OUP user menu

★ Editor's choice ★

Apixaban vs. warfarin with concomitant aspirin in patients with atrial fibrillation: insights from the ARISTOTLE trial

John H. Alexander, Renato D. Lopes, Laine Thomas, Marco Alings, Dan Atar, Philip Aylward, Shinya Goto, Michael Hanna, Kurt Huber, Steen Husted, Basil S. Lewis, John J.V. McMurray, Prem Pais, Hubert Pouleur, Philippe Gabriel Steg, Freek W.A. Verheugt, Daniel M. Wojdyla, Christopher B. Granger, Lars Wallentin
DOI: http://dx.doi.org/10.1093/eurheartj/eht445 224-232 First published online: 20 October 2013

Abstract

Aims We assessed the effect of concomitant aspirin use on the efficacy and safety of apixaban compared with warfarin in patients with atrial fibrillation (AF).

Methods and results In ARISTOTLE, 18 201 patients were randomized to apixaban 5 mg twice daily or warfarin. Concomitant aspirin use was left to the discretion of the treating physician. In this predefined analysis, simple and marginal structured models were used to adjust for baseline and time-dependent confounders associated with aspirin use. Outcome measures included stroke or systemic embolism, ischaemic stroke, myocardial infarction, mortality, major bleeding, haemorrhagic stroke, major or clinically relevant non-major bleeding, and any bleeding. On Day 1, 4434 (24%) patients were taking aspirin. Irrespective of concomitant aspirin use, apixaban reduced stroke or systemic embolism [with aspirin: apixaban 1.12% vs. warfarin 1.91%, hazard ratio (HR) 0.58, 95% confidence interval (CI) 0.39–0.85 vs. without aspirin: apixaban 1.11% vs. warfarin 1.32%, HR 0.84, 95% CI 0.66–1.07; P interaction = 0.10] and caused less major bleeding than warfarin (with aspirin: apixaban 3.10% vs. warfarin 3.92%, HR 0.77, 95% CI 0.60–0.99 vs. without aspirin: apixaban 1.82% vs. warfarin 2.78%, HR without aspirin 0.65, 95% CI 0.55–0.78; P interaction = 0.29). Similar results were seen in the subgroups of patients with and without arterial vascular disease.

Conclusion Apixaban had similar beneficial effects on stroke or systemic embolism and major bleeding compared with warfarin, irrespective of concomitant aspirin use.

  • Concomitant medications
  • Aspirin
  • Atrial fibrillation
  • Stroke
  • Systemic embolism
  • Major bleeding

Introduction

Oral anticoagulation is recommended for the prevention of thromboembolic events in patients with atrial fibrillation (AF) at risk for stroke.13 Oral anticoagulants have been shown to be more effective for stroke prevention than antiplatelet therapy with aspirin alone4,5 or aspirin plus clopidogrel.6 However, patients with AF frequently also have coronary artery disease or other indications for aspirin.7,8 Randomized trials and observational studies have demonstrated high rates of bleeding with the combination of aspirin therapy and vitamin K antagonists (VKAs).915 Clinical trials comparing new oral anticoagulants with placebo in patients with recent acute coronary syndromes already on antiplatelet therapy also showed substantial increases in bleeding and inconsistent reductions in ischaemic events.16,17 Therefore, recent consensus recommendations for patients with AF and a coronary stent recommend combining antiplatelet and anticoagulant therapy for up to 12 months and then discontinuing antiplatelet therapy and continuing only anticoagulation in the majority of patients.1820

In the ARISTOTLE trial, in patients with AF at risk for stroke, apixaban was superior to warfarin for stroke prevention and resulted in less major bleeding.21 The details of the efficacy and safety of apixaban compared with warfarin in patients with AF receiving concomitant aspirin have not been described. The objectives of this prespecified analysis from the ARISTOTLE trial were to (i) describe the use over time and dose of concomitant aspirin in patients with AF overall and in the subgroups of patients with and without arterial vascular disease, and (ii) evaluate the efficacy and safety of apixaban compared with warfarin in patients receiving and not receiving concomitant aspirin overall and in the subgroups of patients with and without arterial vascular disease.

Methods

Patient population and study drug

The design and main results of the ARISTOTLE trial have been published previously.21,22 Briefly, ARISTOTLE was a double-blind, double-dummy, randomized clinical trial of 18 201 patients with AF and at least one additional risk factor for stroke or systemic embolism. Risk factors included age ≥75 years, hypertension, diabetes, heart failure, or reduced left-ventricular systolic function, and prior stroke or systemic embolism.

