Abstract
Current guidelines recommend long-term anticoagulant therapy in patients with unprovoked venous thromboembolism (VTE). The risk of fatal recurrent VTE after treatment discontinuation (versus that of fatal bleeding during anticoagulation) is of particular relevance in the decision to continue or stop anticoagulation after the first 3 months. Our primary aim was to provide a point-estimate of the yearly rate of fatal recurrent VTE and VTE case-fatality rate in patients with unprovoked VTE after anticoagulation cessation. Data were extracted from both randomised controlled trials and observational studies published before May 1, 2017. The pooled fatality rates were calculated using a random-effects model. 18 studies with low-to-moderate bias were included in the primary analysis, totalling 6758 patients with a median (range) follow-up duration of 2.2 (1–5) years. After anticoagulation cessation, the weighted pooled rate of VTE recurrence was 6.3 (95% CI 5.4–7.3) per 100 patient-years and the weighted pooled rate of fatal recurrent VTE was 0.17 (95% CI 0.047–0.33) per 100 patient-years, for a case-fatality rate of 2.6% (95% CI 0.86–5.0). These numbers are a solid benchmark for comparison to the risks associated with long-term anticoagulation treatment for the decision on the optimal duration of treatment of patients with unprovoked VTE.
Abstract
The rate of fatal recurrent VTE after anticoagulation cessation for unprovoked VTE was 0.17 per 100 patient-years http://ow.ly/U1sM30mtbrp
Introduction
The risk of recurrent venous thromboembolism (VTE), which includes deep vein thrombosis (DVT) and pulmonary embolism (PE) persists after cessation of anticoagulant treatment and is particularly high among patients with unprovoked VTE [1–3]. Consequently, treatment guidelines recommend continuation of anticoagulant therapy beyond the first 3 months in patients with unprovoked VTE without high risk for major bleeding [4–6]. This recommendation is based on weighing the risk of recurrent VTE after anticoagulant treatment cessation against the risk of major bleeding during ongoing treatment. For the individual patient, the risk of fatal recurrent VTE versus that of fatal bleeding is of particular relevance when making the decision to prolong treatment or not.
The case-fatality rate of major bleeding events during long-term vitamin K antagonist (VKA) treatment has been estimated to be as high as 9–13%, with a yearly rate of fatal bleeding varying between 0.2% and 1.5% [7, 8]. Importantly, this bleeding risk was found to decrease considerably with the introduction of direct oral anticoagulants (DOACs), which are associated with lower rates of intracranial and fatal bleeding than VKA, while non-inferiority was shown with regard to risk of recurrent VTE [9].
The case-fatality rate of recurrent VTE after cessation of anticoagulant therapy has previously been shown to vary between 3.6% and 5.1% in a mixed cohort of patients with both provoked and unprovoked VTE, with a yearly risk of fatal recurrence ranging between 0.4% and 0.5% [7, 10]. To date, these exact numbers are unknown for patients with unprovoked VTE, although this is the patient category for whom this knowledge is most relevant [4, 5]. Therefore, we conducted a systematic review and meta-analysis of the literature to provide an accurate point-estimate of the case-fatality rate of recurrent VTE as well as a yearly rate of fatal recurrences after anticoagulation cessation in patients with a first unprovoked VTE.
Methods
Data sources and literature search
A systematic literature search was conducted for all relevant publications in PubMed, Embase, Web of Science and the Cochrane Library in May 2017. The subject headings and/or keywords of our search strategy comprised “venous thromboembolism”, “pulmonary embolism”, “deep venous thrombosis”, “anticoagulation” and “recurrence” and were database-specifically translated (online supplementary material).
Study selection and data extraction
Initial results were screened for relevant titles and abstracts by two independent reviewers (SJ and LM). This process was performed in duplicate and disagreements were independently resolved by consensus or by a third reviewer (FA). Studies were included if 1) consecutive patients with objectively confirmed symptomatic DVT or PE were prospectively enrolled (proximal DVT diagnosed in case of evidence of thrombosis in the popliteal or more proximal veins on compression ultrasonography or contrast venography and a diagnosis of PE based on at least one subsegmental filling defect on computed tomography pulmonary angiography (CTPA), high-probability ventilation/perfusion lung scan (V′/Q′) or abnormal pulmonary angiography; 2) patients were subject to dedicated follow-up for symptomatic recurrent VTE and such events were objectively confirmed; 3) the initial anticoagulation treatment (with VKA or DOAC) was continued for ≥3 months and the follow-up period extended for ≥3 months after the anticoagulation therapy was discontinued; 4) fatal VTE events during follow-up after treatment cessation were reported (PE and/or DVT); and 5) ≥100 patients were included. Only full-text publications in the English language were reviewed for potential inclusion. There was no restriction on publication year.
