OUP user menu

★ Editor's choice ★

Management of major bleeding events in patients treated with rivaroxaban vs. warfarin: results from the ROCKET AF trial

Jonathan P. Piccini, Jyotsna Garg, Manesh R. Patel, Yuliya Lokhnygina, Shaun G. Goodman, Richard C. Becker, Scott D. Berkowitz, Günter Breithardt, Werner Hacke, Jonathan L. Halperin, Graeme J. Hankey, Christopher C. Nessel, Kenneth W. Mahaffey, Daniel E. Singer, Robert M. Califf, Keith A. A. Fox,
DOI: http://dx.doi.org/10.1093/eurheartj/ehu083 1873-1880 First published online: 21 March 2014

Abstract

Aims There are no data regarding management and outcomes of major bleeding events in patients treated with oral factor Xa inhibitors.

Methods and results Using data from ROCKET AF, we analysed the management and outcomes of major bleeding overall and according to the randomized treatment. During a median follow-up of 1.9 years, 779 (5.5%) patients experienced major bleeding at a rate of 3.52 events/100 patient-years with a similar event rate in each arm (n = 395 rivaroxaban vs. n = 384 warfarin). The median number of transfused packed red blood cells (PRBC) per episode was similar in both arms [2 (25th, 75th: 2, 4) units]. Overall, few transfusions of whole blood (n = 14), platelets (n = 10), or cryoprecipitate (n = 2) were used. Transfusion of fresh frozen plasma (FFP) was significantly less in the rivaroxaban arm (n = 45 vs. n = 81 units) after adjustment for covariates [odds ratio (OR) 0.43 (95% CI 0.29–0.66); P < 0.0001]. Prothrombin complex concentrates (PCC) were administered less in the rivaroxaban arm (n = 4 vs. n = 9). Outcomes after major bleeding, including stroke or non-central nervous system embolism (4.7% rivaroxaban vs. 5.4% warfarin; HR 0.89; 95% CI 0.42–1.88) and all-cause death (20.4% rivaroxaban vs. 26.1% warfarin; HR 0.69, 95% CI 0.46–1.04) were similar in patients treated with rivaroxaban and warfarin (interaction P = 0.51 and 0.11).

Conclusion Among high-risk patients with atrial fibrillation who experienced major bleeding in ROCKET AF, the use of FFP and PCC was less among those allocated rivaroxaban compared with warfarin. However, use of PRBCs and outcomes after bleeding were similar among patients randomized to rivaroxaban or to warfarin.

  • Factor Xa inhibitor
  • Bleeding
  • Atrial fibrillation

Introduction

There is limited information on the management of bleeding events in patients treated with factor Xa inhibitors. Additionally, there is uncertainty about the extent to which generic approaches to bleeding management or specialized coagulation products are required in such patients. In the Rivaroxaban Once-daily oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET AF), rivaroxaban was shown to be non-inferior to warfarin for the prevention of stroke and systemic embolism, with less intracranial and fatal bleeding.1

Novel agents, including rivaroxaban, have several advantages, including shorter half-lives, more rapid clearance from the circulation, and fewer drug–drug interactions compared with vitamin K antagonists (VKA). Nevertheless, questions remain about the management and outcomes of major bleeding events in patients treated with factor Xa inhibitors. Among the concerns expressed by clinicians is whether or not specific ‘reversal’ agents or specialized blood coagulation products may be needed to manage patients following major bleeding.2 The goal of this analysis is to describe the management and outcomes of major bleeding in the ROCKET AF trial. Further, we sought to determine whether management and subsequent outcomes differed between those treated with rivaroxaban vs. warfarin.