Patients were randomized to warfarin and apixaban placebo (n = 9081) or apixaban 5 mg twice daily and warfarin placebo (n = 9120). Patients with two or three of the following characteristics—age ≥80 years, weight ≤60 kg, creatinine ≥1.5 mg/dL—at baseline were assigned to apixaban 2.5 mg twice daily or matching placebo. Warfarin was adjusted to a target international normalized ratio (INR) of 2.0–3.0 using a blinded encrypted bedside device.

Aspirin

Patients on aspirin ≥165 mg per day and those who required treatment with both aspirin and a P2Y12 receptor antagonist at baseline were not eligible to be enrolled in ARISTOTLE. The use and dosing of aspirin and of P2Y12 receptor antagonists during the trial was left to the discretion of the treating physician. For patients who required aspirin, a ‘low-dose’ (≤165 mg daily) was recommended. Patients who required antiplatelet therapy during the trial and were considered to be at high risk for bleeding could, but were not required to, temporarily discontinue study drug.

Outcomes

The median duration of follow-up in ARISTOTLE was 1.8 years. Outcomes in this analysis include stroke or systemic embolism, ischaemic stroke, myocardial infarction, all-cause mortality, major bleeding classified according to the International Society on Thrombosis and Haemostasis (ISTH) scale, haemorrhagic stroke, major or clinically relevant non-major bleeding, and any bleeding. All endpoints, with the exception of any bleeding, were adjudicated by a clinical events committee blinded to treatment assignment using prespecified criteria.

Statistical analysis

Baseline categorical variables are presented as percentages and continuous variables as medians and 25th, 75th percentiles. Because a significant proportion of patients taking aspirin at baseline stopped it on Day 1, we defined aspirin users as those using aspirin on Day 1. We determined the frequency of aspirin use during the trial overall and among patients with and without a history of arterial vascular disease, which included a history of coronary artery disease, stroke, or peripheral arterial disease. Because we were interested in concomitant aspirin and anticoagulation use, patients were censored when they permanently discontinued study drug. Baseline characteristics were compared using the Wilcoxon rank sum test for continuous variables and the χ2 test for categorical variables. All analyses were performed at the Duke Clinical Research Institute (Durham, NC, USA) using SAS version 9.2 (SAS Institute, Inc., Cary, NC, USA).

Outcomes are presented as the number of events and event rates per year. Hazard ratios (HRs) and 95% confidence intervals (CIs) comparing aspirin users and non-users were derived from a Cox proportional hazards model after adjustment for a prespecified list of baseline variables associated with either the propensity to use aspirin or the outcomes of interest. These variables included age, sex, prior warfarin use, body mass index, prior stroke, left-ventricular ejection fraction ≤40%, diabetes, hypertension, type of AF, prior myocardial infarction, peripheral arterial disease, congestive heart failure, chronic kidney disease, prior percutaneous coronary intervention, prior coronary artery bypass surgery, and history of bleeding. Non-linearity in continuous covariates was handled by fitting restricted cubic splines. Missing data were rare for these covariates (<1.2%) and were handled by single imputation. The interaction between randomized treatment and aspirin use was tested in this Cox proportional hazards model. Hazard ratios for the randomized treatment among aspirin users and non-users were derived from this model with interaction. Analyses were conducted in the overall population and in the subgroups of patients with and without arterial vascular disease.

In addition to adjusting for baseline variables associated with either the propensity to use aspirin or the outcomes of interest, we used marginal structural models to further adjust for potential time-dependent confounders to estimate a HR for the effect of aspirin on outcome, and to test for interaction between aspirin use and the randomized treatment (apixaban vs. warfarin).23 This approach involves a Cox model where aspirin status is a time-dependent covariate (i.e. patients can contribute information to either the aspirin or non-aspirin category depending on whether or not they are taking aspirin at a particular time) and weighting is used to adjust for time-varying confounders (i.e. variables measured during follow-up that may affect both outcome and the chance of receiving aspirin). Post-randomization, time-dependent confounders included whether or not the patient was taking study drug, myocardial infarction, angina, percutaneous coronary intervention, coronary stenting, non-intracranial major bleeding, non-major bleeding, and interactions of prior myocardial infarction by major non-intracranial bleeding, on study drug by angina and by percutaneous coronary intervention, on aspirin in the previous week by angina, myocardial infarction by major non-intracranial bleeding and by non-major bleeding, and age by non-major bleeding. Patients were considered to be taking aspirin in a particular week if they received aspirin for at least 50% of the days of the week. Patient weights were re-calculated at weekly intervals during follow-up in order to balance the case-mix between patients receiving and not receiving aspirin. These analyses were also conducted in the overall population and replicated in the subgroups of patients with and without arterial vascular disease.

