OUP user menu

Efficacy and safety of apixaban compared with aspirin in patients who previously tried but failed treatment with vitamin K antagonists: results from the AVERROES trial

Michiel Coppens, David Synhorst, John W. Eikelboom, Salim Yusuf, Olga Shestakovska, Stuart J. Connolly
DOI: http://dx.doi.org/10.1093/eurheartj/ehu048 1856-1863 First published online: 25 February 2014

Abstract

Aim The AVERROES double-blinded, randomized trial demonstrated that apixaban reduces the risk of stroke or systemic embolism (SSE) by 55% compared with aspirin without an increase in major bleeding in patients with atrial fibrillation either who previously tried but failed vitamin K antagonists (VKA) therapy or who were expected to be unsuitable for VKA therapy. In this pre-specified analysis, we explored the consistency of the results in the subgroup of patients who tried but failed VKA therapy.

Methods and results Of 5599 patients, 2216 (40%) had previously failed VKA treatment [main reasons: poor international normalized ratio (INR) control 42%, refusal 37%, bleeding on VKA 8%]. Compared with those expected to be unsuitable for VKA therapy, those who had previously failed were older, more often male, had higher body mass index, more likely to have moderate renal impairment and a history of stroke and less likely to have heart failure or to be medically undertreated. The effects of apixaban compared with aspirin were consistent in those who previously failed and those who were expected to be unsuitable, for both SSE (P interaction 0.13) and major bleeding (P interaction 0.74) and were also consistent among different subgroups of patients who had previously failed VKA therapy defined by reasons for unsuitability, age, sex, renal function, CHADS2 score, aspirin dose, duration, indication, and quality of INR control of prior VKA use.

Conclusion The efficacy and safety of apixaban compared with aspirin is consistent in subgroups of patients who have previously attempted but failed VKA therapy, irrespective of the reason for discontinuation.

  • Atrial fibrillation
  • Stroke
  • Bleeding
  • Vitamin K antagonists
  • Apixaban
  • Aspirin

Introduction

Vitamin K antagonists (VKAs) are effective for the prevention of ischaemic stroke or systemic embolism (SSE) in patients with atrial fibrillation (AF) but can be challenging to use in clinical practice because of their high propensity for food, alcohol, and drug interactions, need for routine coagulation monitoring and risk of bleeding.1,2 These limitations of VKAs contribute to their underutilization; surveys in North America and Europe have shown that VKAs are used in only 50–60% of patients for whom this treatment is recommended by the guidelines.3,4

Most Phase 3 randomized controlled trials of new oral anticoagulants for stroke prevention in AF have compared the new drug against warfarin,57 and it is therefore uncertain whether these new drugs are also beneficial for patients unsuitable for VKA therapy. In the AVERROES trial, the efficacy and safety of apixaban was compared with that of aspirin in AF patients with additional risk factors for stroke who are deemed to be unsuitable for VKA therapy, either because they had previously tried but discontinued treatment with VKAs or because they were expected to be unsuitable for VKA therapy.8 However, patients who have previously received VKA therapy may differ from patients that never received VKA therapy with regard to risks of stroke and bleeding. In the AVERROES trial, the rates of bleeding were > 50% higher in patients who had previously failed warfarin compared with those who were deemed unsuitable for warfarin but had not previously been treated with VKAs.8 Similarly, patients who were taken off VKA therapy because of a serious bleeding episode may differ from those who failed VKA therapy because of inadequate international normalized ratio (INR) control and from those who refused further VKA therapy. It is unclear whether the efficacy and safety results of the AVERROES trial can be uniformly applied to these patients.

The first aim of this pre-specified subgroup analysis was to examine the consistency of the efficacy and safety of apixaban compared with aspirin in patients who had previously failed VKA therapy, according to different subgroups including the reason for discontinuation. The second aim was to explore the differences between patients that previously failed VKA therapy and those that were expected to be unsuitable but had never received warfarin in the past.

