Article Text
Abstract
Objective To evaluate efficacy and safety of oral anticoagulant regimens and aspirin for extended venous thromboembolism (VTE) treatment.
Methods We searched MEDLINE, Embase, CENTRAL and conference proceedings for randomised controlled trials studying vitamin K antagonists (VKAs), direct oral anticoagulants (DOACs) or aspirin for secondary prevention of VTE beyond 3 months. ORs (95% credible intervals) between treatments were estimated using random-effects Bayesian network meta-analysis.
Results Sixteen studies, totaling more than 22 000 patients, were included. Compared with placebo or observation and with aspirin, respectively, the risk of recurrent VTE was lower with standard-intensity VKAs (0.15 (0.08 to 0.24) and 0.23 (0.09 to 0.54)), low-dose factor Xa inhibitors (0.16 (0.06 to 0.38) and 0.25 (0.09 to 0.66)), standard-dose factor Xa inhibitors (0.17 (0.08 to 0.33) and 0.27 (0.11 to 0.65)) and the direct thrombin inhibitor (0.15 (0.04 to 0.37) and 0.23 (0.06 to 0.74)) although the risk of major bleeding was higher with standard-intensity VKAs (4.42 (1.99 to 12.24) and 4.14 (1.17 to 18.86)). Effect estimates were consistent in male patients and those with index pulmonary embolism or with unprovoked VTE and in sensitivity analyses. In addition, compared with placebo or observation, the risk of all-cause mortality was reduced with standard-intensity VKAs (0.44 (0.20 to 0.87)) and low-dose factor Xa inhibitors (0.38 (0.12 to 0.995)).
Conclusions Standard-intensity VKAs and DOACs are more efficacious than aspirin for extended VTE treatment. Despite a higher risk of major bleeding, standard-intensity VKAs was associated with a lower risk of all-cause mortality. Since overall efficacy and safety of standard-intensity VKAs and DOACs are in equipoise, patient factors, costs and patient preferences should be considered when recommending extending anticoagulation treatment.
- thromboembolic pulmonary vascular disease
- pulmonary vascular disease
Statistics from Altmetric.com
Introduction
Venous thromboembolism (VTE), which comprises deep vein thrombosis and pulmonary embolism (PE), can be effectively treated with anticoagulants. According to current guidelines, such treatment should be continued for at least 3 months in all patients.1–3 Afterwards, extending anticoagulation treatment for secondary prevention of VTE should be considered in patients at high risk of recurrent VTE, such as those with an unprovoked event.4
Vitamin K antagonists (VKAs) have been the oral anticoagulant of choice for extended VTE treatment, although they are associated with an annual rate of major bleeding as high as 2%. Aspirin, as an alternative to VKAs, may have a lower risk of major bleeding at the expense of a higher risk of recurrent VTE.5 Direct oral anticoagulants (DOACs) have similar efficacy and a lower risk of major bleeding than VKAs do for secondary prevention of VTE in patients with acute VTE.6
Physicians currently have multiple potential treatment options for secondary prevention of VTE. Since direct comparisons between these regimens are scanty, comparative efficacy and safety between regimens remain unclear. To help physicians and patients decide on the optimal management strategy, we performed a systematic review and network meta-analysis to evaluate benefits and harms of different oral anticoagulant regimens and aspirin for extended VTE treatment.
Method
The study protocol is documented online in the PROSPERO registry (CRD42017067066). Our report adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.
Data source and search
We updated our prior systematic search of the literature by searching in MEDLINE, Embase and CENTRAL databases from January 2013 to April 2017 using the same search strategy.7 Moreover, conference proceedings of the American Society of Hematology and the International Society on Thrombosis and Haemostasis (ISTH) were searched from 2013 to 2017. No language restrictions were applied.
Eligibility criteria
Eligible studies were randomised controlled trials (RCTs) comparing any oral anticoagulant regimen or aspirin with one another, placebo or observation for extended treatment for secondary prevention of VTE. Studies had to report at least one of outcomes of interest. Extended VTE treatment was defined as management beyond the first 3 months of acute VTE treatment. Data on ximelagatran were not included, because it is not available for clinical practice.
Data extraction and quality assessment
Two authors (K-LW and NvE) independently screened titles and abstracts. For potentially eligible studies, full-text papers were retrieved and assessed independently. Data from eligible studies were extracted independently by the same authors using a standardised form. Potential risks of bias were assessed independently using the Risk of Bias Tool developed by the Cochrane Collaboration. Disagreements between two authors were resolved by consensus.