The ARISTOTLE trial was funded by Bristol-Myers Squibb (Princeton, NJ, USA) and Pfizer, Inc. (New York, NY, USA) and they participated in the trial design and data collection. The analyses presented here were designed by the authors, performed at the Duke Clinical Research Institute, and interpreted by the authors. All authors, including two (M.H., H.P.) employed by the sponsors, commented on the manuscript. The decision to publish the final manuscript was made by the first author (J.H.A.).

Results

Of the 18 201 patients enrolled in ARISTOTLE, 5632 (31%) were using aspirin at baseline. Of the patients using aspirin at baseline, 1198 (21%) stopped within 1 day of enrolment. Thus, 4434 patients (24%) were using aspirin on Day 1. Figure 1 illustrates the use of concomitant aspirin over time in the overall population and among patients with and without a history of arterial vascular disease. The rate of aspirin use was almost twice as high among patients with a history of arterial vascular disease as among patients without a history of arterial vascular disease. Among patients using aspirin on Day 1, the majority were taking 75 (15%), 81 (30%), or 100 mg (44%) daily. Among patients using aspirin on Day 1, 1134 (25.7%) stopped aspirin during follow-up a median (25th, 75th) of 92 (15, 350) days after randomization. Among patients not using aspirin on Day 1, 731 (5.4%) started aspirin during follow-up a median (25th, 75th) of 264 (112, 485) days after randomization.

Figure 1

Net concomitant aspirin use over time overall and among patients with and without arterial vascular disease. This includes patients who start and who stop aspirin. Arterial vascular disease includes a history of coronary artery disease, stroke, or peripheral arterial disease. Includes data from Day 1 post-randomization until each patient's permanent discontinuation of study drug.

Study drug discontinuation was more frequent among aspirin users than among non-aspirin users (29.7 and 24.6%, P < 0.0001). Among patients randomized to warfarin, the median (25th, 75th) time in therapeutic range (INR = 2.0–3.0) was similar among aspirin users and non-aspirin users [65.1 (50.7, 76.0) and 66.3 (52.0, 76.7)].

Use of antiplatelet agents other than aspirin was rare. On Day 1, 272 (1.5%) patients were receiving a P2Y12 receptor antagonist, predominantly clopidogrel. During the trial, a total of 579 patients (3.2%) received a P2Y12 receptor antagonist in combination with study drug for at least 1 day. Of these, only 135 (0.7%) patients took ‘triple therapy’ with aspirin, a P2Y12 receptor antagonist, and study drug for at least 7 days.

Baseline characteristics

Due to randomization, baseline characteristics were well matched between the apixaban and warfarin groups among aspirin users and non-users at baseline (data not shown). Baseline characteristics of patients using and not using aspirin on Day 1 are shown in Table 1. Aspirin users were more likely to be male, have diabetes or hypertension, and were significantly more likely to have a history of myocardial infarction, percutaneous coronary intervention, coronary artery bypass surgery, and peripheral arterial disease than non-users. Aspirin use was more frequent in North America and less frequent in Europe. Most aspirin users with a history of coronary revascularization had their most recent procedures more than 12 months prior to enrolment. Aspirin users were less likely to have previously used VKAs, had higher CHADS2 scores, and were more likely to have recent onset AF than non-users. Aspirin users were more likely to be using concomitant proton pump inhibitors (17.4%) than non-aspirin users (12.8%).