Methods

AVERROES trial design

The AVERROES trial design has been published previously.9 Briefly, the primary study objective was to establish the superiority of apixaban compared with aspirin for stroke prevention in patients with AF and at least one additional risk factor for stroke who were unsuitable for VKA therapy. The AVERROES investigators randomized 5599 patients to receive apixaban of 5 mg twice daily or aspirin at a dose of 81–324 mg once daily (dose determined by local investigator). Patients that met two of the three following criteria including the age of 80 years or older, body weight of 60 kg or less, and serum creatinine level of 133 μmol/L (1.5 mg/dL) or higher, received a study dose of 2.5 mg apixaban or apixaban-placebo twice daily. Among the exclusion criteria were an indication for anticoagulation other than AF, severe renal insufficiency [creatinine of >221 μmol/L (>2.5 mg/dL) or estimated glomerular filtration rate (eGFR) of <25 mL/min] or a serious bleeding episode in the preceding 6 months.9 Baseline data on the reasons for VKA unsuitability and details of prior VKA use were obtained by structured questionnaires.

Outcome definitions

The primary efficacy outcome of the AVERROES trial was SSE. Stroke was a clinical diagnosis that was made on the basis of typical symptoms lasting at least 24 h. Brain imaging, which was available in the vast majority of patients, was not required but was recommended for the diagnosis of stroke; however, it was required to differentiate ischaemic from haemorrhagic events. Systemic embolism was defined as an acute arterial occlusion of an extremity or organ and documented by means of imaging, surgery, or autopsy.

Major bleeding was the primary safety outcome and was defined as clinically overt bleeding accompanied by one or more of the following: decrease in haemoglobin of 1.2 mmol/L (2 g/dL) or more over a 24-h period, transfusion of two or more units of packed red cells, bleeding into a critical site (intra-cranial, intra-spinal, intra-ocular, pericardial, intra-articular, intra-muscular with compartment syndrome, or retroperitoneal), or fatal bleeding. All clinically overt bleeding not meeting these criteria was considered minor bleeding. All efficacy and safety outcomes were confirmed by a Blinded Central Adjudication Committee.

Statistical analyses

Analyses were based on the intention-to-treat principle. Patients were kept in the treatment group to which they were randomized and were followed as long as possible from the randomization until either study end, loss to follow-up or death. We compared baseline characteristics between patients unsuitable for warfarin because they had previously failed VKA therapy and patients who were expected to be unsuitable but had not previously received VKA treatment using the Pearson χ2 test for categorical variables and the Wilcoxon two-sample test for continuous variables. We assessed the effect of apixaban vs. aspirin on the risk or SSE and major bleeding in subgroups of patients who previously failed VKA therapy using Cox proportional-hazards regression models fitted separately within each level of a subgroup variable. We examined the homogeneity of treatment effect with apixaban by including a treatment by subgroup interaction in the Cox model fitted to all patients. Subgroups were defined by age (<65, 65–<75, ≥75 years), sex, eGFR (<50, 50–<80, ≥80 mL/min, as calculated by the CKD-EPI equation10), CHADS2 score (0–1, 2, 3–6), history of stroke or transient ischaemic attack (TIA), aspirin dose (<162 or ≥162 mg), presence of heart failure, indication for prior VKA use (AF, non-AF), duration of prior VKA use (<6, 6–24, >24 months), prior bleeding (any bleeding and major bleeding only), and quality of prior INR control (INR out of range <25, 25 to <40, 40 to <50, ≥50% of time).

Analyses were performed using the SAS software, version 9.2 (SAS Institute, Inc., Cary, NC, USA). A two-sided P-value of 0.05 was considered statistically significant. Results associated with a P-value between 0.05 and 0.001 were interpreted with caution because of multiple testing.

The AVERROES trial complied with the Declaration of Helsinki and was approved by all appropriate regulatory authorities and ethics committees (ClinicalTrials.gov no. NCT004969769). All the patients provided written informed consent before study entry. The authors of the manuscript had full access to the data and planned the statistical analyses.