Data synthesis and Bayesian network meta-analysis
Primary outcomes were recurrent VTE and major bleeding. Secondary outcomes included all-cause mortality, fatal recurrent VTE and fatal bleeding. Study definitions for all outcomes were accepted. Where possible, outcomes were evaluated according to the intention-to-treat principle. If patients were followed for additional prolonged observation after cessation of extended VTE treatment in both arms, we only included data from the initial treatment period.
Placebo or observation, aspirin (100 mg once daily), low-intensity VKAs (targeting an international normalised ratio between 1.5 and 2.0), standard-intensity VKAs (targeting an international normalised ratio between 2.0 and 3.0), low-dose factor Xa inhibitors (apixaban 2.5 mg twice daily and rivaroxaban 10 mg once daily), standard-dose factor Xa inhibitors (apixaban 5 mg twice daily, edoxaban 60 mg (dose reduced to 30 mg by study criteria) once daily, and rivaroxaban 20 mg once daily) and the direct thrombin inhibitor (dabigatran 150 mg twice daily) were compared pairwise using Bayesian network meta-analysis, which has better developed diagnostic tools for assessing model fit of meta-regression and more suitable characteristics for integrating results into decision making.
Primary analyses on recurrent VTE, major bleeding and all-cause mortality were based on the binomial likelihood model. The Poisson likelihood model was used for analyses on fatal recurrent VTE and fatal bleeding as those events were rare and most studies contained at least one zero cell.
Subgroup analyses were performed on patients at potentially higher risk of recurrent VTE, including male patients, patients with index PE and patients with unprovoked VTE. For subgroup analyses on patients with unprovoked VTE, study definitions of unprovoked VTE were accepted. For studies that did not report data on provoked and unprovoked VTE separately, we only included studies that enrolled 70% or more of patients classified as having unprovoked VTE.
Meta-regression was conducted to examine potential effects of baseline patient characteristics, including age and the proportion of male sex, index PE and unprovoked VTE, anticoagulation treatment duration before randomisation, and the inclusion period of the study (referenced by the year of study publication).
Four sensitivity analyses on recurrent VTE and major bleeding were performed: (1) studies assessing standard-intensity VKAs were excluded if median or mean time in therapeutic range was less than 65%8; (2) Hokusai-VTE was excluded because extended VTE treatment was not evaluated in a randomised fashion9; (3) two studies assessing standard-intensity VKAs with an unclear treatment period were excluded10 11; and (4) an analysis accounting for different lengths of follow-up duration was performed by modelling data into the binomial likelihood with the complementary log-log link function assuming a constant event rate for each study arm in every study.
Median ORs or HRs and 95% credible intervals (CrIs) were reported as appropriate. The probability of the estimating rank and the surface under the cumulative ranking curve for a given regimen were calculated.
The baseline hazard of placebo or observation was estimated by modelling data on placebo or observation arms of studies into the Poisson likelihood. Numbers needed to treat to benefit (NNTB) and numbers needed to treat to harm (NNTH) of each regimen were calculated based on the median event rate of placebo or observation and individual treatment-specific effects. NNTB was defined as the number of patients who need to be treated for 1 year to prevent one recurrent VTE, whereas NNTH was defined as the number of patients who need to be treated for 1 year to cause one major bleeding.
Heterogeneity was evaluated by estimating the posterior distribution of the between-study SD. Inconsistency in the network was assessed by comparing statistics for deviance and the deviance information criterion in fitted consistency and inconsistency models. The goodness of model fit was assessed based on residual deviance and the deviance information criterion. Trace plots and the Brooks-Gelman-Rubin statistic were checked to ensure convergence. No corrections for studies with no events were used. All analyses were conducted in random-effects models with vague priors using WinBUGS V.1.4.3 (MRC Biostatistics Unit, Cambridge, UK).
Result
The updated search identified four additional studies, increasing the total number of patients from 12 407 up to 22 396 (online supplementary figure 1). The evidence network for primary outcomes is depicted in figure 1. Study and patient characteristics are reported in table 1 and online supplementary table 1. The definition for recurrent VTE was generally homogeneous across studies and most studies, except DURAC II and PREVENT, defined major bleeding per the ISTH definition though the key inclusion and exclusion criteria for enrolment varied between studies (online supplementary table 2).