View this table:
Table 1

Baseline characteristics of aspirin users and non-users

Aspirin users (n = 4434)Non-users (n =13 699)P-value
Age, median (25th, 75th), years70 (64, 76)70 (62, 76)0.0058
Female sex, n (%)1405 (31.7)4990 (36.4)<0.0001
Weight, median (25th, 75th), kg83 (70, 97)82 (70, 95)0.0015
BMI, median (25th, 75th), kg/m228.7 (25.4, 33.1)28.4 (25.2, 32.4)0.0022
Diabetes, n (%)1282 (28.9)3249 (23.7)<0.0001
Hypertension, n (%)3940 (88.9)11 914 (87.0)0.0010
History of CAD,a n (%)2264 (51.1)4354 (31.8)<0.0001
History of MI, n (%)1046 (23.6)1529 (11.2)<0.0001
History of PCI, n (%)744 (16.8)903 (6.6)<0.0001
 Time from most recent PCI0.0452
  <3 months52 (7.0)81 (9.0)
  3–12 months71 (9.6)112 (12.5)
  >12 months618 (83.4)706 (78.5)
 Proportion stent, n (%)518 (70.7)562 (63.4)0.0019
  Proportion DES184 (35.5)203 (36.1)0.8373
  Proportion BMS334 (64.5)359 (63.9)
History of CABG, n (%)582 (13.1)620 (4.5)<0.0001
History of stroke, n (%)501 (11.3)1624 (11.9)0.3172
History of PAD, n (%)291 (6.6)589 (4.4)<0.0001
Heart failure, n (%)1282 (28.9)3791 (27.7)0.1098
LVEF <40%, n (%)752 (17.0)1858 (13.6)<0.0001
Chronic kidney disease, n (%)331 (7.5)1174 (8.6)0.0208
Prior VKA use, n (%)1970 (44.4)8398 (61.3)<0.0001
CHADS2 score, n (%)<0.0001
 11391 (31.4)4763 (34.8)
 21616 (36.4)4882 (35.6)
 31427 (32.2)4054 (29.6)
AF type, n (%)0.1666
 Paroxysmal707 (15.9)2066 (15.1)
 Persistent or permanent3727 (84.1)11 630 (84.9)
Time from first onset of AF, n (%)<0.0001
 <6 months1501 (33.9)3543 (26.0)
 6–24 months903 (20.4)2645 (19.4)
 >24 months2021 (45.7)7459 (54.7)
Region of enrolment<0.0001
 North America1430 (32.2)3007 (21.9)
 Latin America916 (20.7)2542 (18.6)
 Europe1360 (30.7)5971 (43.6)
 Asia Pacific728 (16.4)2179 (15.9)
Randomized group0.6917
 Apixaban2233 (50.4)6852 (50.0)
 Warfarin2201 (49.6)6847 (50.0)
  • AF, atrial fibrillation; BMI, body mass index; BMS, bare-metal stent; CABG, coronary artery bypass graft surgery; CAD, coronary artery disease; DES, drug-eluting stent; LVEF, left-ventricular ejection fraction; MI, myocardial infarction; PAD, peripheral arterial disease; PCI, percutaneous coronary intervention; VKA, vitamin K antagonist.

  • aDefined as documented history of coronary artery disease or prior MI, PCI, or CABG.

Effect of apixaban vs. warfarin among aspirin users and non-users

Ischaemic events

Outcomes among aspirin users and non-users randomized to apixaban vs. warfarin are shown in Figure 2. Compared with warfarin, apixaban resulted in similar reductions in stroke or systemic embolism among aspirin users and non-users (with aspirin: apixaban 1.12% vs. warfarin 1.91%, HR 0.58, 95% CI 0.39–0.85 vs. without aspirin: apixaban 1.11% vs. warfarin 1.32%, HR 0.84, 95% CI 0.66–1.07; P interaction = 0.10). There was also no evidence of a differential effect of apixaban compared with warfarin on ischaemic stroke (P interaction = 0.19), myocardial infarction (P interaction = 0.19), or death (P interaction = 0.23) among aspirin users and non-users.

Figure 2

Event rates per year, adjusted hazard ratios, and 95% confidence intervals for the effect of apixaban vs. warfarin among patients using and not using aspirin in the overall population (A), and in the subgroups of patients with (B) and without (C) arterial vascular disease. Adjusted for baseline covariates including age, sex, prior warfarin or VKA use, body mass index, prior stroke, left-ventricular ejection fraction ≤40%, diabetes, hypertension, type of AF, prior myocardial infarction, peripheral arterial disease, congestive heart failure, chronic kidney disease, prior percutaneous coronary intervention, prior coronary artery bypass surgery, and history of bleeding. Numbers are event counts and event rates per year. CI, confidence interval; CRNM, clinically relevant non-major; HR, hazard ratio.

