- Email: [email protected]
Net clinical benefit of rivaroxaban compared with warfarin in atrial fibrillation: Results from ROCKET AF
International Journal of Cardiology 257 (2018) 78–83
Contents lists available at ScienceDirect
International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard
Net clinical benefit of rivaroxaban compared with warfarin in atrial fibrillation: Results from ROCKET AF Adam S. Barnett a,1, Derek D. Cyr a,1, Shaun G. Goodman b,1, Bennett S. Levitan c,1, Zhong Yuan c,1, Graeme J. Hankey d,1, Daniel E. Singer e,1, Richard C. Becker f,1, Günter Breithardt g,1, Scott D. Berkowitz h,1, Jonathan L. Halperin i,1, Werner Hacke j,1, Kenneth W. Mahaffey k,1, Christopher C. Nessel l,1, Keith A.A. Fox m,1, Manesh R. Patel a,1, Jonathan P. Piccini a,⁎,1 a
Duke Clinical Research Institute, Durham, NC, United States Terrence Donnelly Heart Centre, St Michael's Hospital, University of Toronto, Toronto, ON, Canada c Department of Epidemiology, Janssen Research & Development, Titusville, NJ, United States d School of Medicine and Pharmacology, The University of Western Australia, Perth, WA, Australia e Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States f University of Cincinnati Heart, Lung & Vascular Institute, Cincinnati, OH, United States g Department of Cardiology and Angiology, University of Muenster, Muenster, Germany h Bayer HealthCare Pharmaceuticals, Parsippany, NJ, United States i Cardiovascular Institute, Mount Sinai Medical Center, New York, NY, United States j Ruprecht-Karls-University, Heidelberg, Germany k Stanford University School of Medicine, Stanford, CA, United States l Janssen Research and Development LLC, Raritan, NJ, United States m Centre for Cardiovascular Science, University of Edinburgh, United Kingdom b
a r t i c l e
i n f o
Article history: Received 19 May 2017 Accepted 27 June 2017 Keywords: Rivaroxaban Atrial fibrillation Net clinical benefit
a b s t r a c t Aims: The aim of this study was to determine the net clinical benefit (NCB) of rivaroxaban compared with warfarin in patients with atrial fibrillation. Methods: This was a retrospective analysis of 14,236 patients included in ROCKET AF who received at least one dose of study drug. We analyzed NCB using four different methods: (1) composite of death, stroke, systemic embolism, myocardial infarction, and major bleeding; (2) method 1 with fatal or critical organ bleeding substituted for major bleeding; (3) difference between the rate of ischemic stroke or systemic embolism minus 1.5 times the difference between the rate of intracranial hemorrhage; and (4) weighted sum of differences between rates of death, ischemic stroke or systemic embolism, intracranial hemorrhage, and major bleeding. Results: Rivaroxaban was associated with a lower risk of the composite outcome of death, myocardial infarction, stroke, or systemic embolism (rate difference per 10,000 patient-years [RD] = −86.8 [95% CI −143.6 to −30.0]) and fatal or critical organ bleeding (−41.3 [−68 to −14.7]). However, rivaroxaban was associated with a higher risk of major bleeding other than fatal or critical organ bleeding (55.9 [14.7 to 97.2]). Method 1 showed no difference between treatments (−35.5 [− 108.4 to 37.3]). Methods 2–4 favored treatment with rivaroxaban (2: −96.8 [−157.0 to −36.8]; 3: −65.2 [−112.3 to −17.8]; 4: −54.8 [−96.0 to −10.2]). Conclusions: Rivaroxaban was associated with favorable NCB compared with warfarin. The NCB was attributable to lower rates of ischemic events and fatal or critical organ bleeding. © 2017 Elsevier B.V. All rights reserved.
1. Introduction
⁎ Corresponding author at: Electrophysiology Section, Duke Center for Atrial Fibrillation, Duke University Medical Center, Duke Clinical Research Institute, PO Box 17969, Durham, NC 27710, United States. E-mail address: [email protected] (J.P. Piccini). 1 All authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation.
http://dx.doi.org/10.1016/j.ijcard.2017.06.110 0167-5273/© 2017 Elsevier B.V. All rights reserved.
Current professional society guidelines recommend treatment with an oral anticoagulant in patients with atrial fibrillation (AF) who are at increased risk of thromboembolism [1–3]. However, anticoagulation is inherently associated with an increased risk of bleeding, and the benefits of anticoagulation must be weighed against the risks. Thus, for patients with AF, the reduced risk of ischemic stroke and systemic thromboembolism (SSE) must be weighed against the increased risk of bleeding when considering antithrombotic therapy. The concept of
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
net clinical benefit (NCB) has been proposed as a method to estimate the overall relative benefits of different antithrombotic agents, taking into account both thromboembolic and bleeding risk [4,5]. NCB is typically calculated as a composite of thromboembolic and bleeding events, often with numerical weights applied to different outcomes to account for their clinical severity; [4] a therapy that lowers the risk of thromboembolism more than it increases the risk of bleeding would therefore have a favorable NCB. Rivaroxaban is an oral direct factor Xa inhibitor that is noninferior to warfarin for the prevention of stroke in patients with AF [6]. 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) trial, rivaroxaban had a similar risk of overall major bleeding. However, within the subcategories of major bleeding, a differential effect was seen. Rivaroxaban was associated with a lower rate of fatal bleeding and intracranial hemorrhage (ICH) but a higher rate of less impactful major bleeding requiring transfusion. Data regarding the NCB of rivaroxaban compared with warfarin, taking into account both ischemic and bleeding risks, are limited to a single publication that employed only one descriptive method using overall event rates [7]. The goal of this post-hoc analysis was to estimate the NCB of rivaroxaban compared with warfarin using several different methods in a patientlevel analysis. In addition, we sought to identify patient characteristics that were associated with an NCB with one agent compared with the other. 2. Methods The design and main results of the ROCKET AF study have been published previously [6,8]. Briefly, ROCKET AF was a multicenter, international, double-blind, double-dummy, randomized trial comparing fixed-dose rivaroxaban (20 mg once daily, or 15 mg once daily in patients with a creatinine clearance of 30–49 mL/min) with adjusted-dose warfarin (target international normalized ratio 2.0–3.0) for the prevention of stroke (ischemic or hemorrhagic) or systemic embolism. To maintain blinding, point-of-care testing was used to determine real international normalized ratios (in patients taking warfarin) or generate sham values (in patients taking rivaroxaban and receiving placebo warfarin). The doses of warfarin and matching placebo tablets were adjusted based on these values. All appropriate national regulatory authorities and the ethics/institutional review boards at all participating centers approved the study. All patients provided written informed consent. 