After selection of all relevant articles, two reviewers (SJ and LM) independently extracted data on first author's name, year of publication, design (prospective/retrospective), number of patients included, age, initial anticoagulation treatment, the total duration of follow-up after cessation of treatment, proportion of unprovoked VTE at baseline (PE/DVT), case-fatality rate of recurrent VTE during follow-up after anticoagulant discontinuation (PE/DVT) and overall mortality during follow-up, as reported by the authors. The authors of publications with missing data were approached by email at least twice, 2 weeks apart. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement for reporting systematic reviews and meta-analysis was used for this study [11].
Study objectives
The primary objective was to determine the case-fatality rate of recurrent VTE after anticoagulation cessation following a first unprovoked VTE diagnosis, as well as the yearly rate of fatal VTE recurrences from selected studies with low to moderate bias. The secondary aims were to determine the overall rate of fatal VTE for all available studies, including those with a high risk of bias, and to differentiate between 1) enrolment periods, comparing studies that started enrolment before and after January 1, 2000 (if reported); 2) cohort studies and randomised controlled trials (RCTs); 3) studies with a follow-up duration that was shorter versus longer than 2.5 years; 4) patients who initially presented with DVT versus unprovoked PE; and 5) different definitions of fatal VTE that were applied.
Study outcomes and definitions
Recurrent PE was predefined as a new intraluminal filling defect on pulmonary angiography or CTPA, a new high-probability perfusion defect on V′/Q′ scan or any new defects after earlier normalisation of the scan [12]. Recurrent DVT was defined as new noncompressibility by ultrasonography of the common femoral and/or popliteal vein, noncompressibility of a previously normalised vein segment or a pronounced increase in vein diameter (≥4 mm) of a previously noncompressible venous segment [12]. Patients with both index DVT and PE were classified as patients with PE when fatal rates were reported separately for this subgroup. Fatal recurrent VTE was predefined as PE diagnosed by autopsy, high-probability V′/Q′ scan, a new intraluminal filling defect detected on pulmonary angiography, computed tomography (CTPA) or venography prior to death or a high clinical suspicion as judged by the investigators of the individual studies. For each study, the definition of unprovoked VTE was evaluated post hoc and compared to criteria provided by the scientific and standardisation committee of the International Society on Thrombosis and Haemostasis (ISTH) [13].
Risk of bias
Two authors (SJ and LM) independently evaluated the risk of bias at a study level in accordance with the Cochrane Collaboration's tool for assessing risk of bias and the PRISMA statement [11, 14]. We focused on the following criteria. 1) Prespecified protocol; 2) clear description of inclusion and exclusion criteria; 3) adequate anticoagulation treatment prior to cessation according to international standards; 4) clear description of follow-up after anticoagulation cessation; 5) clear definitions provided of unprovoked and fatal VTE; 6) loss to follow-up; 7) adjudication of outcomes; and 8) assessment of primary end-point in all patients. Disagreements were resolved through discussion with a third author (FA).
Statistical analyses
Case-fatality rates of each study were calculated by dividing the number of recurrent fatal VTEs by all recurrent VTEs. The case-fatality rates were pooled after Freeman–Tukey double arcsine transformation to stabilise variances, using a random effects model according to the method of DerSimonian and Laird [15, 16]. Pooled case-fatality rates were reported with corresponding 95% confidence intervals. Subsequently, we estimated the rate of recurrent VTE and fatal recurrent VTE per 100 patient-years. We assessed statistical heterogeneity of exposure effects across the various cohort studies by calculating the I2 statistic, which depicts the variance of results from study to study beyond (or rather than) chance. Heterogeneity was considered low when I2 was <25%, intermediate when I2 was 25–75% and high when I2 was >75% [17]. Heterogeneity was explored using meta-regression. We evaluated differences across subgroups under the null hypothesis of no differences (Chi-squared distribution with S (number of subgroups) minus 1 degrees of freedom). All analyses were performed using Stata 14.0 (StataCorp, College Station, TX, USA).
Results
Literature search and study selection
The initial search yielded 7647 potentially relevant articles; 586 in Cochrane, 2839 in Embase, 2307 in PubMed and 1915 in Web of Science (figure 1). After the first screening of title and abstract, 7540 records were excluded, leaving 107 unique articles for detailed assessment. An additional 83 articles were excluded after full review for the following reasons: 28 were abstracts only, with insufficient information, 27 comprised studies of duplicate patients with other reports, eight did not clarify if fatal events were on or off anticoagulation treatment, eight had fewer than 100 patients, four did not distinguish between provoked and unprovoked VTE and authors did not comply with our data request after at least two attempts, and eight were excluded for other reasons. The remaining 24 articles all satisfied our predetermined methodological criteria [18–41].