Supplemental material
The risk of bias assessment is reported in online supplementary figure 2. Six studies were judged to be at low risk of bias for all domains, whereas three open-label studies and one post hoc analysis were judged to be at high risk of bias. The posterior mean deviance distributions were close to 1 for consistency and inconsistency models, indicating no evidence of inconsistency in the network (online supplementary figure 3). Furthermore, rates of recurrent VTE and major bleeding with placebo or observation were similar across studies (table 2), suggesting that the reference group used in our study was at low risk of heterogeneity.
Recurrent VTE and major bleeding
Results of all pairwise comparisons are reported in figure 2. Compared with placebo or observation, all oral anticoagulant regimens were associated with a lower risk of recurrent VTE with summary ORs ranging between 0.15 and 0.36 (figure 3A). On contrary, only low-intensity VKAs and standard-intensity VKAs were associated with a higher risk of major bleeding (OR 4.14 (95% CrI, 1.09 to 18.02) and 4.42 (95% CrI 1.99 to 12.24), respectively) (figure 3B). NNTB ranged from 15 for standard-intensity VKAs or DOACs to 20 for low-intensity VKAs, while NNTH was 91 for low-intensity VKAs and 87 for standard-intensity VKAs.
Compared with aspirin, all oral anticoagulant regimens, except low-intensity VKAs, were associated with a reduced risk of recurrent VTE with summary ORs ranging between 0.23 and 0.27, whereas only standard-intensity VKAs was associated with a significant risk of major bleeding (OR 4.14 (95% CrI 1.17 to 18.86)) (online supplementary figure 4). Compared with aspirin, NNTB for oral anticoagulant regimens, except for low-intensity VKAs, ranged from 26 to 27, while NNTH was 90 for standard-intensity VKAs.
Absolute risk differences between placebo or observation, aspirin and oral anticoagulant regimens are reported in online supplementary table 3. Standard-intensity VKAs, low-dose factor Xa inhibitors and the direct thrombin inhibitor had the highest probability of being the most (25%, 28% and 36%, respectively) or the second most (37%, 20% and 20%, respectively) efficacious regimen, while aspirin and low-dose factor Xa inhibitors had the highest probability of being the safest (34% and 21%, respectively) or the second safest (30% and 18%, respectively) regimen besides placebo or observation (online supplementary figure 5).
All-cause mortality, fatal recurrent VTE and fatal bleeding
A total of 21 699 patients were included in the analysis on all-cause mortality. Online supplementary table 4 reports the pooled all-cause mortality rate. Compared with placebo or observation, standard-intensity VKAs and low-dose factor Xa inhibitors were associated with a reduced risk of all-cause mortality (OR 0.44 (95% CrI 0.20 to 0.87) and 0.38 (95% CrI 0.12 to 0.995), respectively) (figure 4). Surface under the cumulative ranking curve was comparable between standard-intensity VKAs and DOACs. Fatal recurrent VTE occurred in 53 of 22 292 patients (0.24%) and fatal bleeding occurred in 12 of 22 284 patients (0.05%) (table 3). Low-dose factor Xa inhibitors and the direct thrombin inhibitor were associated with a lower risk of fatal recurrent VTE compared with placebo or observation. HRs for fatal bleeding could not be estimated because of very low event rates.
Subgroup analysis and patients with unprovoked VTE
Subgroup analyses on recurrent VTE in male patients (nine studies) and patients with index PE (nine studies) yielded consistent results with the primary analysis (online supplementary table 5). Major bleeding could not be evaluated in these subgroups because of incomplete reporting.
Meta-regression and sensitivity analysis
Meta-regression evaluating heterogeneity due to patient characteristics at enrolment and the study inclusion period indicated that effect estimates of all regimens for recurrent VTE and major bleeding were not modified by these factors (online supplementary figure 6 and online supplementary table 7).
Sensitivity analyses that excluded studies in which time in therapeutic range of their standard-intensity VKA treatment was less than 65%, Hokusai-VTE, and additional studies with an unclear treatment period or were accounted for different lengths of follow-up duration of various studies did not alter results significantly (online supplementary figures 7–10).
Discussion
In this updated systematic review and network meta-analysis of extended VTE treatment, including over 22 000 patients, we found that all oral anticoagulant regimens but not aspirin were associated with a lower risk of recurrent VTE, while only VKAs were associated with a higher risk of major bleeding. Effect estimates were consistent across patient subgroups. In addition, standard-intensity VKAs and low-dose factor Xa inhibitors were associated with a reduced risk of all-cause mortality.
