Phase III Trials of New Oral Anticoagulants in the Acute Treatment and Secondary Prevention of VTE: Comparison and Critique of Study Methodology and Results

The design of the VTE studies, characteristics of patients randomized and the variation in study design impose a number of limitations in relation to the generalizability of the results obtained to a clinical practice setting.

Mortality is an important indicator of level of illness in any group of patients enrolled into a clinical study. The rates reported in the acute VTE studies are, therefore, of great interest to fully understand the relevance of the studies in clinical practice. The studies that had around 6 months of follow-up including the EINSTEIN studies reported mortality rates slightly above 2% and the AMPLIFY and RE-COVER studies reported rates slightly below 2% [21‐23, 25, 27]. The study that followed all patients for 12 months, Hokusai-VTE, reported approximately 3.2% total mortality [26]. However, the different lengths of follow-up and the different analyses of the study populations, with and without patients off treatment, do not allow direct between-study comparisons of mortality rates.

In addition to mortality, the range of anatomical extent of PE at baseline also provides an important insight into the relevance of the studies in clinical practice. Both the EINSTEIN-PE and Hokusai-VTE studies used the same criteria to define anatomical extent of PE, with extensive PE defined as involvement of multiple lobes with 25% or more of the entire vasculature. In EINSTEIN-PE, extensive PE was present in approximately 24% of patients and in Hokusai-VTE it was present in approximately 45.8% of patients. The AMPLIFY study used different criteria to define extensive PE, which were at least two lobes with at least 50% of vasculature for each lobe, and approximately 37.2% of patients had extensive PE according to these criteria in AMPLIFY [22, 25, 26]. Despite the varying criteria utilized, the highest reported proportion of patients with extensive PE was in the Hokusai-VTE study [18, 21, 22].

Patient age is a key factor and it has previously been found that the half-life and exposure of the NOACs dabigatran, rivaroxaban and apixaban are higher in the elderly [29]. Also, elderly patients are more likely to suffer greater bleeding complications both with and without anticoagulation [30]. Although no upper-age limits were set with regard to randomisation of patients, elderly and younger patients were under-represented and mean ages ranged from 54 to 58 years (Table 3). The placebo group of the EINSTEIN-Extension (NCT00439725) trial had the highest mean age in any of the trials at 58.4 years [21]. With regard to race and ethnicity, patients were predominantly Caucasian in most of the studies even though the trials were multinational. For example, 94.8% of patients in the RE-COVER study were Caucasian [23]. Although the larger Hokusai-VTE study had a predominantly Caucasian study population (approximately 70%), the study also had a varied ethnic composition with 21% Asian and around 3.5% Black or African–American patients, and hence was the largest and most internationally representative VTE study [26].

Renal function is also a key factor in treatment with a NOAC and patients with renal impairment were under-represented in clinical trials. The AMPLIFY and AMPLIFY-Extension (NCT00633893) trials excluded patients with a creatinine clearance of <25 ml/min [20]. In AMPLIFY, the apixaban arm had 64% of patients with a creatinine clearance >80 ml/min and approximately 6% of patients had a creatinine clearance <50 ml/min [25]. A similar proportion was observed in the AMPLIFY-Extension study, approximately 71% of patients in the apixaban arm had a creatinine clearance >80 ml/min and only 5–6% of patients had a creatinine clearance <50 ml/min [20]. The Hokusai-VTE trial also included few patients with renal impairment, less than 7% of patients had a creatinine clearance of ≥30 ml to <50 ml/min. Although patients with a creatinine clearance <30 ml/min were excluded from the EINSTEIN trials, some patients with a lower creatinine clearance were enrolled [21, 22]. Subsequent pharmacodynamic modeling resulted in rivaroxaban being utilized at an estimated glomerular filtration rate as low as 15 ml/min. Although no weight restrictions existed in the trials, it is also clear that participants with a body mass index (BMI) ≥35 kg/m² were not frequently randomized. The mean BMI of the dabigatran group in the RE-COVER trial was 28.9 [standard deviation (SD) ± 5.7] kg/m² [23]. In Hokusai-VTE, approximately 13% of patients had a body weight ≤60 kg and 15% weighed >100 kg [22]. The mean weight of the apixaban 2.5 mg group in the AMPLIFY-Extension trial was 85.7 (SD ± 19.8) kg. In the AMPLIFY study, approximately 72% of the apixaban group weighed 60–100 kg and 19% weighed ≥100 kg [25]. In the EINSTEIN studies, approximately 14% of patients had a body weight >100 kg [17, 18]. A BMI ≥30 kg/m² was found in around 30% of patients in the EINSTEIN, AMPLIFY and RE-COVER studies, and in AMPLIFY and RE-COVER around 12% had a BMI ≥35 kg/m² [21‐23, 25]. The proportion of obese patients is broadly consistent with levels found in population studies.