2.1. Patient population Complete inclusion and exclusion criteria for ROCKET AF have been published [8]. Briefly, patients with electrocardiographically documented nonvalvular AF at moderate to high risk of stroke were recruited at 1178 participating sites in 45 countries. Elevated stroke risk was indicated by a history of stroke, transient ischemic attack, or systemic embolism, or at least two of the following risk factors: heart failure or left ventricular ejection fraction ≤35%, hypertension, age ≥ 75 years, or diabetes mellitus (i.e., CHADS2 score ≥ 2). Those with a high risk for bleeding (including previous intracranial bleeding, surgical trauma within 30 days, and gastrointestinal bleeding within 6 months) were excluded. The present study included all patients who received at least one dose of study drug (on-treatment population N = 14,236). In a sensitivity analysis, we repeated the study with all patients who were randomized (intention-to-treat population, N = 14,264). 2.2. Outcomes The efficacy endpoints included in this analysis were all-cause death, stroke (including both ischemic and hemorrhagic events), myocardial infarction, systemic embolism, and the composite of these four endpoints. The safety endpoints included were fatal or critical organ bleeding, bleeding requiring transfusion of ≥2 units of whole blood or packed red blood cells or causing a drop in hemoglobin ≥2 g/dL, and a composite of these two endpoints (defined as major bleeding). Critical organ bleeding was defined as bleeding in any of the following anatomic locations: intracranial, spinal, ocular, pericardial, articular, retroperitoneal, or intramuscular with compartment syndrome. Outcomes were assessed as time-to-first-event from the first dose of the study drug to 2 days following permanent study drug discontinuation (first dose to last dose + 2 days). Patient outcomes were censored after the first event. 2.3. Calculation of net clinical benefit We chose to use four different methods to estimate NCB given the multitude of approaches previously published and the lack of a widely accepted standardized approach. In Method 1, NCB was defined as the unweighted composite of all-cause death, stroke, myocardial infarction, systemic embolism, or major bleeding, with each event type
79
weighted equally and counted to the composite only once. In Method 2, we modified the composite endpoint in Method 1 by substituting fatal or critical organ bleeding for major bleeding. This was done to exclude less-impactful, non-fatal extracranial hemorrhage events, which typically do not lead to “irreversible harm.” Therefore in Method 2, NCB was defined as the unweighted composite of all-cause death, stroke, myocardial infarction, systemic embolism, fatal bleeding, or critical organ bleeding. This approach of focusing on events that are fatal or cause irreversible harm is similar to the approach the U.S. Food and Drug Administration has described for its approval of other anticoagulants [9,10]. In Method 3, NCB was calculated using an approach based on a previous study by Singer et al., which compared warfarin versus no anticoagulation in patients with AF [4]. In this method, NCB was defined as the difference between the annualized rate of SSE minus 1.5 times the difference between the annualized rate of ICH. Thus, NCB was defined as: (SSErivaroxaban − SSEwarfarin) + 1.5 × (ICHrivaroxaban − ICHwarfarin). The weighting factor of 1.5 reflects the greater relative impact in terms of death and disability of an ICH relative to an ischemic stroke. A numerically negative result indicates that treatment with rivaroxaban is favored compared with warfarin. In Method 4, NCB was defined as the sum of the rate differences of SSE, ICH, major bleeding (MB) excluding ICH, and all-cause death (DE) in the two treatment groups, with each rate difference multiple by a weighting factor. This formula was based on a previous study of left atrial appendage closure by Gangireddy et al. [11] The weight assigned to each rate difference reflects the relative impact of that event type in terms of death and disability. Death is assigned the highest weight (1.00), and SSE is given a reference weight of 0.20. The weights of ICH and MB are based on an analysis of the Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events (ACTIVE) trials [12]. In this analysis, the hazard ratio (HR) for death after ICH was 3.08 compared with SSE, and the HR for DE after MB was 0.67 compared with SSE. Therefore, ICH is assigned a weight of 0.6 and MB a weight of 0.1. Therefore, NCB was calculated as: 1 × (DErivaroxaban − DEwarfarin) + 0.2 × (SSErivaroxaban − SSEwarfarin) + 0.6 × (ICHrivaroxaban − ICHwarfarin) + 0.1 × (MBrivaroxaban − MBwarfarin). We also conducted additional sensitivity analyses using weights of 0.1 and 0.3 for SSE, 0.3 and 0.9 for ICH, and 0.05 and 0.15 for MB. For methods 1, 2, and 4, ICH was not counted within MB to avoid “double counting” of this endpoint in both stroke and major bleeding.
2.4. Statistical analysis Baseline characteristics were summarized as counts (percentages) for categorical variables and as median values with 25th and 75th percentiles for continuous variables. Event rates were calculated as the number of events per 100 patient-years. The rate difference for each endpoint was calculated by subtracting the event rate per 100 patient-years of the warfarin group from the event rate of the rivaroxaban group (rivaroxaban minus warfarin). Thus, a negative rate difference indicates a lower event rate in the rivaroxaban group. The rate difference was multiplied by a factor of 100 and expressed as the number of additional events per 10,000 patient-years, to enable easier interpretation of the small rate differences as natural frequencies. The 95% confidence interval (CI) of the rate difference was calculated using empirical bootstrap resampling with 1000 replicates. Kaplan-Meier (K-M) cumulative event rates throughout the duration of the ROCKET AF follow-up period were calculated according to randomized treatment. The cumulative K-M rates were converted to cumulative rate differences by subtracting the K-M estimates for each treatment. The excess number of events (per 10,000 patients) between treatments over time was visualized by plotting the cumulative rate differences.
3. Results From December 18, 2006, through June 17, 2009, a total of 14,264 patients underwent randomization (intention-to-treat population) and a total of 14,236 patients received at least one dose of a study drug (on-treatment population). The median duration of treatment exposure was 590 days, and the median follow-up period was 707 days. The proportion of patients who stopped their assigned therapy before an endpoint event and before the termination date was 23.7% in the rivaroxaban group and 22.2% in the warfarin group. A total of 32 patients were lost to follow-up. Because of violations in Good Clinical Practice guidelines at one site that made the data unreliable, 93 patients (50 in the rivaroxaban group and 43 in the warfarin group) were excluded from all efficacy analyses. The key clinical characteristics of the patients included in the analysis (on-treatment population) are shown in Table 1. The median age was 73 years, and 39.7% of patients were female. Comorbid conditions were prevalent: 90.5% of patients had hypertension, 62.5% had heart failure, 39.9% had diabetes mellitus, and 54.7% of patients had a previous SSE or transient ischemic attack. The mean and median CHADS2 scores were 3.5 and 3.0, respectively, indicating a high-risk population. Previous use of vitamin K antagonists was reported by 62.4% of patients.