The most recent version of the American College of Chest Physicians guideline recommends extending anticoagulation treatment in patients with unprovoked VTE if they have a low or moderate bleeding risk.2 Our results support such an approach since all oral anticoagulant regimens were consistently associated with a reduced risk of recurrent VTE. It should be noted that the primary analysis was based on all patients in whom extended treatment for secondary prevention of VTE was considered, including those with a transient risk for whom extended VTE treatment is not necessarily recommended.1–3 However, meta-regression using the proportion of patients with unprovoked VTE enrolled in studies as a covariate yielded similar results. Additional analyses also showed consistent findings in other subgroups at high risk of recurrent VTE, including male patients and those with index PE,12 13 indicating that overall effect estimates apply to all patients for whom extended VTE treatment is considered.
In the setting of atrial fibrillation or acute VTE, DOACs are associated with less severe major bleeding, a lower case fatality rate of major bleeding and fewer fatal bleeding events compared with standard-intensity VKAs.14 15 For extended VTE treatment, our study suggested that DOACs had a numerically, though not significantly, lower risk of major bleeding than standard-intensity VKAs did. It has to be noted that VKA experienced patients are at lower risk of major bleeding with ongoing treatment than VKA naive patients,16 17 while the quality of VKA management may improve with longer treatment duration and in patients who already have prior exposure.18 Therefore, we were not able to observe significant differences between VKAs and DOACs in the risk of major bleeding by the current study size. Even though the risk of major bleeding with VKAs may be lower in patients already challenged to anticoagulation treatment, their associated case fatality rate, nevertheless, remains greater than 10% during and beyond the first 3 months.19 20 Furthermore, when deciding on longer term secondary prevention of VTE, DOACs appeared safer than standard-intensity VKAs, taking clinical relevant bleeding or any bleeding into consideration.21
Despite a higher risk of major bleeding, standard-intensity VKAs were associated with a lower risk of all-cause mortality. Mortality directly attributed to anticoagulation treatment should remain to be fatal recurrent VTE and fatal bleeding, which were infrequently reported in included studies. Consequently, we were not able to confirm whether the reduction in the risk of all-cause mortality was attributed to a lower risk of fatal recurrent VTE. In addition, rates of all-cause mortality with placebo or observation and with standard-intensity VKAs seemed higher in earlier studies than in contemporary studies. A larger sized RCT will likely be required to detect a small absolute difference in the risk of all-cause mortality between regimens.
Based on two prior RCTs,22 23 aspirin is sometimes considered for extended VTE treatment because physicians and patients may perceive its bleeding risk to be lower than that with oral anticoagulants. Although aspirin, by the frequentist method, reduces the risk of recurrent VTE (HR 0.68 (95% CI 0.51 to 0.90)),5 our study using the Bayesian method demonstrated a comparable but more conservative estimate (OR 0.62 (95% CrI 0.29 to 1.28)). Even though the sensitivity analysis using the fixed-effects model in our prior study yielded a favourable result for aspirin, aspirin was, nevertheless, less efficacious than oral anticoagulant regimens.7 Moreover, the risk of major bleeding appeared similar between aspirin and DOACs. This finding is consistent with RCTs directly comparing aspirin with DOAC regimens for extended VTE treatment and for stroke prevention in patients with atrial fibrillation.24 25 Based on present findings, for every 1000 patients treated with a DOAC rather than with aspirin for 1 year, recurrent VTE could be prevented in approximately 38 patients, among whom one might succumb to the event.19 20 Accordingly, DOACs should be considered over aspirin for their more favourable benefit–harm profiles.
In our study, low-dose factor Xa inhibitors appeared to be as efficacious as standard-dose factor Xa inhibitors, whereas the risk of major bleeding was not statistically different between these regimens (OR 0.76 (95% CrI 0.19 to 2.66)). These findings were primarily based on two RCTs that directly compared low-dose of apixaban and rivaroxaban with their respective standard-dose regimens.25 26 Similarly, in Hokusai-VTE, efficacy and safety were similar between reduced-dose edoxaban, given per study criteria, and standard-dose edoxaban.27 However, these studies were not designed and powered to demonstrate non-inferiority on efficacy and superiority on safety between these dosing regimens. In addition, standard-dose factor Xa inhibitors seemed to prevent more hospitalisations and reduced the length of hospital stay than low-dose factor Xa inhibitors.28 29 Therefore, physicians should individualise their decision of which dose to use when avoiding bleeding.