Overall, the limited inclusion of elderly patients, patients with renal impairment, very obese patients and patients from non-white ethnic groups may reduce the generalizability of the results of VTE studies in clinical practice.

The VTE study publications also provide limited information on drug interactions, which may have been very useful information for clinical practice. The AMPLIFY and EINSTEIN trials excluded patients taking strong cytochrome P-450 inhibitors/inducers, which means that they should not be recommended for patients who require these agents [21, 22, 25, 26]. In addition, although concomitant low-dose aspirin use was permitted in all of the studies, only the Hokusai-VTE study investigators reported any analysis of the effect of aspirin on efficacy and bleeding rates. These subgroup analyses did not find any effect of concomitant aspirin on the efficacy and safety of edoxaban [26].

There are other facets of trial design that are important to consider when translating trial evidence into clinical practice. The duration and use of a parenteral heparin lead in prior to commencing NOAC therapy varied across the trials. In the RE-COVER trial the median length of parenteral anticoagulation post-randomisation was 6 days and in Hokusai-VTE the median length was 7 days, therefore, there is currently no evidence to support the immediate use of initial dabigatran or edoxaban monotherapy in VTE treatment [23, 26]. The EINSTEIN-PE and EINSTEIN-DVT trials excluded patients who had received more than 48 h of parenteral anticoagulation and the AMPLIFY study also excluded patients who received more than 1 day of LMWH therapy, or more than 36 h of continuous intravenous heparin [21, 22, 25]. However, it is known that the rate of VTE recurrence is highest during initial parenteral therapy and during the transition to VKA in this treatment strategy [31]. This raises the possibility that recurrence rates could be higher than those observed in the trials. Only few patients in the clinical trials used rivaroxaban and apixaban as a monotherapy because approximately 80–90% received a pre-randomisation dose of parenteral heparin. However, a subgroup analysis of EINSTEIN did not confirm differences between patients who only received rivaroxaban and those who had initial parenteral therapy [21]. The optimal strategy for using or not using a heparin lead in has not been determined in a randomized trial and, therefore, this is an open issue. In clinical practice some physicians may prefer to use NOAC monotherapy for VTE treatment in lower risk patients suitable for outpatient therapy, while others may prefer a heparin lead in, especially for sicker patients requiring inpatient therapy.