Similar beneficial effects of apixaban compared with warfarin were seen in the cohorts of patients with and without a history of arterial vascular disease (Figure 2B and C).

Results from marginal structural model analyses accounting for whether a patient was actually taking aspirin at the time of their event resulted in similar findings (Table 2). Using these techniques, apixaban had a consistent effect among aspirin users and non-users on stroke or systemic embolism, ischaemic stroke, myocardial infarction, and death both in the overall population and among the subgroups of patients with and without a history of arterial vascular disease (Table 2).

View this table:
Table 2

Marginal structural model analyses assessing effect of apixaban vs. warfarin in aspirin users and non-users overall and in subgroups of patients with and without a history of arterial vascular disease

Aspirin users Apixaban vs. Warfarin HR (95% CI)Non-users Apixaban vs. Warfarin HR (95% CI)Interaction P-value
Stroke or systemic embolism0.59 (0.40–0.87)0.85 (0.67–1.08)0.1114
 With arterial vascular disease0.58 (0.37–0.92)0.96 (0.71–1.30)0.0772
 Without arterial vascular disease0.60 (0.28–1.28)0.72 (0.49–1.05)0.6817
Ischaemic stroke0.72 (0.45–1.88)0.99 (0.74–1.32)0.2685
 With arterial vascular disease0.71 (0.41–1.23)1.09 (0.77–1.56)0.2013
 Without arterial vascular disease0.75 (0.29–1.93)0.83 (0.51–1.36)0.8515
MI1.15 (0.70–1.88)0.77 (0.51–1.14)0.2144
 With arterial vascular disease1.08 (0.62–1.88)0.73 (0.45–1.19)0.3045
 Without arterial vascular disease1.38 (0.45–4.25)0.84 (0.41–1.72)0.4644
Death1.03 (0.72–1.46)0.85 (0.70–1.03)0.3727
 With arterial vascular disease0.93 (0.59–1.46)1.00 (0.77–1.29)0.7937
 Without arterial vascular disease1.17 (0.66–2.08)0.71 (0.53–0.95)0.1277
Major bleeding0.74 (0.57–0.95)0.68 (0.57–0.81)0.6312
 With arterial vascular disease0.77 (0.57–1.06)0.74 (0.58–0.94)0.8198
 Without arterial vascular disease0.66 (0.42–1.04)0.63 (0.50–0.81)0.8754
Haemorrhagic stroke0.43 (0.21–0.87)0.53 (0.32–0.87)0.6332
 With arterial vascular disease0.44 (0.19–1.02)0.50 (0.25–1.01)0.8204
 Without arterial vascular disease0.46 (0.12–1.81)0.57 (0.28–1.15)0.7929
Major or CRNM bleeding0.73 (0.60–0.89)0.67 (0.59–0.76)0.4623
 With arterial vascular disease0.78 (0.61–0.99)0.75 (0.63–0.89)0.8207
 Without arterial vascular disease0.64 (0.45–0.90)0.60 (0.50–0.71)0.7580
Any bleeding0.69 (0.62–0.77)0.72 (0.68–0.77)0.4857
 With arterial vascular disease0.71 (0.62–0.81)0.72 (0.65–0.79)0.9224
 Without arterial vascular disease0.65 (0.54–0.78)0.73 (0.67–0.79)0.2811
  • CI, confidence interval; CRNM, clinically relevant non-major; HR, hazard ratio; MI, myocardial infarction.

Bleeding

Compared with warfarin, apixaban resulted in consistent reductions in bleeding among both aspirin users and non-users (Figure 2A). There were similar and statistically significant reductions in major bleeding (with aspirin: apixaban 3.10% vs. warfarin 3.92%, HR 0.77, 95% CI 0.60–0.99 vs. without aspirin: apixaban 1.82% vs. warfarin 2.78%, HR without aspirin 0.65, 95% CI 0.55–0.78; P interaction = 0.29). Consistent benefits were seen with apixaban compared with warfarin among both aspirin users and non-users for haemorrhagic stroke (P interaction = 0.52), major or clinically relevant non-major bleeding (P interaction = 0.15), and any bleeding (P interaction = 0.70).

Similar beneficial effects of apixaban compared with warfarin on bleeding were seen in the subgroups of patients with and without a history of arterial vascular disease (Figure 2B and C).