Our study has a number of strengths and limitations. This updated study, incorporating the most recent RCTs assessing regimens for secondary prevention of VTE, is twice as large as our prior analysis. Accordingly, it provides a more robust evidence network enabling estimations of risk differences between various regimens with greater precision. However, our findings may not apply to patients considered to be at very high risk of recurrent VTE, such as those with antiphospholipid antibody syndrome, multiple recurrent VTE events or cancer, since these patients were under-represented and often excluded from RCTs. Meanwhile, our findings may not be totally generalizable to all patients with VTE in the community since we only included RCTs, which generally applied strict criteria for patient enrolment and in which patients at high risk of bleeding were often not included. As with any meta-analysis, differences in study design, selection criteria, patient characteristics and treatment duration may have potentially impacted our conclusions. Nonetheless, multiple subgroup, meta-regression and sensitivity analyses did not modify our conclusions. Finally, treatment duration with DOACs was less than 1 year in five of six studies. Therefore, longer term clinical experience with DOACs for secondary prevention of VTE remains limited.
Conclusion
Standard-intensity VKAs and DOACs are more efficacious than aspirin for extended VTE treatment. Standard-intensity VKAs are associated with a higher risk of major bleeding but a lower risk of all-cause mortality. Treatment benefits are comparable across an array of patients at high risk of recurrent VTE. Given that there is still equipoise regarding the more benefit–harm improved default, patient factors, costs and patient preferences should be considered when recommending extended VTE treatment.
Key messages
What is already known on this subject?
Current guidelines recommend extending anticoagulation treatment in patients with unprovoked venous thromboembolism (VTE) who have a low or moderate bleeding risk. Randomised controlled trials show that vitamin K antagonists (VKAs), direct oral anticoagulants (DOACs) and probably aspirin prevent recurrent VTE when considering extended VTE treatment. However, the regimen of choice is debated because of differences in efficacy and safety across these agents.
What might this study add?
This network meta-analysis of 16 studies including a total of more than 22 000 patients show that, compared with no treatment, for every 1000 patients, treatment with standard-intensity VKAs or DOACs for a year can prevent 66–69 patients from recurrent VTE, while 12 excess patients might experience major bleeding with standard-intensity VKAs. On the contrary, extended treatment with aspirin rather than with standard-intensity VKAs or DOACs would lead to 37–39 additional patients experiencing recurrent VTE. Comparative efficacy of DOACs is consistent in patients considered to be at high risk of recurrent VTE in addition to those with unprovoked VTE.
How might this impact on clinical practice?
Our results suggest that the benefit–harm balance favours standard-intensity VKAs or DOACs but not aspirin. Therefore, aspirin should not be considered for extended VTE treatment. Our findings may further justify extending anticoagulation treatment with standard-intensity VKAs or DOACs in patients for whom extended VTE treatment is considered but for whom current guidelines do not have separate and specific recommendations. Longer term (beyond 1 year) comparative efficacy and safety between standard-intensity VKAs and DOAC regimens require further studies.
References
Footnotes
K-LW and NE contributed equally.
Contributors The study was conceived and designed by K-LW, NvE, HRB and MC. The data were analysed by all authors. The first draft of the manuscript was prepared by KLW and NvE, which was edited with inputs from CC, LAC, HRB and MC. All authors gave final approval of the version to be published and have contributed to the manuscript. K-LW, NvE and MC are guarantors.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests K-LW reports honoraria from AstraZeneca, Bayer, Boehringer Ingelheim, Daiichi-Sankyo, Orient EuroPharm, Novartis and Tanabe. NvE reports a personal fee from Daiichi-Sankyo and Pfizer. CC reports being a partner at Cornerstone Research Group Inc, which consults for various pharmaceutical and medical device companies. LAC reports honoraria from Bayer, Bristol-Myers Squibb, Leo Pharma and Pfizer. HRB reports research support from Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi-Sankyo, Eli Lilly, GlaxoSmithKline, IONIS, Novartis, Pfizer, Roche and Sanofi and is a consultant for Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi-Sankyo, Eli Lilly, GlaxoSmithKline, IONIS, Novartis, Pfizer, Roche, Sanofi and Thrombogenics. MC reports research funding from Bristol-Myers Squibb and LEO Pharma and honoraria from Bayer, Bristol-Myers Squibb, Leo Pharma, Pfizer and Sanofi.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement We do not have any additional data.