Results

In the AVERROES trial, 5599 patients were randomized to treatment with apixaban or aspirin. Of those, 2216 patients (40%) had previously received and failed treatment with VKA; 3383 patients (60%) had not previously received VKA treatment but were expected to be unsuitable. The AVERROES trial was terminated early after two consecutive interim analyses showed overwhelming evidence of a treatment benefit in favour of apixaban over aspirin for the primary efficacy outcome (>4 SD). The mean follow-up duration was 1.1 years.8

Compared with patients who had not previously received VKA treatment, those who had failed VKA therapy were older (mean age 70.3 vs. 69.7 years, P = 0.01), more often male (60 vs. 58%, P = 0.04), had a higher body mass index (28.9 vs. 27.9 kg/m2, P < 0.001), had a similar CHADS2 score (2.1 vs. 2.0, P = 0.40), and a higher proportion had moderate renal impairment (17 vs. 15% with an eGFR of 25–50 mL/min, P = 0.004, Table 1). Other differences included more patients with prior stroke or TIA (15 vs. 13%, P = 0.02), fewer patients with heart failure (35 vs. 42%, P < 0.001), and fewer patients that were medically undertreated (more use of verapamil or diltiazem, beta-blocker, and statin). Forty-eight per cent of patients enrolled in North America or Western Europe had never received VKA therapy compared with 68% of patients included in other regions (Table 1).

View this table:
Table 1

Baseline characteristics of patients according to unsuitability for treatment with vitamin K antagonists

Previously failed VKA treatment (n = 2216)Expected unsuitable for VKA treatment (n = 3383)P-value*
Age, year70.3 ± 9.569.7 ± 9.60.01
Heart rate (b.p.m.)74 ± 1474 ± 140.27
Systolic blood pressure (mmHg)132 ± 17132 ± 160.39
Body mass index (kg/m2)a28.9 ± 5.727.9 ± 5.6<0.001
Male sex1333 (60)1944 (58)0.04
Classification of atrial fibrillation
 Paroxysmal543 (25)969 (29)0.001
 Persistent464 (21)713 (21)
 Permanent1209 (55)1699 (50)
Renal function
 eGFR <50 mL/min378 (17)506 (15)0.004
 eGFR 50 to <80 mL/min1242 (56)1842 (55)
 eGFR ≥80 mL/min594 (27)1034 (31)
Previous bleeding episodes on VKA therapy533 (24)n/an/a
 Major bleeding96 (4)n/a
 Minor bleeding437 (20)n/a
Risk factors for stroke
 Prior stroke or transient ischaemic attack331 (15)433 (13)0.02
 Hypertension, receiving treatment1907 (86)2930 (87)0.50
 Heart failure768 (35)1403 (42)<0.001
 Diabetes mellitus, receiving treatment448 (20)648 (19)0.34
 Left ventricular ejection fraction ≤35%117 (5)171 (5)0.72
 Peripheral arterial disease71 (3)81 (2)0.08
 Mitral stenosis48 (2)66 (2)0.44
CHADS2 score (mean score)b2.1 ± 1.12.0 ± 1.00.40
 0–1830 (38)1196 (35)0.02
 2741 (34)1258 (37)
 3–6644 (29)926 (27)
Concomitant treatment
 ACE inhibitor or ARB1405 (64)2171 (65)0.46
 Verapamil or diltiazem252 (11)247 (7)<0.001
 β-Blocker1311 (59)1786 (53)<0.001
 Digoxin617 (28)958 (28)0.65
 Amiodarone238 (11)388 (12)0.38
 Statin890 (40)974 (29)<0.001
 Regular alcohol consumption (at least once/week)639 (29)718 (21)<0.001
Region
 North America or Western Europe1082 (49)980 (29)<0.001
 Other regions1134 (51)2403 (71)
Study dose of aspirin or aspirin-placebo
 <162 mg1325 (60)2277 (67)<0.001
 ≥162 mg890 (40)1106 (33)
Study dose of 2.5 mg apixaban twice daily or apixaban-placebo140 (6)221 (7)0.75
  • Continuous variables are expressed as means ± standard deviation, categorical variables as number of patients (%). ACE, angiotensin-converting enzyme; ARB, angiotensin-receptor blocker; eGFR, estimated glomerular filtration rate calculated with the CKD-EPI formula.10

  • aThe body mass index is the weight in kilograms divided by the square of the height in metres.

  • bCHADS2 is a stroke risk score (ascending risk from 1 to 6), which gives one point each for a history of heart failure, hypertension, the age of ≥75 years, and diabetes, and two points for either stroke or transient ischaemic attack. There were 15 patients (nine patients with VKA therapy demonstrated unsuitable) with a CHADS2 score of 0, 117 of whom had either peripheral artery disease or left ventricular ejection fraction of ≤35%.