For bleeding, results from marginal structural model analyses that accounted for whether a patient was actually taking aspirin at the time of their bleeding event resulted in similar findings (Table 2). Apixaban had a consistent effect among aspirin users and non-users on causing less major bleeding, haemorrhagic stroke, major or clinically relevant non-major bleeding, and any bleeding both in the overall population and in the subgroups of patients with and without a history of arterial vascular disease (Table 2).

Outcomes among aspirin users and non-users

Outcomes among aspirin users and non-users are shown in Table 3. After adjustment for the propensity to use aspirin and for differences in prognostically important baseline variables associated with aspirin use, aspirin users had similar rates of stroke or systemic embolism, ischaemic stroke, myocardial infarction, and death and higher rates of bleeding when compared with non-users. After adjusting for both baseline confounders and post-randomization variables associated with aspirin use, aspirin users tended to have higher rates of stroke or systemic embolism, ischaemic stroke, and myocardial infarction, similar rates of death, and higher rates of bleeding than non-aspirin users. In general, similar results were seen in the subgroups of patients with and without a history of arterial vascular disease (Table 3); although in patients with arterial vascular disease, aspirin users tended to have higher rates of myocardial infarction and lower rates of death than non-aspirin users.

View this table:
Table 3

Outcomes among aspirin users and non-users in all patients and in subgroups of patients with and without a history of arterial vascular disease

Aspirin usersa (n = 4434)Aspirin non-usersa (n = 13 699)Adjusted HR (95% CI)MSM adjusted HR (95% CI)
Stroke or systemic embolism108 (1.51)276 (1.22)1.18 (0.94–1.49)1.46 (1.15–1.85)
 With arterial vascular disease78 (1.76)168 (1.62)1.05 (0.80–1.38)1.48 (1.11–1.97)
 Without arterial vascular disease30 (1.10)108 (0.87)1.21 (0.80–1.84)1.40 (0.91–2.14)
Ischaemic stroke69 (0.96)189 (0.83)1.08 (0.80–1.43)1.40 (1.05–1.88)
 With arterial vascular disease51 (1.15)125 (1.21)0.93 (0.67–1.30)1.35 (0.96–1.91)
 Without arterial vascular disease18 (0.66)64 (0.52)1.23 (0.72–2.10)1.48 (0.85–2.56)
MI59 (0.82)104 (0.46)1.28 (0.92–1.80)1.72 (1.21–2.44)
 With arterial vascular disease47 (1.06)72 (0.69)1.47 (1.01–2.14)1.56 (1.05–2.32)
 Without arterial vascular disease12 (0.44)32 (0.26)1.61 (0.81–3.19)2.09 (1.08–4.07)
Death135 (1.87)411 (1.80)0.89 (0.73–1.09)1.00 (0.81–1.24)
 With arterial vascular disease82 (1.83)226 (2.17)0.77 (0.60–1.00)0.87 (0.67–1.14)
 Without arterial vascular disease53 (1.94)185 (1.49)1.17 (0.85–1.60)1.27 (0.91–1.77)
Major bleeding252 (3.50)528 (2.30)1.41 (1.21–1.66)1.65 (1.40–1.94)
 With arterial vascular disease168 (3.77)266 (2.54)1.46 (1.20–1.78)1.75 (1.41–2.17)
 Without arterial vascular disease84 (3.07)262 (2.10)1.42 (1.10–1.83)1.52 (1.17–1.97)
Haemorrhagic stroke84 (0.47)72 (0.32)1.47 (0.96–2.25)2.08 (1.35–3.21)
 With arterial vascular disease24 (0.54)38 (0.36)1.36 (0.81–2.29)2.47 (1.41–4.30)
 Without arterial vascular disease10 (0.37)34 (0.27)1.23 (0.60–2.54)1.50 (0.73–3.12)
Major or CRNM bleeding445 (6.34)1030 (4.57)1.27 (1.13–1.43)1.36 (1.21–1.54)
 With arterial vascular disease299 (6.91)507 (4.93)1.36 (1.18–1.57)1.39 (1.19–1.63)
 Without arterial vascular disease146 (5.43)523 (4.27)1.25 (1.03–1.51)1.33 (1.09–1.62)
Any bleeding1541 (27.38)3837 (20.02)1.30 (1.22–1.39)1.31 (1.24–1.41)
 With arterial vascular disease995 (28.8)1788 (20.3)1.39 (1.29–1.51)1.38 (1.27–1.51)
 Without arterial vascular disease546 (25.2)2049 (19.8)1.28 (1.16–1.41)1.26 (1.14–1.40)
  • CI, confidence interval; CRNM, clinically relevant non-major; HR, hazard ratio; MI, myocardial infarction; MSM, marginal structural model.