80
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
Table 1 Patient characteristics at baseline (on-treatment population, N = 14,236). Characteristic
Rivaroxaban (N = 7111)
Warfarin (N = 7125)
Age, year, median (25th, 75th percentiles) Female sex, n (%) Race, n (%) White Black Asian Other Region, n (%) Asia/Pacific Islands Eastern Europe Latin America North America Western Europe Type of atrial fibrillation, n (%) Persistent Paroxysmal Newly diagnosed CHADS2 score, mean ± standard deviation CHADS2 score, n (%) 1 2 3 4 5 6 Presenting characteristics, median (25th, 75th percentiles) Body mass index, kg/m2 Systolic blood pressure, mm Hg Diastolic blood pressure, mm Hg Heart rate, bpm Creatinine clearance, mL/mina Baseline comorbidities, n (%) Prior stroke / transient ischemic attack / systemic embolism Carotid occlusive disease Hypertension Diabetes Prior myocardial infarction Chronic heart failure Chronic obstructive pulmonary disease Prior medications Vitamin K antagonist Aspirin
73.0 (65.0,78.0) 2819 (39.6)
73.0 (65.0,78.0) 2826 (39.7)
5906 (83.1) 94 (1.3) 894 (12.6) 217 (3.1)
5952 (83.5) 85 (1.2) 887 (12.4) 201 (2.8)
1052 (14.8) 2746 (38.6) 939 (13.2) 1334 (18.8) 1040 (14.6)
1052 (14.8) 2747 (38.6) 938 (13.2) 1339 (18.8) 1049 (14.7)
5771 (81.2) 1242 (17.5) 98 (1.4) 3.5 ± 0.9
5754 (80.8) 1269 (17.8) 102 (1.4) 3.5 ± 0.9
1 (0.0) 923 (13.0) 3047 (42.8) 2087 (29.3) 930 (13.1) 123 (1.7)
2 (0.0) 932 (13.1) 3156 (44.3) 1998 (28.0) 879 (12.3) 158 (2.2)
28.3 (25.2,32.1) 130.0 (120.0,140.0) 80.0 (70.0,85.0) 76.0 (67.0,86.0) 67.0 (52.0,88.0)
28.1 (25.1,31.8) 130.0 (120.0,140.0) 80.0 (70.0,85.0) 76.0 (67.0,86.0) 67.0 (52.0,86.0)
3905 (54.9) 291 (4.1) 6419 (90.3) 2869 (40.3) 1178 (16.6) 4457 (62.7) 751 (10.6)
3889 (54.6) 300/ (4.2) 6468 (90.8) 2814 (39.5) 1282 (18.0) 4437 (62.3) 742 (10.4)
4431 (62.3) 2578 (36.3%)
4458 (62.6) 2616 (36.7%)
a
Creatinine clearance was calculated using the Cockcroft–Gault formula.
3.1. Outcomes Event rates and rate differences for the efficacy and safety endpoints examined in this study are shown in Table 2. The composite efficacy endpoint of all-cause death, stroke (ischemic or hemorrhagic), myocardial infarction, or systemic embolism occurred in 471 patients in the rivaroxaban group (4.25 per 100 patient-years) and in 575 patients in the warfarin group (5.12 per 100 patient-years). The rate difference for this endpoint was − 86.8 per 10,000 patient-years (95% CI -143.6 to −30.0) favoring treatment with rivaroxaban. Thus, treatment with rivaroxaban would be expected to prevent approximately 86.8 adverse events per 10,000 patient-years compared with treatment with warfarin. The rivaroxaban group had a lower cumulative event rate over the duration of the study (Supplemental Fig. 1A, log rank P = 0.003). This finding is also shown in Supplemental Fig. 2A, which plots the cumulative difference between the number of events in the rivaroxaban group and the warfarin group over time. There was a difference in the event rates of systemic embolism favoring treatment with rivaroxaban (Table 2, rate difference − 15.0; 95% CI −24.0 to −5.9). There were numerically, but not statistically significant, lower rates of all-cause death, stroke, and myocardial infarction in patients treated with rivaroxaban.
The composite safety endpoint of major bleeding occurred in 395 patients in the rivaroxaban group (3.60 per 100 patient-years) and 386 patients in the warfarin group (3.45 per 100 patient-years). There was no statistically significant difference in the rates of major bleeding (rate difference 14.2 per 10,000 patient-years; 95% CI − 35.2 to 63.7). The cumulative event rates were similar in the rivaroxaban and warfarin arms (Supplemental Fig. 1B, log rank P = 0.576), and a plot of the cumulative rate difference (Supplemental Fig. 2A) does not clearly favor warfarin or rivaroxaban. Considering both the more and less severe forms of major bleeding, the rate difference for the endpoint of fatal or critical organ bleeding was − 41.3 per 10,000 patient-years (95% CI -68.0 to − 14.7) and favored treatment with rivaroxaban. The cumulative rate of fatal or critical organ bleeding was also lower in the rivaroxaban group (Supplemental Fig. 1C, log rank P = 0.003) along with the cumulative rate difference (Supplemental Fig. 2B). The rate difference for major bleeding other than fatal or critical organ bleeding was 55.9 per 10,000 patient-years (95% CI 14.7 to 97.2) and favored treatment with warfarin. The cumulative rate of this category of major bleeding was also lower in the warfarin group (Supplemental Fig. 1D, log rank P = 0.008). This can also be seen in the plot of the cumulative rate difference (Supplemental Fig. 2C).