  • *P-value is from the Pearson χ2 test for categorical variables and from the Wilcoxon two-sample test for continuous variables.

Among 2216 patients who previously failed VKA therapy, a bleeding episode during VKA therapy was the reason for unsuitability in 8% of patients, although 24% (n = 533) had experienced an episode of minor (n = 437) or major (n = 96) bleeding during prior VKA treatment (Table 2). Other reasons for unsuitability were inability to maintain the INR in the target range (42%), patient refusal of further VKA treatment (37%), and adverse events not related to bleeding (8%). Atrial fibrillation was the most common indication for prior VKA treatment (97%), and 44% of patients were treated with VKA for a period shorter than 6 months. These variables were equally distributed between the randomized apixaban and aspirin treatment groups (Table 2).

View this table:
Table 2

Reason for unsuitability for vitamin K antagonist use and details of prior vitamin K antagonist use in patients who previously failed vitamin K antagonist treatment

All (n = 2216)Aspirin (n = 1107)Apixaban (n = 1109)
Reasons for VKA unsuitability
 Unable to maintain INR in target range932 (42)468 (42)464 (42)
 INR outside target range of 2.0–3.0
  <25% of time537 (24)282 (25)255 (23)
  25 to <40%262 (12)127 (11)135 (12)
  40 to <50%195 (9)94 (8)101 (9)
  ≥50%525 (24)270 (24)255 (23)
  unknown697(31)334 (30)363 (33)
 Adverse event during VKA therapy, not related to bleeding180 (8)94 (9)86 (8)
 Bleeding event during VKA therapy173 (8)82 (7)91 (8)
 Patient refuses VKA treatment819 (37)418 (37)401 (36)
Details of prior VKA use
 Duration of past VKA use
  <6 months978 (44)493 (45)485 (44)
  6–24 months547 (25)275 (25)272 (25)
  >24 months687 (31)339 (31)348 (31)
 Bleeding episode on VKA533 (24)264 (24)269 (24)
  Major bleeding96 (4)48 (4)48 (4)
  Minor bleeding437 (20)216 (20)221 (20)
 Indication for past VKA use
  Atrial fibrillation2088 (94)1039 (94)1049 (95)
  Non-atrial fibrillation68 (3)40 (4)28 (3)
  Both56 (3)28 (3)28 (3)
  • All data are presented as number of patients (%). VKA, vitamin K antagonist; INR, international normalized ratio.

The effects of apixaban compared with aspirin were consistent in those who previously failed warfarin and those who were expected to be unsuitable for warfarin, for both SSE (P interaction 0.13) and major bleeding (P interaction 0.74; Table 3, adapted from Connolly et al.8).

View this table:
Table 3

Incidence and relative risk of stroke or systemic embolism and major bleeding in apixaban- and aspirin-treated patients according to previous vitamin K antagonist therapy

 AspirinApixabanApixaban vs. Aspirin
NnRate (/100 patient-years)NnRate (/100 patient-years)Hazard ratio (95% CI)P-valueInteraction P-value
Stroke or systemic embolism
 All27911133.72808511.60.45 (0.32–0.62)<0.001
 Tried and failed VKA therapy1107524.21109171.40.32 (0.19–0.56)<0.0010.13
 Expected to be unsuitable for VKA therapy1684613.31699341.80.55 (0.36–0.84)0.005
Major bleeding
 All2791391.22808441.41.13 (0.74–1.75)0.57
 Tried and failed VKA therapy1107211.71109221.81.06 (0.58–1.93)0.840.74
 Expected to be unsuitable for VKA therapy1684181.01699221.21.22 (0.66–2.31)0.52
  • N is the number of patients; n is the number of events. Events are defined as the first occurrence of stroke or systemic embolism. Time to event is the time between randomization date and the event first occurrence. The rate is the number of outcome events per 100 patient-years of follow-up. Data adapted from Connolly et al.8

  • P-value and interaction P-value are from the likelihood ratio test. Hazard ratios were estimated with Cox proportional-hazards regression models fit separately to each subgroup. Hazard ratio confidence interval limits are the profile likelihood limits. Interaction between the effect of apixaban vs. aspirin and subgroups was assessed with the Cox model fit to all patients.