Flow chart of the clinical search. VTE: venous thromboembolism.
Included studies
Table 1 shows the characteristics of the included studies. 15 were cohort studies [21, 23–26, 28, 29, 32, 33, 35, 37–41] and nine were RCTs [18–20, 22, 27, 30, 31, 34, 36]. The 24 articles were published between 1995 and 2017 and included a total of 8914 patients with unprovoked VTE (range 117–914 patients per study). The median follow-up duration after treatment cessation was 2.5 years (range 1–7.7 years). The evaluation of the presence of bias is shown in table 2. Of the 24 studies, 18 were considered to be at low or moderate risk of bias and were included in the primary analysis. Five studies did not involve an independent adjudication committee [24, 28, 32, 33, 40]. Most studies did not meet the criteria of the ISTH definition of unprovoked VTE [20, 21, 24, 26, 28–30, 32, 34–39, 41]. One study did not provide any definition of unprovoked VTE [22].
Characteristics and outcomes of the included studies
Evaluation of presence of bias for all 24 identified relevant studies
Primary outcome: rate of fatal recurrent VTE in studies with low or moderate risk of bias
The 18 studies with low or moderate risk of bias enrolled a total of 6758 patients with a median follow-up of 2.2 years (range 133–914 years). Table 3 shows the rates of recurrent VTE and fatal recurrent VTE per subgroup. The weighted pooled rate of recurrent VTE in studies with low or moderate risk of bias was 6.3 (95% CI 5.4–7.3, I2=72.6%) per 100 patient-years and the rate of fatal recurrent VTE was 0.17 (95% CI 0.047–0.33, I2=83.57%) per 100 patient-years, for a case-fatality rate of 2.6% (95% CI 0.86–5.0, I2=66.6%; figure 2).
Fatal venous thromboembolism (VTE) rates per subgroup
Meta-analysis of case-fatality incidences after anticoagulant cessation in studies with low-to-moderate risk of bias. CFR: case-fatality rate; VTE: venous thromboembolism.
Secondary outcomes
The overall weighted pooled fatal rate of VTE recurrence among all 24 studies was 6.2 (95% CI 5.4–7.2, I2=86.8%) per 100 patient-years and the rate of fatal recurrent VTE was 0.13 (95% CI 0.036–0.25, I2=72.7%) per 100 patient-years, for a case-fatality rate of 2.0% (95% CI 0.69–3.8, I2: 65.2%; online supplementary figure S1).
Studies that initiated enrolment before the year 2000 had a numerically higher, but not significantly different, pooled rate of fatal VTE than studies that started inclusion within or after the year 2000 (0.27, 95% CI 0.038–0.59; I2=83.1 versus 0.039, 95% CI 0.0028–0.1 per 100 patient-years; I2=0; p=0.70 for interaction), as well as case-fatality rate (3.7%, 95% CI 0.95–7.6%; I2=76.5 versus 0.71%, 95% CI 0.063–1.8%; I2=0; p=0.21 for interaction; online supplementary figure S2). Notably, the analysis of the more recent studies showed good homogeneity (both I2=0) while the results of earlier studies were quite heterogeneous (I2>75). The rate of fatal recurrent VTE was similar in cohort and RCT studies (0.11, 95% CI 0.009–0.29; I2=79.5% versus 0.14, 95% CI 0.021–0.33; I2=49.7% per 100 patient-years; p=0.96 for interaction) and studies with shorter and longer than 2.5 years follow-up duration (0.11, 95% CI 0.018–0.27; I2=52.4% versus 0.13, 95% CI 0.076–0.35; I2=81.7% per 100 patient-years; p=0.94 for interaction). Likewise, the case-fatality rates did not differ for cohort and RCT studies (1.7%, 95% CI 0.19–4.2%; I2=74.6% versus 2.5%, 95% CI 0.69–5.0%; I2=26.8%; p=0.87 for interaction) as well as for studies with shorter and longer than 2.5 years follow-up duration (2.2%, 95% CI 0.22–5.4%; I2=76.6% versus 1.8%, 95% CI 0.46–3.8%; I2=34.9%; p=0.69 for interaction).