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
81
Table 2 Summary of efficacy and safety endpoints (on-treatment population, N = 14,236). Rivaroxaban (N = 7111) Event Efficacy Death, stroke, myocardial infarction, or systemic embolism All-cause death Strokec Ischemic stroke Myocardial infarction Systemic embolism Ischemic stroke or systemic embolism Safety Major bleeding Fatald or critical organ bleeding Fatald bleeding Critical organ bleeding Intracranial hemorrhage Non-fatal, non-critical organ bleeding a b c d
Rate difference (95% confidence interval)b
Warfarin (N = 7125)
No. events
Event ratea
No. events
Event ratea
471 208 184 149 101 5 160
4.25 1.87 1.65 1.34 0.91 0.05 1.44
575 250 221 161 126 22 194
5.12 2.21 1.96 1.43 1.12 0.19 1.72
−86.8 (−143.6, −30.0) −34.2 (−71.6, 3.1) −30.3 (−65.5, 4.8) −8.7 (−39.4, 22.1) −20.9 (−47.3, 5.5) −15.0 (−24.0, −5.9) −28.0 (−60.8, 4.9)
395 93 21 91 55 302
3.60 0.83 0.19 0.82 0.49 2.74
386 141 43 133 84 245
3.45 1.25 0.38 1.18 0.74 2.18
14.2 (−35.2, 63.7) −41.3 (−68.0, −14.7) −19.1 (−33.0, −5.2) −36.0 (−62.1, −10.0) −24.8 (−45.3, −4.3) 55.9 (14.7, 97.2)
Per 100 patient-years. Per 10,000 patient-years. Negative values favor rivaroxaban. Ischemic or hemorrhagic stroke. Narrow definition of fatal bleeding.
3.2. Net clinical benefit of rivaroxaban In Method 1, there was a trend towards NCB with rivaroxaban, although this did not reach significance (Table 3, rate difference − 35.5 per 10,000 patient-years; 95% CI − 108.4 to 37.3). In Method 2, rivaroxaban had an NCB of − 96.8 per 10,000 patient-years (95% CI −157.0 to −36.8); thus, treatment with rivaroxaban would be expected to prevent approximately 97 adverse events per 10,000 patient-years compared with warfarin. This finding was significant across almost all patient subgroups, with the exception of race marked as “other” (Supplemental Fig. 3). Of note, there was a large rate difference favoring warfarin in black patients that did not reach statistical significance (rate difference 617.4 per 10,000 patient-years; 95% CI − 85.3 to 1321.0). The event rates in this subgroup were very low (15 and 6 for rivaroxaban and warfarin, respectively). In Method 3, using the NCB method previously described by Singer et al. [4], rivaroxaban was associated with − 65.2 (95% CI − 112.3 to − 17.8; Table 3) fewer stroke equivalents per 10,000 patient-years compared with warfarin. In Method 4, using the NCB method previously described by Gangireddy et al. [11], rivaroxaban was associated with an NCB of − 54.8 (95% CI − 96.0 to − 10.2) fewer death equivalents per 10,000 patient-years (middle row, Table 4). The use of smaller weights for event types other than death reduced the NCB to − 44.5, but this was
still significant (95% CI − 82.0 to −4.8). The use of higher weights increased the NCB to −65.1 (95% CI −112.2 to −15.8). 3.3. Intention-to-treat population Similar results were achieved using the intention-to-treat population (Supplemental Tables 1–4). Rivaroxaban again had a lower rate of death, stroke, myocardial infarction, or systemic embolism, although the absolute value of the rate difference per 10,000 patient-years (RD) was numerically smaller compared with the on-treatment population (RD = −64.3, 95% CI −128.4 to −0.3). This was true as well for fatal or critical organ bleeding (RD = − 47.5, 95% CI − 75.0 to − 20.1). Rivaroxaban was favored in NCB methods 2–4, similar to the ontreatment population, although the absolute value of the rate differences were numerically smaller, and the 95% CI for method 3 barely crossed zero (Method 2: RD = −80.4, 95% CI −147.3 to −13.4; Method 3: RD = −47.2, 95% CI −100.2 to 2.0; Method 4: RD = −64.0, 95% CI −122.3 to −6.4). No difference was observed for Method 1 (RD = −14.1, 95% CI −91.6 to 63.4). 4. Discussion In this retrospective analysis, we estimated the NCB of rivaroxaban compared with warfarin using four different methods that included
Table 3 Net clinical benefit Methods 1–3 (on-treatment population, N = 14,236). Rivaroxaban (N = 7111)
Warfarin (N = 7125)
No. events
Event ratea
No. events
Event ratea
Method 1 Death + stroke + myocardial infarction + systemic embolism + major bleeding
811
7.42
863
7.78
−35.5 (−108.4, 37.3)
Method 2 Death + stroke + myocardial infarction + systemic embolism + fatal bleedingc + critical organ bleed
528
4.76
643
5.73
−96.8 (−157.0, −36.8)
Event
Method 3d based on Singer et al. [4] Net clinical benefit a
Net clinical benefit (95% confidence interval)b
−65.2 (−112.3, −17.8)
Per 100 patient-years. b Per 10,000 patient-years. Negative values favor rivaroxaban. c Narrow definition of fatal bleeding. d Net clinical benefit = (SSErivaroxaban − SSEwarfarin) + 1.5 × (ICHrivaroxaban − ICHwarfarin), where SSE = ischemic stroke or systemic embolism and ICH = intracranial hemorrhage.
82
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
Table 4 Net clinical benefit Method 4 based on Gangireddy et al. [11] (on-treatment population, N = 14,236). Net clinical benefit (95% confidence interval)a
Coefficients DE (A) SSE (B) ICH (C) MB (D) 1 1 1
0.1 0.2 0.3
0.3 0.6 0.9
0.05 0.10 0.15
−44.5 (−82.0, −4.8) −54.8 (−96.0, −10.2) −65.1 (−112.2, −15.8)
Net clinical benefit = A × (DErivaroxaban − DEwarfarin) + B×(SSErivaroxaban − SSEwarfarin) + C × (ICHrivaroxaban − ICHwarfarin) + D × (MBrivaroxaban − MBwarfarin), where DE = death, SSE = ischemic stroke or systemic embolism, ICH = intracranial hemorrhage, and MB = major bleeding excluding ICH. a Per 10,000 patient-years. Negative values favor rivaroxaban.