Another major difference in trial design is the use of open- or double-blinded methodology. Double-blinded trials are traditionally viewed as the ‘gold standard’ in design. Blinding protects against detection and reporting bias arising from patients and investigators knowing which treatment is being received. However, blinding results in the loss of information that may reflect true differences in a randomized trial with respect to the quality of life experienced with different regimens. The EINSTEIN-DVT and EINSTEIN-PE trials were open-label and, therefore, a direct comparison of the effects of rivaroxaban and VKA on quality of life could be performed. A higher rate of recurrent VTE episodes was suspected in the rivaroxaban arm than in the VKA arm in the EINSTEIN-DVT trial, indicating a possible diagnostic-suspicion bias against rivaroxaban. However, the blinded central adjudicating committee did not confirm higher VTE recurrences in the rivaroxaban arm [21]. Double-blinded trials are complex in the context of anticoagulant therapy. The AMPLIFY, Hokusai-VTE and RE-COVER trials used double-dummy methodology to maintain blinding and minimize bias. In these studies, the arm receiving active NOAC also received placebo warfarin together with ‘sham INR’ monitoring. In addition to making patient recruitment harder, patient selection bias may occur in double-dummy trials as investigators may doubt the ability of elderly or frail patients to follow complex instructions [32]. In general, double-dummy trials also tend to have higher discontinuation rates, as is seen in the RE-COVER trial (15.2% discontinuation) compared with the open-label EINSTEIN-PE (11.5% discontinuation) and EINSTEIN-DVT (12.8% discontinuation) trials. However, the double-dummy Hokusai-VTE study had a discontinuation rate of 4.2%, which was the lowest in any of the VTE studies. This may be explained by the extremely low proportion of patients who withdrew consent (<0.9%) or who were lost to follow-up (<0.2%) in Hokusai-VTE [26]. Double-dummy design may also bias towards a higher time in therapeutic (TTR) range, in relation to INR control, by selecting more compliant patients [32]. The mean TTR observed in clinical studies using a VKA to treat VTE has been estimated at 60% [33]. Although the methods of assessing TTR vary from trial to trial, the RE-MEDY trial with dabigatran had the highest mean TTR of 65% [24]. A selection-to-continue bias present in the extended treatment RE-MEDY trial may have also contributed to the higher TTR achieved. The EINSTEIN-DVT open-label trial with rivaroxaban had a lower TTR of 57% than the EINSTEIN-PE which had a TTR of 63%, indicating more stringent control in PE patients compared with DVT patients. Clinical trial results for TTR are higher than normally seen in clinical practice. The mean TTR varies in reported series, however, a very large study in the United States demonstrated that in a real-world population, less than 50% of warfarin patients achieved INR values within the therapeutic range [34]. Furthermore, adherence is often poor, missed and extra doses are not uncommon and subsequently INR is frequently out of range [35]. It is important to consider that missing doses of NOACs would also result in patients’ anticoagulation becoming sub-therapeutic. NOACs have a shorter half-life than warfarin, and therefore a missed dose may pose a greater short-term risk of VTE recurrence [36].

All of the trials comparing a NOAC with VKAs were designed to assess for non-inferiority and utilized intention-to-treat analyses, but it is recommended to also perform a per-protocol analysis because this strengthens an equivalence finding if both analyses are in agreement [37]. The evidence for non-inferiority of NOACs may depend on the quality of VKA treatment reflected by TTR. In addition, premature discontinuation, small sample size, lack of blinding and effects of concomitant medication all may increase the chance of finding non-inferiority when a true difference exists [38]. The incentive to reduce the impact of these factors is greater when a trial aims to demonstrate superiority. The importance of sample size was observed in the non-inferiority RE-COVER trial, which may have initially been too small as a low rate of recurrent VTE was observed [23]. The replica RE-COVER II study was then conducted to confirm the findings of RE-COVER [27]. The VTE trial with the largest study population was Hokusai-VTE in 8,292 patients, which was more than double the size of RE-COVER [22].