  • AF, atrial fibrillation; CI, confidence interval; Egfr, estimated glomerular filtration rate calculated with the CKD-EPI formula;10 INR, international normalized ratio; TIA, transient ischaemic attack; VKA, vitamin K antagonist.

In patients who failed warfarin therapy (n = 2216), apixaban compared with aspirin reduced the risk of SSE by 68% [event rate 1.4 vs. 4.2 per 100 patient-years; hazard ratio (HR) 0.32, 95% confidence interval (CI) 0.19–0.56, P < 0.001; Table 3], and the incidence of major bleeding was similar in both treatment groups (event rate 1.8 vs. 1.7 per 100 patient-years; HR: 1.06, 95% CI: 0.58–1.93, P = 0.84; Table 2). In an on-treatment analysis, which included only events that occurred in patients while they were receiving the study treatment (i.e. ≤2 days after permanent discontinuation of the study medication), the event rate of major bleeding of 1.8 per 100 patient-years in the apixaban group, when compared with 1.3 per 100 patient-years in the aspirin group (HR: 1.43, 95% CI: 0.76–2.76, P = 0.27). Patients who were unsuitable for VKA therapy because of prior bleeding episodes had higher incidences of both SSE and major bleeding than patients without previous bleeding episodes. However, the superiority of apixaban vs. aspirin with regard to stroke prevention and the similar incidence of major bleeding was preserved in those high-risk patients, as well as in the patients that were unsuitable due to inability to maintain the target INR (Figures 1 and 2). Except in subgroups that had few outcome events, the effects of apixaban vs. aspirin were consistent in the subgroups examined, including reason for VKA unsuitability, age, sex, renal function, CHADS2 score, history of stroke, or heart failure, and aspirin dose (Figures 1 and 2).

Figure 1

Incidence and relative risk of stroke or systemic embolism in apixaban- and aspirin-treated patients who previously failed VKA therapy, according to subgroups. N is the number of patients; n is the number of events. Events are defined as the first occurrence of stroke or systemic embolism. Time to event is the time between randomization date and the event first occurrence. The rate is the number of outcome events per 100 patient-years of follow-up. P-value and interaction P-value are from the likelihood ratio test. Hazard ratios were estimated with Cox proportional-hazards regression models fit separately to each subgroup. Hazard ratio confidence interval limits are the profile likelihood limits. Interaction between the effect of apixaban vs. aspirin and subgroups was assessed with the Cox model fit to all patients. AF, atrial fibrillation; CI, confidence interval; eGFR, estimated glomerular filtration rate calculated with the CKD-EPI formula;10 INR, international normalized ratio; TIA, transient ischaemic attack; VKA, vitamin K antagonist.

Figure 2

Incidence and relative risk of major bleeding in apixaban- and aspirin-treated patients who previously failed VKA therapy, according to subgroups. N is the number of patients; n is the number of events. Events are defined as the first occurrence of stroke or systemic embolism. Time to event is the time between randomization date and the event first occurrence. The rate is the number of outcome events per 100 patient-years of follow-up. P-value and interaction P-value are from the likelihood ratio test. Hazard ratios were estimated with Cox proportional-hazards regression models fit separately to each subgroup. Hazard ratio confidence interval limits are the profile likelihood limits. Interaction between the effect of apixaban vs. aspirin and subgroups was assessed with the Cox model fit to all patients. AF, atrial fibrillation; CI, confidence interval; eGFR, estimated glomerular filtration rate calculated with the CKD-EPI formula;10 INR, international normalized ratio; TIA, transient ischaemic attack; VKA, vitamin K antagonist.

Discussion

The AVERROES trial demonstrated that apixaban compared with aspirin reduced stroke by 55% and did not increase major bleeding in patients with AF deemed unsuitable for warfarin. In the present analyses, we demonstrate that patients deemed unsuitable because they had previously failed warfarin compared with those who were warfarin naive and expected to be unsuitable were older, heavier, and more likely to be male. They were also more likely to have had a prior stroke or TIA, and less likely to have had heart failure or to be medically undertreated. It is hard to speculate on the reasons that patients who failed VKA therapy seem to have more stroke risk factors than VKA-naive patients, and in fact, the opposite might have anticipated because in the most frail patients VKA therapy may not be attempted. This suggests that other factors, not related to stroke risk such as access to and cost of healthcare, may be more important determinants of prior antithrombotic therapy. For example, 71% of VKA-naive patients were enrolled outside North America or Western Europe, compared with 51% of patients that tried and failed VKA therapy. While some of these characteristics are important risk factors for stroke and bleeding, the differences were rather small and unlikely to be clinically relevant. Moreover, despite the differences in baseline characteristics, the effects of apixaban compared with aspirin were consistent in patients who previously failed warfarin and in those who were expected to be unsuitable. Apixaban compared with aspirin also produced consistent treatment effects among different subgroups of patients who had previously failed warfarin therapy defined by reasons for failure of warfarin treatment, age, renal function, and CHADS2 score.