Methods

The rationale and design of the ROCKET AF trial has been previously reported.3 ROCKET AF was a multicentre, international, double-blind, double-dummy, event-driven, non-inferiority trial designed to compare the efficacy and safety of rivaroxaban vs. warfarin in patients with non-valvular AF. ROCKET AF randomized 14 264 patients at 1178 sites with AF at moderate-to-high risk for stroke. Patients were required to have a history of stroke, transient ischaemic attack (TIA) or systemic embolism, or at least two of the following: heart failure or a left ventricular ejection fraction ≤35%, hypertension, age ≥75 years, or diabetes mellitus. Furthermore, the proportion of patients who had not had a previous ischaemic stroke, TIA, or systemic embolism, and who had no more than two risk factors was limited to 10% of the cohort for each region; the remainder of patients was required to have prior TIA, stroke, or systemic embolism or more than two risk factors. Exclusion criteria included: active internal bleeding, conditions predisposing to increased bleeding risk (e.g. major surgical procedure or trauma ≤30 days; clinically significant gastrointestinal bleeding ≤6 months; history of intracranial, intraocular, spinal, or atraumatic intra-articular bleeding; bleeding diathesis; known intracranial neoplasm, arteriovenous malformation, or aneurysm; platelet count <90 000/µL at the screening visit; sustained uncontrolled hypertension), aspirin >100 mg daily or aspirin in combination with thienopyridines, anticipated need for chronic treatment with a non-steroidal anti-inflammatory drug, anaemia (haemoglobin <10 g/dL) at the screening visit, calculated creatinine clearance <30 mL/min at the screening visit, significant liver disease (e.g. acute clinical hepatitis, chronic active hepatitis, cirrhosis), or alanine aminotransferase greater than three times the upper limit of normal.

Patients were randomized to receive rivaroxaban 20 mg daily (or 15 mg daily in patients with a creatinine clearance of 30–49 mL/min) or adjusted-dose warfarin [target international normalized ratio (INR), 2.0–3.0]. A point-of-care device was used to generate encrypted values that were sent to an independent study monitor, who provided sites with either real INR values (for patients in the warfarin group in order to adjust the dose) or sham values (for patients in the rivaroxaban group receiving placebo warfarin). All patients provided written informed consent.

Study population

The analysed population was the on-treatment (or ‘safety’) population: patients who received at least 1 dose of study drug. The analysis excludes 50 patients from 1 Good Clinical Practice-violating site. Sites and investigators were instructed to manage major bleeding events with cessation of study medication, discontinuation of antiplatelet drugs, and empiric administration of vitamin K or FFP when needed. Sites were advised that factor concentrates might be helpful in cases of major or life-threatening bleeding refractory to FFP.

Event and endpoint definitions

The clinical event of interest was major bleeding defined as clinically overt bleeding associated with any of the following: fatal outcome, involvement of a critical anatomic site (intracranial, spinal, ocular, pericardial, articular, retroperitoneal, or intramuscular with compartment syndrome), fall in haemoglobin concentration >2 g/dL, transfusion of >2 units of whole blood or packed red blood cells (PRBC), or permanent disability.4 A clinical events committee was blinded to treatment assignment and adjudicated all bleeding events using the primary source data and documents. The committee adjudicated both the anatomic site and severity of the bleed using the International Society on Thrombosis and Haemostasis (ISTH) criteria.4

Efficacy endpoints were considered from randomization until site notification and included all-cause death, stroke, or non-central nervous system (CNS) embolism, myocardial infarction, and surgical intervention for bleeding (e.g. haematoma evacuation, fasciotomy, or surgery for correction of haemorrhage). Safety outcomes were measured during the safety period (from first drug dose to last dose +2 days), unless specified otherwise. Safety outcomes included the duration of hospitalization and transfusion, including all products and factors.

Statistical methods

Baseline characteristics are presented as median (25th, 75th percentiles) for continuous variables and as percent (number) for categorical variables, and are summarized by treatment for subjects who experienced at least one major bleed and for those that did not have any major bleed. A subject could have experienced more than one major bleed; hence, information from all bleeds in the study was used to summarize location of major bleeds, duration of hospitalization, transfusion information after a major bleed, and pharmacologic management after a major bleed by treatment.

A generalized estimating equation (GEE) logistic regression model was utilized to find the important predictors of receipt of FFP or cryoprecipitate transfusions following a major bleed. Candidate covariates included randomized treatment, sex, race, prior aspirin, prior VKA, prior stroke/TIA, history of heart failure, diabetes, hypertension, age, body mass index, creatinine clearance, and days to major bleed. Model selection was performed using stepwise selection with P = 0.05 as entry and removal criteria.

Outcomes occurring after major bleed were summarized as rates and time to the outcomes were summarized as median (25th, 75th percentiles), with the date of major bleed being day 1. Hazard ratios for randomized treatment were estimated using adjusted proportional hazards model for the outcome of interest where major bleed was included as a time-dependent covariate. In addition to major bleed, the models included treatment, treatment by major bleed interaction, sex, history of chronic obstructive pulmonary disease, history of stroke or TIA, history of peripheral arterial disease, atrial fibrillation type, history of heart failure, region, history of alcohol use, age, diastolic blood pressure, systolic blood pressure, heart rate, body mass index, creatinine clearance, and interactions of each of these covariates with treatment.

