addressed. We were not able to distinguish between different subgroups of patients (eg, those with thrombophilia vs idiopathic venous thrombosis vs secondary venous thrombosis). Although these groups likely have different risks for recurrent thrombosis, the diminishing risk over time will remain. When relative risks for recurrence for subgroups of patients become available, the risk for recurrence over time per subgroup can be calculated. However, this will require more prospectively collected information in future studies.
The decreasing risk over time for developing recurrent venous thromboembolism may have important clinical implications. As the risk of vitamin K antagonist–related bleeding does not decrease over time after 3 months from start of therapy130 (but in fact increases with age), the benefit of prolonged treatment with vitamin K antagonists diminishes over time. Therefore, continuation of treatment beyond 3 months should be considered with caution. In the debate about the optimal duration of treatment with vitamin K antagonists, it is also reasonable to take into account the burden of treatment experienced by the patient. It is questionable whether the reduced risk of a recurrent venous thromboembolic event when vitamin K antagonist treatment is prolonged can counterbalance the burden of treatment and the increased risk of bleeding, with its often far-reaching consequences for the quality of life for the patient.
Corresponding author: Carlo J. J. van Dongen, MSc, Department of Clinical Epidemiology and Biostatistics, Room J2-204, Academic Medical Center, University of Amsterdam, PO Box 22700, 1100 DE Amsterdam, the Netherlands (e-mail: email@example.com).
Accepted for publication August 16, 2002.
This study is part of a project sponsored by the Dutch Heart Association, the Hague, the Netherlands.
Preliminary data of this review were presented at the 18th Congress of The International Society on Thrombosis and Haemostasis in Paris, France, July 10, 2001.
We thank Per Lindmarker, MD, for providing additional information.