  • aNumbers are event counts and event rates per year.

Discussion

In the ARISTOTLE trial, apixaban reduced stroke or systemic embolism and caused less bleeding than warfarin in patients with AF and at least one risk factor for stroke.21 In this analysis of the concomitant use of aspirin and oral anticoagulation from ARISTOTLE, the benefits of apixaban over warfarin in terms of reducing stroke or systemic embolism, causing less bleeding, and reducing mortality were consistent irrespective of concomitant aspirin use. We found similar results, with important benefits of apixaban over warfarin, in the subgroups of patients with and without arterial vascular disease. If there is a strong indication for a combination of aspirin and oral anticoagulation, apixaban seems to be a safer alternative than warfarin in patients with AF irrespective of concomitant aspirin use.

In ARISTOTLE, aspirin was used in combination with long-term anticoagulation in 20–25% of patients. Use of concomitant aspirin was twice as common among patients with a history of arterial vascular disease as in patients without arterial vascular disease. Among patients with arterial vascular disease, most did not have a recent acute coronary event or revascularization procedure and even among patients without a history of arterial vascular disease, a substantial number were taking aspirin. Current consensus recommendations advise against using concomitant aspirin therapy in patients with AF receiving oral anticoagulation without vascular disease or with a remote (>12 month) history of an acute coronary event or coronary stent.1820,24

The use of aspirin in ARISTOTLE was not randomized and was left to the discretion of the treating physician. Still we were interested in getting as close as possible to establishing causal relationships between concomitant aspirin use and the effects of apixaban vs. warfarin on both ischaemic events and bleeding in patients with AF. Recognizing that patients started and discontinued aspirin and interrupted study drug during the trial, we evaluated outcomes after adjusting for both baseline and time-dependent covariates that were associated with aspirin use where the patients contributed to either the aspirin or non-aspirin category depending on aspirin use at a different time, and weighting was used to adjust for time-varying confounders.

This is a non-randomized comparison of aspirin users and non-users. Patients receiving aspirin were different and had a higher risk for both ischaemic events and bleeding than patients not receiving aspirin. The relationship between aspirin use and bleeding is consistent with the known risks of aspirin and the results of prior studies and meta-analyses of randomized trials.15,2528 Even after adjustment for measured confounders, however, we observed a trend towards a higher rate of myocardial infarction and ischaemic stroke in patients receiving aspirin. Given the known beneficial effects of aspirin on the prevention of myocardial infarction and ischaemic stroke in other populations, this may either represent residual confounding or mean that aspirin is not beneficial in patients with AF who are also taking an anticoagulant.29 The consistently and substantially higher rates of bleeding with concomitant use of aspirin and oral anticoagulation confirm prior findings915 and should lead to a careful assessment of the indications for aspirin in patients with AF who are receiving oral anticoagulation.

This analysis addresses only aspirin, predominantly ‘low-dose’ aspirin, and not other antiplatelet regimens. There are substantial proportions of patients with AF who have either acute coronary syndromes or who undergo percutaneous coronary intervention. Clopidogrel use was an exclusion criterion at randomization and only started in a small proportion of patients included in the ARISTOTLE trial thus limiting our ability to assess the outcomes associated with concomitant apixaban or warfarin and either clopidogrel or dual antiplatelet therapy. Additional randomized clinical trials are needed to determine the ideal antithrombotic regimens for patients with AF and indications for antiplatelet therapy.

Limitations

This analysis has a number of limitations. First, we were only able to assess concomitant ‘low-dose’ aspirin use and not other antiplatelet regimens. Secondly, the use of aspirin was neither randomized nor blinded. Although we adjusted for potential measured confounders, additional confounding almost certainly exists. Thirdly, the fact that aspirin use was not blinded may have led to differential post-randomization use of aspirin among patients assigned to apixaban and warfarin. Fourthly, as with all subgroup analyses, there may be limited power to detect an interaction between apixaban vs. warfarin and aspirin use. Fifthly, although we recorded when patients started and stopped aspirin, we did not collect the reasons for changes in aspirin therapy. Finally, these results from a randomized clinical trial may not be generalizable to other populations.