All analyses were performed using SAS software, version 9.2 (SAS Institute Inc., Cary, NC, USA).

Results

Baseline characteristics

Among the 14 143 randomized patients in this analysis (n = 50 excluded from the GCP violating site), 779 experienced major bleeding (5.5%), 57 experienced 2 major bleeds (0.4%), and 2 patients had more than 2 bleeds (0.01%). Figure 1 illustrates the time to major bleeding according to the randomized treatment. The major bleeding rate was 3.52 events per 100 patient-years. Among 779 bleeding events, 775 (99.5%) occurred while patients were taking the assigned study medication. The baseline characteristics of patients with bleeding events stratified by both the classification of the bleeding event and the treatment assignment are shown in Table 1. The median CHADS2 score was 3 (25th, 75th: 3, 4) in those patients who experienced a major bleed and those who did not. HAS-BLED scores were similar in those with and without major bleeding [3 (3, 4) vs. 3 (3, 4)]. The baseline characteristics of those patients with a major bleeding event, including creatinine clearance, were similar in those randomized to rivaroxaban and warfarin, except those patients with a major bleed on rivaroxaban were less frequently Asian (11.2 vs. 18.3%) and more frequently white (84.3 vs. 78.1%) compared with those on warfarin.

View this table:
Table 1

Baseline characteristics of ROCKET AF patients according to major bleeding events