In 19 studies, fatal recurrent VTE could be distinguished for patients initially presenting with DVT versus PE [13, 18, 19, 23–25, 27–39, 41]. The case-fatality rates of patients initially presenting with DVT and PE were 2.3% (95% CI 0.52–4.8%, I2=60.39%) and 0.12% (95% CI 0–1.8%, I2=34.9%; p=0.57 for interaction; online supplementary figure S3). When focusing on studies with low or moderate risk of bias only, this numerical difference decreased considerably (2.7%, 95% CI 0.50–6.1%; I2=63.52% versus 1.6%, 95% CI 0–5.7%; I2=48.43%; p=0.66 for interaction; online supplementary figure S4).
Fatal VTE definition
The definition of fatal VTE varied widely across studies (online supplementary table S1). Only 12 (54%) studies reported a definition of fatal VTE [18, 19, 22, 24, 25, 27, 30, 31, 34, 36–38], of which 11 (92%) included autopsy and/or clinical suspicion [18, 19, 22, 24, 25, 27, 30, 31, 36–38] and five (42%) involved “sudden unexplained death” [25, 27, 34, 36, 37]. Studies including “sudden unexplained death” in their fatal VTE definition were found to have the highest case-fatality rates (3.6%, 95% CI 0.018–11%; I2=81.15%), while studies without a clear definition of fatal recurrent VTE reported the lowest rates (0.95%, 95% CI 0.067–2.5%; I2=27.06%; p=0.29 for interaction; online supplementary table S2). This difference in case-fatality rates was observed in both index PE and index DVT patients.
Discussion
In this systematic review and meta-analysis, we determined the risk of fatal recurrent VTE in patients with unprovoked VTE after cessation of anticoagulation treatment. We observed a pooled rate of fatal recurrent VTE of 0.17 per 100 patient-years with a case-fatality rate of 2.6% in studies with low to moderate risk of bias. Where most meta-analyses performed in our study showed relevant heterogeneity among the included studies, the secondary analysis focussing on more recent studies (patient enrolment after January 1, 2000; total of 4508 patients) showed good homogeneity. The numerically lower pooled rate of fatal recurrence (0.039 per 100 patient-years) and case-fatality rate (0.71%) found in this subanalysis may be explained by improved patient care over the years, earlier presentation at the hospital or detection of smaller and less dangerous PE blood clots by more advanced CTPA technology.
The present study revealed similar rates of fatal recurrent VTE in cohort studies compared to RCTs, thus supporting the external validity of our findings. The fatal rates of studies with longer and shorter follow-up durations did not differ as well, indicating that our main finding is valid for long-term follow-up (at least beyond the first 2 years after treatment continuation). Furthermore, we use the finding of a lower rate of fatal recurrent VTE in more recent studies as an argument to hypothesise that the identified rates in our main analysis represent an overestimation of the “true” risk rather than an underestimation. Therefore, our findings provide clinicians, guidelines committees, investigators and policymakers with a solid and valid benchmark of the mortality risk due to recurrent VTE after cessation of treatment to be compared with the risks associated with long-term anticoagulation treatment for patients with unprovoked VTE [4, 5]. Importantly, since risk of VTE recurrence changes over time with the bulk of recurrences occurring in the first years, and the risk of bleeding remains more stable, the ultimate answer to the question of the most optimal duration of anticoagulation for unprovoked VTE is to be determined in future RCTs with long-term follow-up.
We found a nonsignificant higher risk of fatal recurrent VTE after an index DVT diagnosis than after an index PE diagnosis, which was unexpected [7, 10]. This difference is mostly explained by biases of the data pooling due to major methodological differences between the included studies. Other explanations may be that PE is often overdiagnosed due to adoption of increasingly advanced CT technology [42]. In addition, a selection of “healthier” PE patients for whom anticoagulation discontinuation was deemed to be safe in observational studies could have contributed to the lower observed fatal rates of recurrent VTE. Lastly, many of the patients with DVT may actually have had PE as well, although this was not objectively confirmed and therefore nor reported in the original study publications.
Remarkably, the reported rate of fatal recurrent VTE was largely dependent on the definition adopted across the various studies. Overall, studies without a clear definition reported the lowest rates, while studies in which unexplained death was adjudicated as recurrent VTE showed the highest rates. Half of the included studies did not report a definition of fatal VTE, whereas the remaining studies used various definitions ranging from autopsy findings alone to “sudden unexplained death”. With no widely accepted definition of “fatal VTE”, it is impossible to rank these different definitions, although it seems reasonable to assume that studies focussing on autopsy findings may provide underestimated rates of fatal recurrent VTE, while the opposite is true for studies adjudicating all unexplained death as being provoked by recurrent VTE. Moreover, the adjudication process itself might also be difficult and could possibly lead to different rates of PE-related deaths among studies. Our findings thus urgently call for an effort to standardise this definition for future studies in order to allow for valid interstudy comparisons [43].