both ischemic and bleeding events. Rivaroxaban was associated with a favorable NCB in three out of the four methods. The overall NCB of rivaroxaban was attributable to both a reduced risk of ischemic events and a lower risk of fatal or critical organ bleeding compared with warfarin. However, rivaroxaban was associated with a higher risk of less clinically significant bleeding. Accordingly, an NCB was not observed in the one method that counted extracranial bleeding as equivalent to other hemorrhagic and ischemic event types (Method 1). Overall, these results support the use of rivaroxaban, which is associated with a decreased risk of ischemic and fatal or critical organ bleeding events, over warfarin in patients with AF. Antithrombotic agents are highly effective in preventing thromboembolism in patients with AF. However, this benefit can be partially or completely offset by the increased risk of bleeding. Therefore, to assess the overall benefit of oral anticoagulants, one must consider both the reduced risk of thromboembolism and the increased risk of bleeding. In this analysis, we used several different methods to estimate the NCB of rivaroxaban compared with warfarin in patients with AF at increased risk of thromboembolism, taking into account the clinical significance of the events involved. First, we compared the annualized rates of the composite of death, stroke, myocardial infarction, systemic embolism, and major bleeding in the two treatment groups (Method 1). Using this method, treatment with rivaroxaban was associated with a lower composite event rate, although this finding did not reach statistical significance. This method of calculating NCB weights major extracranial hemorrhages as equivalent to intracranial events, although the latter are far more likely to be fatal or result in long-term disability. Rivaroxaban had a higher rate of extracranial hemorrhages but a lower rate of intracranial events. We then conducted an analysis of NCB focused on events that can cause irreversible harm by substituting fatal or critical organ bleeding for major bleeding in the composite (Method 2). Extracranial bleeding events that did not lead to death, stroke, myocardial infarction, or embolism were excluded. This analysis favored rivaroxaban, with an estimated 96.8 events prevented per 10,000 patient-years compared with warfarin. This result is attributable to a reduced risk of both ischemic events and fatal or critical organ bleeding events in the rivaroxaban group (Table 2). Importantly, this finding was consistent across almost all patient subgroups, with the exception of race categorized as “other”, a relatively small group (N = 418, Supplemental Fig. 3). The strongest NCB with rivaroxaban was observed in patients with newly diagnosed AF (RD = −876.1 per 10,000 patient-years, 95% CI −1582.0 to −170.0; Supplemental Fig. 3), although this group of patients was relatively small (N = 202). Another way to estimate the NCB of rivaroxaban is to compare the risk of events that cause irreversible harm to the risk of events that are generally reversible. With rivaroxaban, 97 fewer patients per 10,000 patient-years would experience an event that is generally fatal or would cause irreversible harm (i.e., death, stroke, myocardial infarction, systemic embolism, or fatal or critical organ bleeding). In contrast,
56 more patients treated with rivaroxaban per 10,000 patient-years would experience transient transfusions and reductions in hemoglobin (i.e., major bleeding other than fatal or critical organ bleeding). This gives a ratio of 97/56 generally fatal or irreversible harm events prevented versus transient bleeding events caused, favoring rivaroxaban. While simply comparing event rates of ischemic and bleeding composite outcomes is informative, it does not account for the greater relative clinical impact of certain outcomes. For example, ICH is associated with a higher mortality rate relative to ischemic stroke [13,14]. Therefore, this outcome is sometimes multiplied by a weight N1 to reflect the greater clinical significance of these events [4]. In that case, a therapy must prevent more than one thromboembolic event to offset a single ICH in order to achieve a favorable NCB. We conducted two additional analyses using methods that apply weights to different outcomes to account for different relative clinical impacts. The first method (Method 3), previously published by Singer et al. in an analysis of the NCB of warfarin, applies a weight of 1.5 to the rate difference for ICH [4]. This method favored rivaroxaban, with an estimated 65.2 SSE equivalents prevented per 10,000 patient-years compared with warfarin. In the second analysis (Method 4), we calculated NCB as a sum of the rate differences of death, SSE, ICH, and major bleeding, each multiplied by a different weight [11]. Three different sets of weights were used, and all favored rivaroxaban. Using the baseline set of weights with SSE assigned to 0.2, rivaroxaban use resulted in an estimated 54.8 fewer death equivalents per 10,000 patient-years.
4.1. Limitations This analysis has several limitations. While we believe that these findings apply to most patients with AF, it is unclear whether they can be extrapolated to AF patients that are dissimilar from patients enrolled in ROCKET AF, such as patients at low risk of stroke. We used the ontreatment population for all analyses, which differed slightly from the intention-to-treat population and therefore may have been subject to residual confounding. However, we repeated all analyses using the intention-to-treat population and achieved similar results. We did not stratify warfarin patients by time in therapeutic range. It is possible that the NCB of rivaroxaban may be different if compared with warfarin populations with a higher or lower time in therapeutic range than the overall ROCKET AF population. However, in the ROCKET AF trial, the effect of rivaroxaban did not differ across quartiles of the duration of time that international normalized ratio values were within the therapeutic range.
5. Conclusions In conclusion, taking into account both ischemic and bleeding risk, rivaroxaban was associated with a favorable NCB in three out of four different methods used in this study. The NCB of rivaroxaban was attributable to lower rates of ischemic events and fatal or critical organ bleeding. There was an increased risk of lower severity bleeding with rivaroxaban but a decreased risk of fatal or critical organ bleeding. These results support the use of rivaroxaban over warfarin in patients with AF who are clinically similar to the cohort enrolled in ROCKET AF. In this patient population, treatment with rivaroxaban confers a lower risk of ischemic events and fatal or critical organ bleeding but may increase the risk of non-fatal or non-critical organ bleeding.
Funding ROCKET AF was supported by Johnson & Johnson Pharmaceutical Research and Development and Bayer HealthCare.
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
Disclosures JPP receives research funding from AHRQ, ARCA Biopharma, Boston Scientific, Gilead, Janssen Pharmaceuticals, ResMed, Spectranetics, and St Jude Medical and provides consulting to BMS/Pfizer, GSK, Johnson & Johnson, Medtronic, and Spectranetics. ZY and BL are salaried employees of Janssen Research & Development. SDB is employed as a clinical research physician at Bayer HealthCare Pharmaceuticals, Inc., Parsippany, NJ. SGG has received research grant support and speaker/ consulting honoraria from Bayer, Johnson & Johnson, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi-Sankyo, and Pfizer. DES provides consulting to Boehringer Ingelheim, Bristol-Myers Squibb, Merck, Johnson and Johnson and Medtronic and has received research grants from Boehringer Ingelheim, Bristol-Myers Squibb, and Medtronic. The remaining authors have nothing to disclose.
[3]
[4]
[5]
[6]
[7]
Conflict of interest The authors report no relationships that could be construed as a conflict of interest Appendix A. Supplementary data
[8]
[9] [10]
Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ijcard.2017.06.110.