Patients who have failed warfarin therapy because of bleeding complications or inadequate INR control are usually excluded from trials of anticoagulation and might be expected to have different risks of adverse events when treated with an anticoagulant such as apixaban than those who are expected to be unsuitable for warfarin therapy. One of the major strengths of the present analysis is the large number of these patients included in the trial, with >500 patients having previously experienced bleeding during warfarin therapy. Our data demonstrating uniformly consistent estimates for efficacy and safety irrespective of the reasons for warfarin failure suggest that the results of the AVERROES trial can be broadly generalized to patients who have failed warfarin therapy. The only exception was a statistically significant interaction for major bleeding between randomized treatment and quality of previous INR control that suggested more major bleeding in apixaban-treated patients. To explore if this interaction is truly significant or due to random chance, we looked at this interaction for the composite endpoint of major and clinically relevant non-major bleeding, where no such interaction was observed (see Supplementary material online, Table S1).

The optimal antithrombotic therapy for patients that are unsuitable for VKA therapy remains a challenge. The results of the AVERROES trial show that aspirin monotherapy is inferior to twice daily apixaban8 and accordingly, the most recent update from the European Society of Cardiology guideline now advises against aspirin monotherapy.11 Furthermore, the effect of addition of clopidogrel to aspirin in unsuitable patients was studied in the ACTIVE-A study.12 Although combined clopidogrel and aspirin treatment led to a 30% reduction in stroke compared with aspirin alone, combination therapy also increased the risk of major bleeding by 55%.12 Therefore, apixaban seems a more attractive alternative for patients unsuitable for VKA therapy. Previous comparisons between VKA-naive- and VKA-experienced patients, such as the one from the RE-LY trial,13 show similarly consistent treatment effects of anticoagulant therapy. However, patients in the present analysis were not simply VKA experienced; they were both experienced and unsuitable for VKA therapy. The results in this unsuitable population, therefore, provide valuable additional data that may help to guide these patients.

The early termination and consequent lower than anticipated number of outcome events of the AVERROES trial is the most important limitation of this analysis. Sample size calculations for randomized trials are usually aimed to provide sufficient statistical power for the primary outcome only and thereby, subgroup analyses are generally underpowered. The early termination further limits the power to explore subgroups. Thus, this subgroup analysis within the subgroup of patients who previously failed VKA therapy should be considered exploratory and results should be cautiously interpreted, especially in subgroups with low-outcome events. Nevertheless, the consistency of treatment effects in most key subgroups is re-assuring.

In conclusion, this analysis indicates that the beneficial effects of apixaban vs. aspirin found in the AVERROES trial apply across a wide spectrum of patients, irrespective of the reason for unsuitability of VKA therapy, including patients with prior bleeding episodes and patients with poor INR control.

Funding

The AVERROES trial was supported by Bristol-Myers Squibb and Pfizer.

Conflict of interest: M.C. has received grant support and lecturing fees from Boehringer Ingelheim. D.S. was formerly employed by Bristol-Myers Squibb. J.W.E. has received consulting and lecturing fees and grant support from Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi Sankyo, Eli Lilly, Johnson & Johnson, Novartis, Pfizer, and Portola. S.Y. has received consulting fees and grant support from AstraZeneca, Bristol-Myers Squibb, Boehringer Ingelheim, Glaxo-Smith Kline, Novartis, and Sanofi-Aventis. O.S. has declared to have no conflicts of interest. S.J.C. has received payment for serving on the board, consulting and lecturing fees, and grant support from Boehringer Ingelheim, Bristol-Myers Squibb, Merck, Portola, and Sanofi-Aventis.

References

View Abstract