CharacteristicMajor bleed (N = 779)No major bleedsa (N = 13 364)Interaction P-value
Rivaroxaban (N = 395)Warfarin (N = 384)Rivaroxaban (N = 6666)Warfarin (N = 6698)
Age, median (25th–75th), years76 (69–80)75 (68–79)73 (65–78)72 (65–78)0.2124
Female sex, no. (%)135 (34.2%)132 (34.4%)2656 (39.8%)2667 (39.8%)0.7159
Race, no. (%)0.0231
 White332 (84.3%)299 (78.1%)5524 (82.9%)5610 (83.9%)
 Asian44 (11.2%)70 (18.3%)850 (12.8%)817 (12.2%)
 Black6 (1.5%)3 (0.8%)88 (1.3%)82 (1.2%)
 Other12 (3.0%)11 (2.9%)198 (3.0%)180 (2.7%)
BMI, median (25th–75th), kg/m228.4 (25.4–32.2)27.7 (24.9–31.7)28.3 (25.1–32.1)28.1 (25.1–31.8)0.1268
Systolic BP, median (25th–75th), mm Hg130 (120–140)130 (120–142)130 (120–140)130 (120–140)0.4818
Diastolic BP, median (25th–75th), mm Hg78 (70–80)80 (70–85)80 (70–85)80 (70–85)0.0283
Type of AF, no. (%)0.9604
 Persistent323 (81.8%)314 (81.8%)5416 (81.2%)5409 (80.8%)
 Paroxysmal66 (16.7%)64 (16.7%)1162 (17.4%)1195 (17.8%)
 Newly diagnosed6 (1.5%)6 (1.6%)88 (1.3%)94 (1.4%)
Previous medication use, no. (%)
 Aspirin171 (43.3%)158 (41.1%)2396 (35.9%)2448 (36.5%)0.5113
 VKA270 (68.4%)248 (64.6%)4131 (62.0%)4189 (62.5%)0.1505
 CHADS2 score, median (25th–75th)3 (3–4)3 (3–4)3 (3–4)3 (3–4)0.4955
CHADS2 score, no. (%)
 258 (14.7%)49 (12.8%)864 (13.0%)882 (13.2%)
 3171 (43.3%)172 (44.8%)2854 (42.8%)2959 (44.2%)
 4115 (29.1%)109 (28.4%)1958 (29.4%)1879 (28.1%)
 543 (10.9%)48 (12.5%)875 (13.1%)827 (12.3%)
 68 (2.0%)6 (1.6%)114 (1.7%)149 (2.2%)
HAS-BLED score, median (25th–75th)3 (2–4)3 (2–4)3 (2–3)3 (2–3)0.9141
HAS-BLED score, no. (%)
 01 (0.3%)0 (0.0%)43 (0.6%)47 (0.7%)
 125 (6.3%)20 (5.2%)461 (6.9%)466 (7.0%)
 2103 (26.1%)98 (25.5%)2047 (30.7%)1981 (29.6%)
 3161 (40.9%)167 (43.5%)2750 (41.3%)2795 (41.8%)
 486 (21.8%)84 (21.9%)1188 (17.8%)1215 (18.2%)
 517 (4.3%)14 (3.6%)164 (2.5%)178 (2.7%)
 61 (0.3%)1 (0.3%)6 (0.1%)10 (0.1%)
Medical history
 Previous stroke, SE, TIA, no. (%)186 (47.1%)186 (48.4%)3695 (55.4%)3683 (55.0%)0.6112
 Stroke/TIA, no. (%)178 (45.1%)183 (47.7%)3544 (53.2%)3509 (52.4%)0.3255
 Non-CNS SE, no. (%)11 (2.8%)9 (2.3%)262 (3.9%)268 (4.0%)0.6678
 Heart failure, no. (%)233 (59.0%)231 (60.2%)4195 (62.9%)4178 (62.4%)0.7214
 Coronary artery disease, no. (%)166 (42.0%)151 (39.3%)2478 (37.2%)2632 (39.3%)0.1507
 Hypertension, no. (%)356 (90.1%)347 (90.4%)6016 (90.2%)6082 (90.8%)0.9664
 Diabetes mellitus, no. (%)165 (41.8%)168 (43.8%)2677 (40.2%)2625 (39.2%)0.4193
 Previous MI, no. (%)82 (20.8%)69 (18.0%)1087 (16.3%)1200 (17.9%)0.1020
 Peripheral vascular disease, no. (%)35 (8.9%)24 (6.3%)361 (5.4%)409 (6.1%)0.0594
  Chronic obstructive pulmonary disease, no. (%)65 (16.5%)46 (12.0%)679 (10.2%)687 (10.3%)0.1091
 Creatinine clearance, median (25th–75th)64 (49–85)62 (49–77)68 (52–88)68 (52–87)0.0382
 Creatinine clearance <50 mL/min, no. (%)99 (25.1%)99 (25.8%)1386 (20.8%)1357 (20.3%)0.7869
 LV dysfunction (<35% EF), no. (%)4 (3.2%)2 (1.6%)25 (1.0%)23 (1.0%)0.4010
 Alcohol use, no. (%)133 (33.7%)143 (37.2%)2380 (35.7%)2348 (35.1%)0.3565
 Smoking, no. (%)169 (42.8%)144 (37.5%)2270 (34.1%)2166 (32.3%)0.2401
  • AF, atrial fibrillation; BMI, body mass index; BP, blood pressure; CAD, coronary artery disease; CHF, congestive heart failure; CNS, central nervous system; EF, ejection fraction; LV, left ventricular; MI, myocardial infarction; SE, systemic embolism; TIA, transient ischaemic attack; VKA, vitamin K antagonist.

  • aNo major bleeds includes patients with non-major clinically relevant bleeding, minimal bleeding, or no bleeding.

Figure 1

Kaplan–Meier plot demonstrating time to ISTH major bleeding events according to the randomized treatment (rivaroxaban or warfarin).

Location of major bleeding

The distribution of multiple bleeds and the location of major bleeding events are shown in Table 2. The occurrence of two or more major bleed events was infrequent in the rivaroxaban and warfarin patients (0.5 vs. 0.0%). Major bleeding in the rivaroxaban arm was more frequently located in the upper gastrointestinal tract (38.1 vs. 25.7%), less commonly intracranial (12.8 vs. 20.5%), and less commonly intraocular or retinal (4.4 vs. 6.6%).