[11]
References
[12]
[1] C.T. January, L.S. Wann, J.S. Alpert, H. Calkins, J.E. Cigarroa, J.C. Cleveland, J.B. Conti, P.T. Ellinor, M.D. Ezekowitz, M.E. Field, K.T. Murray, R.L. Sacco, W.G. Stevenson, P.J. Tchou, C.M. Tracy, C.W. Yancy, ACC/AHA Task Force Members, 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: executive summary: a report of the American College of Cardiology/American Heart Association Task Force on practice guidelines and the Heart Rhythm Society, Circulation 130 (2014) 2071–2104. [2] P. Kirchhof, S. Benussi, D. Kotecha, A. Ahlsson, D. Atar, B. Casadei, M. Castella, H.-C. Diener, H. Heidbuchel, J. Hendriks, G. Hindricks, A.S. Manolis, J. Oldgren, B.A. Popescu, U. Schotten, B. Van Putte, P. Vardas, Authors/Task Force Members, Document reviewers: 2016 ESC guidelines for the management of atrial fibrillation
[13]
[14]
83
developed in collaboration with EACTS: the task force for the management of atrial fibrillation of the European Society of Cardiology (ESC) developed with the special contribution of the Europea, Eur. Heart J. 37 (38) (2016) 2893–2962. L. Macle, J. Cairns, K. Leblanc, T. Tsang, A. Skanes, J.L. Cox, J.S. Healey, A. Bell, L. Pilote, J.G. Andrade, L.B. Mitchell, C. Atzema, D. Gladstone, M. Sharma, S. Verma, S. Connolly, P. Dorian, R. Parkash, M. Talajic, S. Nattel, A. Verma, CCS Atrial Fibrillation Guidelines Committee, 2016 focused update of the Canadian Cardiovascular Society guidelines for the management of atrial fibrillation, Can. J. Cardiol. 32 (2016) 1170–1185. D.E. Singer, Y. Chang, M.C. Fang, L.H. Borowsky, N.K. Pomernacki, N. Udaltsova, A.S. Go, The net clinical benefit of warfarin anticoagulation in atrial fibrillation, Ann. Intern. Med. 151 (2009) 297–305. P.S. Wells, M.H. Prins, B. Levitan, J. Beyer-Westendorf, T.A. Brighton, H. Bounameaux, A.T. Cohen, B.L. Davidson, P. Prandoni, G.E. Raskob, Z. Yuan, E.G. Katz, M. Gebel, A.W. Lensing, Long-term anticoagulation with rivaroxaban for preventing recurrent VTE: a benefit-risk analysis of EINSTEIN Extension, Chest 150 (5) (2016) 1059–1068. M.R. Patel, K.W. Mahaffey, J. Garg, G. Pan, D.E. Singer, W. Hacke, G. Breithardt, J.L. Halperin, G.J. Hankey, J.P. Piccini, R.C. Becker, C.C. Nessel, J.F. Paolini, S.D. Berkowitz, Fox KAA, R.M. Califf, ROCKET AF Investigators, Rivaroxaban versus warfarin in nonvalvular atrial fibrillation, N. Engl. J. Med. 365 (2011) 883–891. G. Renda, M. Di Nicola, R. De Caterina, Net clinical benefit of non-vitamin K antagonist oral anticoagulants versus warfarin in phase III atrial fibrillation trials, Am. J. Med. 128 (2015) 1007–1014.e2. ROCKET AF Study Investigators, Rivaroxaban-once daily, oral, direct factor Xa inhibition compared with vitamin K antagonism for prevention of stroke and embolism trial in atrial fibrillation: rationale and design of the ROCKET AF study, Am. Heart J. 159 (2010) 340–347.e1. B.N. Beasley, E.F. Unger, R. Temple, Anticoagulant options—why the FDA approved a higher but not a lower dose of dabigatran, N. Engl. J. Med. 364 (2011) 1788–1790. E.F. Unger, Weighing benefits and risks—the FDA's review of prasugrel, N. Engl. J. Med. 361 (2009) 942–945. S.R. Gangireddy, J.L. Halperin, V. Fuster, V.Y. Reddy, Percutaneous left atrial appendage closure for stroke prevention in patients with atrial fibrillation: an assessment of net clinical benefit, Eur. Heart J. 33 (2012) 2700–2708. S.J. Connolly, J.W. Eikelboom, J. Ng, J. Hirsh, S. Yusuf, J. Pogue, R. de Caterina, S. Hohnloser, R.G. Hart, ACTIVE (Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events) Steering Committee and Investigators, Net clinical benefit of adding clopidogrel to aspirin therapy in patients with atrial fibrillation for whom vitamin K antagonists are unsuitable, Ann. Intern. Med. 155 (2011) 579–586. E.M. Hylek, A.S. Go, Y. Chang, N.G. Jensvold, L.E. Henault, J.V. Selby, D.E. Singer, Effect of intensity of oral anticoagulation on stroke severity and mortality in atrial fibrillation, N. Engl. J. Med. 349 (2003) 1019–1026. M.C. Fang, A.S. Go, Y. Chang, E.M. Hylek, L.E. Henault, N.G. Jensvold, D.E. Singer, Death and disability from warfarin-associated intracranial and extracranial hemorrhages, Am. J. Med. 120 (2007) 700–705.