Current guideline recommendations with regard to extended duration of treatment after unprovoked VTE will be confirmed beyond doubt if these studies show that long-term treatment with DOACs is, indeed, associated with a yearly rate of fatal bleeding <0.047–0.33%. Until then, anticoagulation duration should be individualised based on a patient-specific balance between bleeding and recurrent thrombotic risk. Valid bleeding and thrombotic risk tools have been developed and, although not validated in RCTs, could be helpful to assess these risks and thereby identify patients who may benefit from short- or long-term anticoagulation treatment [44–47].
Strong points of this analysis include the strict selection criteria applied and the large number of patients studied. Source data were only derived from high-quality studies. Moreover, we were able to compare fatal rates in four relevant subgroups. Our study has several limitations in addition to the issue of varying definitions of fatal recurrent VTE. In particular, we did not have the availability of patient-level data, which would have allowed us to evaluate the prognostic role of risk factors such as age and sex. In addition, although we performed rigorous inclusion criteria and focused only on high-quality studies, the meta-analyses presented were subject to relevant heterogeneity caused by several between-study differences, especially for those studies that enrolled patients before January 1, 2000.
Conclusions
This meta-analysis revealed a pooled rate of fatal recurrent VTE of 0.17 (95% CI 0.047–0.33) per 100 patient-years for patients with unprovoked VTE after discontinuation of anticoagulation therapy in studies with low to moderate risk of bias. This was consistent with a case-fatality rate of 2.6% (95% CI 0.86–5.0%). Notably, we observed utilisation of varying fatal VTE definitions which was associated with moderate to high between-study heterogeneity, affecting the reported rates of fatal recurrent VTE. Current guideline recommendations on the duration of treatment of unprovoked VTE would be strengthened if future studies show that long-term anticoagulation treatment with DOACs is indeed associated with a rate of fatal bleeding <0.33% per year, representing the upper limit of the 95% confidence interval the pooled incident rate of fatal recurrent VTE after anticoagulation discontinuation.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
supplementary material Supplementary_Manuscript_Fatal_recurrent_VTE_after_anticoagulation_cessation_Resp_J_19-7_marked
Acknowledgements
We would like to acknowledge the help received from Cecilia Becattini (University of Perugia, Perugia, Italy), Tim Brighton (Prince of Wales Hospital, Sydney, Australia), Paul Kyrle (University of Vienna, Vienna, Austria), Lisbeth Eischer (University of Vienna), Daniel Ribeiro (University Hospital Federal de Minas Gerais, Belo Horizonte, Brazil), Jean-Philippe Galanaud (Montpellier University Hospital, Montpellier, France), Jean-Luc Bosson (Grenoble University Hospital, Grenoble, France), Daniela Poli (University of Florence, Florence, Italy) and Laura Young (University of Auckland, Auckland, New Zealand), who contributed to data retrieval for the present analysis.
Footnotes
This article has supplementary material available from err.ersjournals.com
Provenance: Submitted article, peer reviewed.
Author contributions: S.J. van der Wall and F.A. Klok were responsible for the concept and design of the study, quality assessment and interpretation of the results and writing of the manuscript. L.M. van der Pol was responsible for data extraction, quality assessment and critical revision of the manuscript. Y.M. Ende-Verhaar was responsible for statistical analysis and critical revision of the manuscript. S. Schulman, P. Prandoni, M. Rodger, S.C. Cannegieter and M.V. Huisman were responsible for interpretation of the results, and critical revision of the manuscript.
Conflict of interest: S.J. van der Wall has nothing to disclose.
Conflict of interest: L.M. van der Pol has nothing to disclose.
Conflict of interest: Y.M. Ende-Verhaar has nothing to disclose.
Conflict of interest: S.C. Cannegieter has nothing to disclose.
Conflict of interest: S. Schulman has nothing to disclose.
Conflict of interest: P. Prandoni has nothing to disclose.
Conflict of interest: M. Rodger has nothing to disclose.
Conflict of interest: M.V. Huisman reports grants from ZonMW, and grants and personal fees from Pfizer-BMS, Boehringer-Ingelheim and Daiichi-Sankyo, outside the submitted work.
Conflict of interest: F.A. Klok reports grants from Bayer, Bristol-Myers Squibb, Boehringer-Ingelheim, Daiichi-Sankyo, MSD and Actelion, outside the submitted work.
- Received October 7, 2018.
- Accepted October 22, 2018.
- Copyright ©ERS 2018.
ERR articles are open access and distributed under the terms of the Creative Commons Attribution Non-Commercial Licence 4.0.
References