The duration of VTE treatment is a key issue for every patient to balance the risk of recurrence and bleeding effectively. It is important to optimize treatment duration dependent on patients’ characteristics. Although the risk of VTE recurrence is highest within the initial 6 months, it does not return to normal after this period and it is also of note that the optimal duration of VTE treatment still remains unclear in clinical guidelines [31]. Most of the acute VTE trials with the NOACs did not have long follow-up periods, and important information about the recurrence rates of VTE upon cessation of NOAC treatment compared with VKA was not obtained. A more complete picture of the effect of length of treatment could have been determined if all patients in the NOAC trials were followed for the same length of time and all events were collected over that time period. The only exception was Hokusai-VTE, which had a flexible treatment duration of 3–12 months and all patients were analyzed at 12 months [26]. In the double-blind, double-dummy Hokusai-VTE trial, 40% of patients completed 12 months of treatment, approximately 62% had >6 months of therapy and 26% had 3 to ≤6 months of therapy [26]. This suggests that in a patient population comprised of large numbers of patients with extensive PE (46%) and extensive DVT (42%), the extent of disease does not determine length of therapy. The investigators’ previous double-dummy clinical trial experience and the flexible treatment duration in Hokusai-VTE may have been two of the reasons that very low levels of patient discontinuation and withdrawal of consent were observed, as compared with the other VTE studies. Insights into the appropriate duration of treatment for provoked, unprovoked, limited and extensive PE and DVT could be made from further analyses of the Hokusai-VTE study, including recurrence rates after cessation of treatment. In addition, the extension trials with dabigatran, rivaroxaban and apixaban add to the growing evidence in support of continued anticoagulation in certain clinical settings.

Promising results have been obtained with the NOAC agents dabigatran, rivaroxaban, apixaban and edoxaban in acute VTE treatment and secondary prevention studies (Table 5). The RE-COVER trial compared heparin/dabigatran 150 mg twice daily with heparin/warfarin in the prevention of recurrent or fatal VTE and found that heparin/dabigatran was non-inferior to heparin/warfarin (p < 0.001 for non-inferiority) [23]. Dabigatran was also associated with significantly fewer major or clinically relevant non-major bleeding, as well as any bleeding events (Table 5) [23]. The replica RE-COVER II trial confirmed that the efficacy of heparin/dabigatran in acute VTE was non-inferior to heparin/warfarin and dabigatran also had a lower risk of bleeding [27]. Similar results have been observed with rivaroxaban, which was compared with standard enoxaparin/VKA therapy in acute symptomatic proximal DVT in EINSTEIN-DVT and in acute symptomatic PE with or without symptomatic DVT in EINSTEIN-PE [21, 22]. Both studies found rivaroxaban was non-inferior compared with enoxaparin/VKA in the prevention of symptomatic, recurrent VTE. In both studies, rivaroxaban was associated with comparable levels of major or clinically relevant non-major bleeding and in EINSTEIN-PE the rivaroxaban group had significantly fewer major bleeding events than standard therapy [21, 22]. The AMPLIFY study demonstrated that apixaban was non-inferior to conventional therapy with enoxaparin followed by warfarin, in the prevention of recurrent VTE or related death. Apixaban treatment was also associated with a significantly lower risk of major or clinically relevant non-major bleeding than conventional therapy [25]. The Hokusai-VTE study demonstrated that edoxaban was non-inferior to warfarin in the prevention of recurrent symptomatic VTE and edoxaban was also associated with significantly fewer major or clinically relevant non-major bleeding events than warfarin at 12 months. Approximately 17.5% of the Hokusai-VTE study population qualified for dose reduction from edoxaban 60 mg to 30 mg. Among the dose-adjusted group, 22/733 (3.0%) edoxaban patients and 30/719 (4.2%) warfarin patients had recurrent VTE events (HR 0.73; 95 % CI 0.42–1.26), which correlated with results in the overall study population [26].

The duration of treatment is an uncertain issue in VTE treatment because the benefit of preventing VTE recurrence and risk of bleeding must be balanced for each patient. Following on from the acute VTE treatment studies, continued long-term therapy with the NOACs dabigatran, rivaroxaban and apixaban have been performed and provided promising results. The RE-MEDY study (Tables 3, 5) compared dabigatran with warfarin in a 6- to 36-month extended treatment period after 3–12 months of initial VTE therapy [24]. Dabigatran showed non-inferior efficacy compared with warfarin in the primary outcome of symptomatic, recurrent VTE or VTE-related death and significantly fewer bleeding events and major or clinically relevant non-major bleeds were observed with dabigatran than warfarin [24]. However, patients in the dabigatran group had a significantly higher rate of acute coronary syndrome events than those in the warfarin group (p = 0.02) [20]. A similar observation was made in patients with atrial fibrillation who received dabigatran compared with warfarin during the RE-LY trial [35].