Contents lists available at ScienceDirect
International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard
Net clinical benefit of rivaroxaban compared with warfarin in atrial fibrillation: Results from ROCKET AF Adam S. Barnett a,1, Derek D. Cyr a,1, Shaun G. Goodman b,1, Bennett S. Levitan c,1, Zhong Yuan c,1, Graeme J. Hankey d,1, Daniel E. Singer e,1, Richard C. Becker f,1, Günter Breithardt g,1, Scott D. Berkowitz h,1, Jonathan L. Halperin i,1, Werner Hacke j,1, Kenneth W. Mahaffey k,1, Christopher C. Nessel l,1, Keith A.A. Fox m,1, Manesh R. Patel a,1, Jonathan P. Piccini a,⁎,1 a
Duke Clinical Research Institute, Durham, NC, United States Terrence Donnelly Heart Centre, St Michael's Hospital, University of Toronto, Toronto, ON, Canada c Department of Epidemiology, Janssen Research & Development, Titusville, NJ, United States d School of Medicine and Pharmacology, The University of Western Australia, Perth, WA, Australia e Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States f University of Cincinnati Heart, Lung & Vascular Institute, Cincinnati, OH, United States g Department of Cardiology and Angiology, University of Muenster, Muenster, Germany h Bayer HealthCare Pharmaceuticals, Parsippany, NJ, United States i Cardiovascular Institute, Mount Sinai Medical Center, New York, NY, United States j Ruprecht-Karls-University, Heidelberg, Germany k Stanford University School of Medicine, Stanford, CA, United States l Janssen Research and Development LLC, Raritan, NJ, United States m Centre for Cardiovascular Science, University of Edinburgh, United Kingdom b
a r t i c l e
i n f o
Article history: Received 19 May 2017 Accepted 27 June 2017 Keywords: Rivaroxaban Atrial fibrillation Net clinical benefit
a b s t r a c t Aims: The aim of this study was to determine the net clinical benefit (NCB) of rivaroxaban compared with warfarin in patients with atrial fibrillation. Methods: This was a retrospective analysis of 14,236 patients included in ROCKET AF who received at least one dose of study drug. We analyzed NCB using four different methods: (1) composite of death, stroke, systemic embolism, myocardial infarction, and major bleeding; (2) method 1 with fatal or critical organ bleeding substituted for major bleeding; (3) difference between the rate of ischemic stroke or systemic embolism minus 1.5 times the difference between the rate of intracranial hemorrhage; and (4) weighted sum of differences between rates of death, ischemic stroke or systemic embolism, intracranial hemorrhage, and major bleeding. Results: Rivaroxaban was associated with a lower risk of the composite outcome of death, myocardial infarction, stroke, or systemic embolism (rate difference per 10,000 patient-years [RD] = −86.8 [95% CI −143.6 to −30.0]) and fatal or critical organ bleeding (−41.3 [−68 to −14.7]). However, rivaroxaban was associated with a higher risk of major bleeding other than fatal or critical organ bleeding (55.9 [14.7 to 97.2]). Method 1 showed no difference between treatments (−35.5 [− 108.4 to 37.3]). Methods 2–4 favored treatment with rivaroxaban (2: −96.8 [−157.0 to −36.8]; 3: −65.2 [−112.3 to −17.8]; 4: −54.8 [−96.0 to −10.2]). Conclusions: Rivaroxaban was associated with favorable NCB compared with warfarin. The NCB was attributable to lower rates of ischemic events and fatal or critical organ bleeding. © 2017 Elsevier B.V. All rights reserved.
1. Introduction
⁎ Corresponding author at: Electrophysiology Section, Duke Center for Atrial Fibrillation, Duke University Medical Center, Duke Clinical Research Institute, PO Box 17969, Durham, NC 27710, United States. E-mail address: [email protected] (J.P. Piccini). 1 All authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation.
http://dx.doi.org/10.1016/j.ijcard.2017.06.110 0167-5273/© 2017 Elsevier B.V. All rights reserved.
Current professional society guidelines recommend treatment with an oral anticoagulant in patients with atrial fibrillation (AF) who are at increased risk of thromboembolism [1–3]. However, anticoagulation is inherently associated with an increased risk of bleeding, and the benefits of anticoagulation must be weighed against the risks. Thus, for patients with AF, the reduced risk of ischemic stroke and systemic thromboembolism (SSE) must be weighed against the increased risk of bleeding when considering antithrombotic therapy. The concept of
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
net clinical benefit (NCB) has been proposed as a method to estimate the overall relative benefits of different antithrombotic agents, taking into account both thromboembolic and bleeding risk [4,5]. NCB is typically calculated as a composite of thromboembolic and bleeding events, often with numerical weights applied to different outcomes to account for their clinical severity; [4] a therapy that lowers the risk of thromboembolism more than it increases the risk of bleeding would therefore have a favorable NCB. Rivaroxaban is an oral direct factor Xa inhibitor that is noninferior to warfarin for the prevention of stroke in patients with AF [6]. 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) trial, rivaroxaban had a similar risk of overall major bleeding. However, within the subcategories of major bleeding, a differential effect was seen. Rivaroxaban was associated with a lower rate of fatal bleeding and intracranial hemorrhage (ICH) but a higher rate of less impactful major bleeding requiring transfusion. Data regarding the NCB of rivaroxaban compared with warfarin, taking into account both ischemic and bleeding risks, are limited to a single publication that employed only one descriptive method using overall event rates [7]. The goal of this post-hoc analysis was to estimate the NCB of rivaroxaban compared with warfarin using several different methods in a patientlevel analysis. In addition, we sought to identify patient characteristics that were associated with an NCB with one agent compared with the other. 2. Methods The design and main results of the ROCKET AF study have been published previously [6,8]. Briefly, ROCKET AF was a multicenter, international, double-blind, double-dummy, randomized trial comparing fixed-dose rivaroxaban (20 mg once daily, or 15 mg once daily in patients with a creatinine clearance of 30–49 mL/min) with adjusted-dose warfarin (target international normalized ratio 2.0–3.0) for the prevention of stroke (ischemic or hemorrhagic) or systemic embolism. To maintain blinding, point-of-care testing was used to determine real international normalized ratios (in patients taking warfarin) or generate sham values (in patients taking rivaroxaban and receiving placebo warfarin). The doses of warfarin and matching placebo tablets were adjusted based on these values. All appropriate national regulatory authorities and the ethics/institutional review boards at all participating centers approved the study. All patients provided written informed consent. 2.1. Patient population Complete inclusion and exclusion criteria for ROCKET AF have been published [8]. Briefly, patients with electrocardiographically documented nonvalvular AF at moderate to high risk of stroke were recruited at 1178 participating sites in 45 countries. Elevated stroke risk was indicated by a history of stroke, transient ischemic attack, or systemic embolism, or at least two of the following risk factors: heart failure or left ventricular ejection fraction ≤35%, hypertension, age ≥ 75 years, or diabetes mellitus (i.e., CHADS2 score ≥ 2). Those with a high risk for bleeding (including previous intracranial bleeding, surgical trauma within 30 days, and gastrointestinal bleeding within 6 months) were excluded. The present study included all patients who received at least one dose of study drug (on-treatment population N = 14,236). In a sensitivity analysis, we repeated the study with all patients who were randomized (intention-to-treat population, N = 14,264). 