View this table:
Table 2

Location of major bleeds by randomized treatmenta

CharacteristicRivaroxaban (n = 431)Warfarin (n = 409)
Number of major bleedsb
 1361 (91.4%)359 (93.5%)
 232 (8.1%)25 (6.5%)
 >22 (0.5%)0 (0.0%)
Bleeding details
 Bleeding associated with cardiac surgery (including CABG)0 (0.0%)2 (0.5%)
 Bleeding associated with non-cardiac surgery19 (4.4%)27 (6.6%)
 Epistaxis14 (3.2%)14 (3.4%)
 GI: Upper (haematemesis or melena)164 (38.1%)105 (25.7%)
 GI: Lower51 (11.8%)33 (8.1%)
 Gingival1 (0.2%)2 (0.5%)
 Haematoma13 (3.0%)26 (6.4%)
 Haemoptysis5 (1.2%)4 (1.0%)
 Increased or prolonged menstrual or abnormal vaginal bleeding3 (0.7%)1 (0.2%)
 Intra-articular16 (3.7%)21 (5.1%)
 Intracranial55 (12.8%)84 (20.5%)
 Intramuscular (with compartment syndrome)2 (0.5%)1 (0.2%)
 Intramuscular (without compartment syndrome)2 (0.5%)4 (1.0%)
 Intraocular/retinal19 (4.4%)27 (6.6%)
 Macroscopic (gross) haematuria27 (6.3%)21 (5.1%)
 Pericardial0 (0.0%)1 (0.2%)
 Puncture site2 (0.5%)4 (1.0%)
 Rectal28 (6.5%)8 (2.0%)
 Retroperitoneal1 (0.2%)3 (0.7%)
 Skin (ecchymosis other than instrumented site)2 (0.5%)3 (0.7%)
 Subconjunctival or other ocular0 (0.0%)1 (0.2%)
 Other7 (1.6%)19 (4.6%)
  • aDenominator will be based on number of bleeds, which may be more than the number of subjects with major bleeds.

  • bDenominator for just this row will be number of patients with major bleeds. CABG indicates coronary artery bypass graft; GI, gastrointestinal.

Transfusion and hospitalization

Transfusion and hospitalization after major bleeding data are provided in Table 3. The median duration of hospitalization was shorter in the rivaroxaban arm [5 (25th, 75th: 4, 10) days] compared with the warfarin arm [6 (25th, 75th: 4, 11) days]. The median numbers of transfused PRBC per major bleeding episode were similar in the rivaroxaban and warfarin arms [2 (25th, 75th: 2, 4) vs. 2 (25th, 75th: 2, 4) units]. A greater number of major bleeds were transfused with PRBC in the rivaroxaban arm compared with the warfarin arm (n = 176 vs. n = 144 units). Few transfusions of whole blood (n = 14), platelets (n = 10), or cryoprecipitate (n = 2) were used in either arm. However, transfusion of FFP was significantly less common in the rivaroxaban arm (n = 45 vs. n = 81 units). The number of major bleeds receiving 4 or more units of packed red cells or whole blood was comparable in the rivaroxaban and warfarin arms (13.7 vs. 12.5%).

View this table:
Table 3

Hospitalization and transfusion for major bleeding event by randomized treatment

 Rivaroxaban (n = 431)Warfarin (n = 409)
N = subjects, median (25%, 75%)
Duration of hospitalization within 5 days of major bleedN = 101

5 (4–10) days
N = 91

6 (4–11) days
Transfusions within 5 days of major bleed
 Packed red blood cellsN = 176

2 (2–4) units
N = 143

2 (2–4) units
 Whole blood cellsN = 8

2 (1–3) units
N = 6

2 (2–4) units
 PlateletsN = 4

3 (2–5) units
N = 6

5 (3–6) units
 Fresh frozen plasmaN = 45

2 (1–2) units
N = 81

2 (2–4) units
 CryoprecipitateN = 1

1 (1–1) units
N = 1

1 (1–1) units
Number of total units of packed red blood cells or whole blood cells transfused
 ≥2159 (36.9%)129 (31.5%)
 ≥459 (13.7%)51 (12.5%)

Clinical factors associated with FFP transfusion in those major bleeds that required transfusions are shown in Table 4. After accounting for multiple clinically relevant covariates, we found that rivaroxaban was associated with lower odds of FFP transfusion [odds ratio (OR) 0.43 (95% CI 0.29–0.66); P < 0.0001], as was a prior history of stroke or TIA [OR 0.55 (95% CI 0.37–0.84); P = 0.0056]. Prior aspirin and prior VKA use were associated with higher odds of FFP transfusion (Table 4).