Placebo-controlled extension studies have also been performed with the NOACs to investigate benefits compared with treatment cessation (Tables 4, 6). The RE-SONATE study compared dabigatran with placebo in a 6- to 18-month extension period that followed 6–18 months of initial VTE treatment [24]. RE-SONATE showed that long-term treatment was significantly more effective than placebo (p < 0.001) in prevention of symptomatic recurrent VTE and related deaths. However, dabigatran treatment was associated with a significantly higher rate of any bleeding and major or clinically relevant bleeding. In the EINSTEIN-Extension study with a 6- or 12-month extension, rivaroxaban was associated with significantly fewer symptomatic, recurrent VTE events than placebo (p < 0.001) [21]. The rivaroxaban arm had a significantly higher incidence of major or clinically relevant non-major bleeding than placebo (p < 0.001), as well as comparable non-fatal major bleeding to placebo. In the 12-month AMPLIFY-Extension study in symptomatic DVT or PE patients [16], both of the apixaban 5 mg and 2.5 mg doses were superior to placebo in prevention of symptomatic, recurrent VTE or death from any cause (p < 0.001 for both comparisons). The rate of major bleeding was similar in the three treatment groups at 0.5% (n = 4) with placebo, 0.2% (n = 2) with apixaban 5 mg and 0.1% (n = 1) with apixaban 2.5 mg. Although the placebo group had the highest rate of major bleeds, no significant difference was observed and such fluctuations are likely to occur by chance. In addition, the rate of major or clinically relevant non-major bleeding was 2.7% in the placebo group, 4.3% in the apixaban 5 mg group and 3.2% in the apixaban 2.5 mg group [16]. Based on these bleeding data, some commenters have speculated that the intermediate dose of apixaban 2.5 mg may be most beneficial in extended treatment. Taken together, the placebo-controlled RE-SONATE, EINSTEIN-Extension and AMPLIFY-Extension studies suggested that patients at clinical equipoise between treatment continuation and cessation may benefit from additional treatment. This was demonstrated because the benefit of the NOACs with an 80% reduction in preventing recurrent venous thromboembolic events exceeded the risk of major bleeding and because the net clinical benefit was evident for rivaroxaban and apixaban compared with placebo [20, 21]. However, the optimal duration of extended therapy with NOACs still requires further clarification as only dabigatran was compared with a VKA in RE-MEDY, the other extension studies had a limited duration of 12–18 months and patients were not followed up after stopping anticoagulant treatment in the EINSTEIN-Extension the AMPLIFY-Extension studies [17, 20, 21].

To help ascertain the benefit of the NOACs in clinical practice, it is important to analyze observed rates of VTE recurrence in relation to bleeding (Figs. 1, 2). In the extended treatment studies, rates of VTE recurrence were very low in the anticoagulant groups (rivaroxaban, apixaban and dabigatran) and were significantly higher, around 10% in the placebo groups. However, rates of clinically relevant bleeding with rivaroxaban and dabigatran were higher than placebo. The lowest bleeding rate in this group of studies was seen with apixaban. Although the extended treatment RE-MEDY study found less bleeding with dabigatran than warfarin, VTE recurrence rates were slightly higher with dabigatran than with warfarin. In the acute VTE studies, bleeding rates with the NOACs were all less than with VKA treatment. The AMPLIFY study with apixaban, the EINSTEIN-DVT study with rivaroxaban and Hokusai-VTE with edoxaban suggested lower absolute rates of both VTE recurrence and bleeding compared with VKA treatment (Fig. 1).