Conclusions

In ARISTOTLE, concomitant aspirin was used in ∼20–25% of patients with AF treated with an anticoagulant and was associated with a higher risk of bleeding. We observed similar effects of apixaban, compared with warfarin, on stroke or systemic embolism, major bleeding, or mortality irrespective of concomitant aspirin use. Adequately powered randomized clinical trials are needed to better define the optimal antithrombotic regimen and its duration in patients with both AF and atherosclerotic coronary artery disease, and particularly in those with an acute coronary syndrome or recent coronary stenting.

Funding

This work was supported by Bristol-Myers Squibb and Pfizer.

Conflict of interest: The ARISTOTLE trial was funded by Bristol-Myers Squibb and Pfizer. J.H.A.: receives sponsored research support through Duke University from Bristol-Myers Squibb and Pfizer; serves or has served as a consultant to Bristol-Myers Squibb and Pfizer. R.D.L.: grants from Bristol-Myers Squibb, AstraZeneca, Boehringer Ingelheim, and Daiichi Sankyo; consulting fees from Bristol-Myers Squibb. L.T.: none. M.A.: consulting fees from Bayer AG, Boehringer-Ingelheim, MSD, and Sanofi-Aventis; travel support from Boston Scientific, Bristol-Myers Squibb, and St Jude Medical. D.A.: consulting fees or honoraria from Bristol-Myers Squibb. P.A.: research grants, honoraria, and advisory boards for Bristol-Myers Squibb, Pfizer, Bayer, Johnson & Johnson, Boehringer Ingelheim, Daiichi Sankyo, AstraZeneca, Sanofi-Aventis, Merck, Eli Lilly, and The Medicines Company. S.G.: board member of Bristol-Myers Squibb and Sanofi-Aventis; grants from Boehringer Ingelheim, Otsuka, Eisai, Sanofi-Aventis, and Daiichi Sankyo; consulting fees from Eisai; lecture fees from Eisai, Otsuka, Daiichi Sankyo, Sanofi-Aventis, Bayer, Novartis, AstraZeneca, Asteras, Pfizer, Medtronics-Japan, Tanabe-Mitsubishi, Takeda, Mochida, and MSD; payments from Bayer and Sanofi-Aventis for developing educational presentations. M.H.: full-time employee of Bristol-Myers Squibb and receives stock as part of his compensation. K.H.: lecture fees from AstraZeneca, Bristol-Myers Squibb/Pfizer, Boehringer Ingelheim, Bayer, Daiichi Sankyo, and Sanofi-Aventis. S.H.: advisory board membership for AstraZeneca, Bristol-Myers Squibb, Pfizer, and Bayer; research support from GlaxoSmithKline, Pfizer, and Sanofi-Aventis. B.S.L.: advisory board member for Bayer HealthCare. J.J.V.M.: none. P.P.: none. H.P.: full-time Pfizer employee and owns stock in this company. P.G.S.: travel support from Bristol-Myers Squibb; board membership for Bayer, Bristol-Myers Squibb/Pfizer, AstraZeneca, and Boehringer Ingelheim; consulting fees from Bristol-Myers Squibb, Eisai, Ablynx, Amarin, Astellas, Eil Lilly, Medtronic, Novartis, Roche, Servier, The Medicines Company, Sanofi, and AstraZeneca; grants from Servier, Sanofi, and New York University School of Medicine; and lecture fees from Pfizer, Amgen, Otsuka, and Aterovax. F.W.A.V.: lecture fees from Bayer and AstraZeneca; consulting fees from Bayer and Daiichi Sankyo. D.M.W.: none. C.B.G.: grants from AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, the Medtronic Foundation, Merck, Sanofi-Aventis, Astellas, and The Medicines Company; consulting fees from AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Hoffmann-La Roche, Novartis, Otsuka Pharmaceutical, Sanofi-Aventis, and The Medicines Company; support from the Medtronic Foundation and Merck for travel, accommodations, or meeting expenses. L.W.: grants from Bristol-Myers Squibb, Pfizer, AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Schering-Plough, and Merck; consulting fees from Regado Biosciences, Portola, CSL Behring, Athera Biotechnologies, Boehringer Ingelheim, AstraZeneca, and GlaxoSmithKline; lecture fees from Bristol-Myers Squibb, Pfizer, AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Schering-Plough, and Merck.

Acknowledgements

The authors acknowledge Elizabeth Cook from the Duke Clinical Research Institute for her editorial support with this manuscript.

References

View Abstract