View this table:
Table 4

Multivariable model for receipt of fresh frozen plasma for management of major bleedinga

ParameterOdds ratio95% CIP-value
Rivaroxaban (vs. warfarin)0.430.29–0.66<0.0001
Prior aspirin1.721.12–2.640.0132
Prior VKA2.051.25–3.360.0042
Prior stroke/TIA0.550.37–0.840.0056
  • aGEE repeated measures model for fresh frozen plasma or cryoprecipitate transfusion in case of on treatment major bleeding events (observations included n = 840 transfusions, events n = 126). We included the following covariates in the model to begin with: treatment, sex, race, prior aspirin, prior VKA, prior stroke/TIA, history of heart failure, diabetes, hypertension, age, body mass index, creatinine clearance, and days to major bleed. The covariates in the table above remained in the model. CI, confidence interval; TIA, transient ischaemic attack; VKA, vitamin K antagonist.

Unblinding

Throughout the course of the trial, 25 patients (0.017%) required unblinding of their treatment drug assignment. Three patients in the rivaroxaban group were unblinded within a day of the major bleed and five patients in the warfarin group were unblinded within 2 days of the major bleed. There was a single major bleeding event in the rivaroxaban group with an unblinding date preceding the bleed date.

Pharmacologic management

Table 5 details the pharmacologic therapies given following a major bleeding event according to the randomized treatment (rivaroxaban vs. warfarin). Vitamin K administration within 1 day of major bleeding was less frequent in the rivaroxaban arm (7.4 vs. 13.2%). Aspirin discontinuation within 1 day of major bleeding was slightly lower in the rivaroxaban arm vs. the warfarin arm (6.5 vs. 8.5%). Prothrombin complex concentrates were administered less frequently in the rivaroxaban arm [n = 4 (0.9%) vs. n = 9 (2.2%)]. Few patients received recombinant factor VIIa, factor VIII, or factor IX within 1 day of major bleeding in either treatment arm (rivaroxaban n = 1, warfarin n = 5). Results were similar when considering larger treatment windows within 2 and 3–5 days of major bleeding (Table 5).

View this table:
Table 5

Pharmacologic management post-major bleed

CharacteristicRivaroxaban (n = 431)Warfarin (n = 409)
Within 1 day post-bleed
 Vitamin K32 (7.4%)54 (13.2%)
 Protamine0 (0.0%)0 (0.0%)
 Desmopressin0 (0.0%)0 (0.0%)
 Epsilon-aminocaproic acid2 (0.5%)3 (0.7%)
 Tranexamic acid2 (0.5%)11 (2.7%)
 Prothrombin complex concentrates4 (0.9%)9 (2.2%)
 Recombinant factor VIIa0 (0.0%)1 (0.2%)
 Factor VIII concentrate1 (0.2%)1 (0.2%)
 Factor IX concentrate0 (0.0%)3 (0.7%)
Within 2 days post-bleed
 Vitamin K34 (7.9%)60 (14.7%)
 Protamine0 (0.0%)0 (0.0%)
 Desmopressin0 (0.0%)0 (0.0%)
 Epsilon-aminocaproic acid2 (0.5%)3 (0.7%)
 Tranexamic acid3 (0.7%)14 (3.4%)
 Prothrombin complex concentrates4 (0.9%)9 (2.2%)
 Recombinant factor VIIa0 (0.0%)1 (0.2%)
 Factor VIII concentrate1 (0.2%)1 (0.2%)
 Factor IX concentrate0 (0.0%)3 (0.7%)
 Aspirin discontinued31 (7.2%)38 (9.3%)
Within 3–5 days post-bleed
 Vitamin K6 (1.4%)18 (4.4%)
 Protamine0 (0.0%)0 (0.0%)
 Desmopressin0 (0.0%)1 (0.2%)
 Epsilon-aminocaproic acid0 (0.0%)0 (0.0%)
 Tranexamic acid3 (0.7%)8 (2.0%)
 Prothrombin complex concentrates0 (0.0%)0 (0.0%)
 Recombinant factor VIIa0 (0.0%)0 (0.0%)
 Factor VIII concentrate0 (0.0%)0 (0.0%)
 Factor IX concentrate0 (0.0%)0 (0.0%)
 Aspirin discontinued38 (8.8%)41 (10.0%)

Outcomes following a major bleeding event

Outcomes following a major bleeding event, the interaction tests for randomized treatment and major bleeding, and the HR (95% CI) for the outcomes in the adjusted model before and after major bleeding events are shown in Table 6. Event rates were similar in the rivaroxaban and warfarin arms for stroke or non-CNS embolism, myocardial infarction/unstable angina, and the composite of all stroke, non-CNS embolism, myocardial infarction/unstable angina, and all-cause death. Similarly, once a major bleed had occurred, all-cause death after bleeding was similar in the rivaroxaban arm compared with warfarin (20.4 vs. 26.1%; interaction P = 0.1098).

View this table:
Table 6

Outcomes post-major bleed

OutcomeaRivaroxaban (n = 431)Warfarin (n = 409)HR (95% CI)bTreatment × major bleed interaction, P-valuec
Stroke or systemic embolismd20 (4.7%)22 (5.4%)
 Time to stroke or SE, median (range), days64 (16–249)15 (1–71)
 Post-major bleed0.888 (0.420, 1.876)0.5135
 Pre/no major bleed1.102 (0.715, 1698)
Composite of all stroke, non-CNS embolism, MI/UA, and all-cause death104 (24.8%)120 (29.9%)
 Time to composite of all stroke, non-CNS embolism, MI/UA, and all-cause death, median (range), days58 (8–248)11 (2–82)
 Post-major bleed0.758 (0.530, 1.082)0.0975
 Pre/no major bleed0.970 (0.768, 1.225)
All-cause death86 (20.4%)105 (26.1%)
 Time to all-cause death, median (range), days60 (8–246)7 (2–88)
 Post-major bleed0.688 (0.455, 1.042)0.1098
 Pre-/no major bleed0.905 (0.686, 1.194)
MI/UA11 (2.6%)7 (1.7%)
 Time to MI/UA, median (range), days282 (9–485)14 (3–26)
 Post-major bleed1.848 (0.572, 5.971)0.5597
 Pre/no major bleed1.374 (0.707, 2.670)
  • aDenominator is based on number of bleeds which may be more than the number of subjects with major bleeds.

  • bHR and 95% CIs for outcomes of interest are calculated post-major bleed and (pre-major bleed or no major bleed) while restricting the other covariates in the model at their mode value, in case of categorical covariates and median value in case of continuous covariates.

  • cP-values are of treatment by major-bleed interaction in an adjusted proportional hazards model for the outcome where major bleed was included as a time-dependent covariate. Other adjustment covariates included in this model are: sex, history of COPD, history of stroke or TIA, history of PAD, atrial fibrillation type, history of CHF, region, history of alcohol, age, diastolic BP, systolic BP, heart rate, BMI, and creatinine clearance and interactions of each of these covariates with treatment.

  • dAmong patients with stroke or systolic embolism in cohort, 86 had a bleeding event and stroke event on the same day. In 77 out of 86 cases, the stroke was an intracranial haemorrhage and the stroke was assumed to precede the intracranial haemorrhage (bleeding event). In nine patients, the stroke followed the bleeding event and those events are included in this analysis. CI, confidence interval; CNS, central nervous system; HR, hazard ratio; MI, myocardial infarction; SE, systemic embolism; UA, unstable angina.

Discussion

Novel oral anticoagulants are attractive alternatives to warfarin; however, concerns exist regarding treatment of major bleeding events in patients taking these agents. In our analysis of patient management following the 779 major bleeding events overall, there are several important clinically relevant findings. First, transfusion of coagulation factors was rare in both warfarin and rivaroxaban patients and the patients were largely managed using supportive approaches to bleeding. Second, transfusion of FFP was significantly more frequent in warfarin patients. Finally, major adverse outcomes, including death, following a major bleeding event were similar in patients treated with rivaroxaban and warfarin. Therefore, despite concerns about a ‘lack of reversibility’ for factor Xa inhibitors, outcomes following a major bleed are similar to those of VKAs.

Factor Xa inhibitors offer several advantages compared with VKAs, including rapid onset of anticoagulation, shorter half-lives, and predictable pharmacokinetics without the need for routine monitoring. The ROCKET AF trial demonstrated that the factor Xa inhibitor rivaroxaban was non-inferior to warfarin for the prevention of stroke and non-CNS embolism in patients with non-valvular AF with a lower risk of intracranial and fatal bleeding. Despite these findings, many physicians express concern about the ability to manage major haemorrhage in these patients due to the lack of an easily administered, Xa-specific reversal strategy. More precisely, there are concerns that major haemorrhage in patients treated with factor Xa inhibitors could be uncorrectable, leading to fatal haemorrhage. In contrast, warfarin has several potential reversal agents, although their efficacy and time to effective reversal are suboptimal. Full reversal of warfarin with vitamin K can take more than 24 h and has questionable efficacy5 and infusion of FFP can take several hours, requires large volumes, and necessitates repeated dosing for full reversal.6 As a result, life-threatening warfarin-associated bleeding is preferably managed with PCC.7 Early studies with healthy volunteers suggest that PCC may also have the potential to reverse the anticoagulant effects in those treated with factor Xa inhibitors.8 However, these agents must also be used cautiously since they carry risk of prothrombotic effects.

In this context, there are several important and reassuring findings in this analysis of management and outcomes of major bleeding events in patients treated with rivaroxaban or warfarin.9,10 The most important of these findings is that transfusion of coagulation factors was very rare in both warfarin and rivaroxaban patients and the patients were successfully managed using general, supportive approaches to bleeding. In the ROCKET AF trial, investigators were instructed to manage major bleeding with cessation of the anticoagulation study drug and empiric administration of vitamin K and FFP if life-threatening bleeding continued. Sites were advised that coagulation factor administration could be considered in patients with refractory bleeding despite infusion of fresh frozen plasma. In 14 143 patients, 779 experienced a major bleed and only 19 cases were treated with coagulation factor administration, including only five in the rivaroxaban-treated patients. These data complement the findings in the overall ROCKET AF trial, which demonstrated a lower risk of intracranial and fatal bleeding in those patients randomized to rivaroxaban.1 Not only is fatal bleeding less likely, administration of haemostatic products was minimal.

In this study of management for major bleeding, we also found that patients randomized to rivaroxaban received FFP less frequently. Following adjustment for prior aspirin use, VKA use, and previous stroke or TIA, we found that patients treated with rivaroxaban were significantly less likely to receive FFP. Patients with prior exposure to aspirin and VKAs were more likely to receive FFP. Studies in healthy human subjects suggest that four-component PCC can reverse the anticoagulant effect of rivaroxaban.8 These data raise the question of whether the factor Xa inhibition can be overcome with FFP administration. It is important to emphasize that the ROCKET AF trial experience does not provide data that FFP reverses the anticoagulant effects of rivaroxaban in patients with haemorrhage. However, these observations do provide some reassurance that major bleeding in rivaroxaban does not require empiric administration of reversal transfusions more than warfarin-treated patients. Finally, it is important to note that FFP transfusion rates were low overall. Although physicians may express a desire for specific ‘reversal agents,’ these interventions are not used in the vast majority of bleed events.

The third and most important finding in this analysis is that there is no evidence of worse outcomes following major bleeding events in patients treated with rivaroxaban. Specifically, there was no evidence of an interaction between randomized treatment (rivaroxaban or warfarin) and major bleeding for stroke or non-CNS embolism or myocardial infarction/unstable angina, or all-cause death. In fact, the event rates in the rivaroxaban patients for stroke or non-CNS embolism and the composite endpoint were numerically lower in the rivaroxaban groups. Furthermore, the duration of hospitalization following a major bleeding event was shorter in the rivaroxaban-treated patients. Following a major bleeding event, the hazard of death was 0.688 (95% CI 0.455, 1.042) for rivaroxaban compared with warfarin. While the interaction term did not reach statistical significance (P = 0.1098), the lower mortality is very similar to data observed in patients treated with dabigatran (pooled odds ratio 0.68; 95% CI , 0.46–1.01; P = 0.057).11 These data also confirm and extend the safety findings observed in the ROCKET AF trial where fatal bleeding was less common in patients treated with rivaroxaban compared with warfarin and there was no evidence of worse outcomes in patients treated with rivaroxaban once a major bleeding event occurred.

Limitations

As with any post hoc analysis, there are several limitations that must be kept in mind when considering these data. First, bleeding events are post-randomization events and the comparison between patients taking rivaroxaban and warfarin may be subject to residual confounding. Second, these results may not be applicable to all major bleeding events, particularly acute traumatic events such as traumatic limb loss. Finally, patients presenting for emergency medical care in the setting of haemorrhage may have had unblinded INR assessment, which may have disclosed or biased the treating physician to assume knowledge of the randomized therapy.

Conclusions

Among high-risk patients with atrial fibrillation who experienced major bleeding in ROCKET AF, the use of FFP and PCC was less among those allocated rivaroxaban compared with warfarin. However, use of PRBC and outcomes after bleeding were similar among patients randomized to rivaroxaban or to warfarin.

Funding

This work was supported by Johnson & Johnson and Bayer HealthCare.

Conflict of interest: none declared.

References

View Abstract