2.2. Outcomes The efficacy endpoints included in this analysis were all-cause death, stroke (including both ischemic and hemorrhagic events), myocardial infarction, systemic embolism, and the composite of these four endpoints. The safety endpoints included were fatal or critical organ bleeding, bleeding requiring transfusion of ≥2 units of whole blood or packed red blood cells or causing a drop in hemoglobin ≥2 g/dL, and a composite of these two endpoints (defined as major bleeding). Critical organ bleeding was defined as bleeding in any of the following anatomic locations: intracranial, spinal, ocular, pericardial, articular, retroperitoneal, or intramuscular with compartment syndrome. Outcomes were assessed as time-to-first-event from the first dose of the study drug to 2 days following permanent study drug discontinuation (first dose to last dose + 2 days). Patient outcomes were censored after the first event. 2.3. Calculation of net clinical benefit We chose to use four different methods to estimate NCB given the multitude of approaches previously published and the lack of a widely accepted standardized approach. In Method 1, NCB was defined as the unweighted composite of all-cause death, stroke, myocardial infarction, systemic embolism, or major bleeding, with each event type
79
weighted equally and counted to the composite only once. In Method 2, we modified the composite endpoint in Method 1 by substituting fatal or critical organ bleeding for major bleeding. This was done to exclude less-impactful, non-fatal extracranial hemorrhage events, which typically do not lead to “irreversible harm.” Therefore in Method 2, NCB was defined as the unweighted composite of all-cause death, stroke, myocardial infarction, systemic embolism, fatal bleeding, or critical organ bleeding. This approach of focusing on events that are fatal or cause irreversible harm is similar to the approach the U.S. Food and Drug Administration has described for its approval of other anticoagulants [9,10]. In Method 3, NCB was calculated using an approach based on a previous study by Singer et al., which compared warfarin versus no anticoagulation in patients with AF [4]. In this method, NCB was defined as the difference between the annualized rate of SSE minus 1.5 times the difference between the annualized rate of ICH. Thus, NCB was defined as: (SSErivaroxaban − SSEwarfarin) + 1.5 × (ICHrivaroxaban − ICHwarfarin). The weighting factor of 1.5 reflects the greater relative impact in terms of death and disability of an ICH relative to an ischemic stroke. A numerically negative result indicates that treatment with rivaroxaban is favored compared with warfarin. In Method 4, NCB was defined as the sum of the rate differences of SSE, ICH, major bleeding (MB) excluding ICH, and all-cause death (DE) in the two treatment groups, with each rate difference multiple by a weighting factor. This formula was based on a previous study of left atrial appendage closure by Gangireddy et al. [11] The weight assigned to each rate difference reflects the relative impact of that event type in terms of death and disability. Death is assigned the highest weight (1.00), and SSE is given a reference weight of 0.20. The weights of ICH and MB are based on an analysis of the Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events (ACTIVE) trials [12]. In this analysis, the hazard ratio (HR) for death after ICH was 3.08 compared with SSE, and the HR for DE after MB was 0.67 compared with SSE. Therefore, ICH is assigned a weight of 0.6 and MB a weight of 0.1. Therefore, NCB was calculated as: 1 × (DErivaroxaban − DEwarfarin) + 0.2 × (SSErivaroxaban − SSEwarfarin) + 0.6 × (ICHrivaroxaban − ICHwarfarin) + 0.1 × (MBrivaroxaban − MBwarfarin). We also conducted additional sensitivity analyses using weights of 0.1 and 0.3 for SSE, 0.3 and 0.9 for ICH, and 0.05 and 0.15 for MB. For methods 1, 2, and 4, ICH was not counted within MB to avoid “double counting” of this endpoint in both stroke and major bleeding.
2.4. Statistical analysis Baseline characteristics were summarized as counts (percentages) for categorical variables and as median values with 25th and 75th percentiles for continuous variables. Event rates were calculated as the number of events per 100 patient-years. The rate difference for each endpoint was calculated by subtracting the event rate per 100 patient-years of the warfarin group from the event rate of the rivaroxaban group (rivaroxaban minus warfarin). Thus, a negative rate difference indicates a lower event rate in the rivaroxaban group. The rate difference was multiplied by a factor of 100 and expressed as the number of additional events per 10,000 patient-years, to enable easier interpretation of the small rate differences as natural frequencies. The 95% confidence interval (CI) of the rate difference was calculated using empirical bootstrap resampling with 1000 replicates. Kaplan-Meier (K-M) cumulative event rates throughout the duration of the ROCKET AF follow-up period were calculated according to randomized treatment. The cumulative K-M rates were converted to cumulative rate differences by subtracting the K-M estimates for each treatment. The excess number of events (per 10,000 patients) between treatments over time was visualized by plotting the cumulative rate differences.
3. Results From December 18, 2006, through June 17, 2009, a total of 14,264 patients underwent randomization (intention-to-treat population) and a total of 14,236 patients received at least one dose of a study drug (on-treatment population). The median duration of treatment exposure was 590 days, and the median follow-up period was 707 days. The proportion of patients who stopped their assigned therapy before an endpoint event and before the termination date was 23.7% in the rivaroxaban group and 22.2% in the warfarin group. A total of 32 patients were lost to follow-up. Because of violations in Good Clinical Practice guidelines at one site that made the data unreliable, 93 patients (50 in the rivaroxaban group and 43 in the warfarin group) were excluded from all efficacy analyses. The key clinical characteristics of the patients included in the analysis (on-treatment population) are shown in Table 1. The median age was 73 years, and 39.7% of patients were female. Comorbid conditions were prevalent: 90.5% of patients had hypertension, 62.5% had heart failure, 39.9% had diabetes mellitus, and 54.7% of patients had a previous SSE or transient ischemic attack. The mean and median CHADS2 scores were 3.5 and 3.0, respectively, indicating a high-risk population. Previous use of vitamin K antagonists was reported by 62.4% of patients.
80
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
Table 1 Patient characteristics at baseline (on-treatment population, N = 14,236). Characteristic
Rivaroxaban (N = 7111)
Warfarin (N = 7125)
Age, year, median (25th, 75th percentiles) Female sex, n (%) Race, n (%) White Black Asian Other Region, n (%) Asia/Pacific Islands Eastern Europe Latin America North America Western Europe Type of atrial fibrillation, n (%) Persistent Paroxysmal Newly diagnosed CHADS2 score, mean ± standard deviation CHADS2 score, n (%) 1 2 3 4 5 6 Presenting characteristics, median (25th, 75th percentiles) Body mass index, kg/m2 Systolic blood pressure, mm Hg Diastolic blood pressure, mm Hg Heart rate, bpm Creatinine clearance, mL/mina Baseline comorbidities, n (%) Prior stroke / transient ischemic attack / systemic embolism Carotid occlusive disease Hypertension Diabetes Prior myocardial infarction Chronic heart failure Chronic obstructive pulmonary disease Prior medications Vitamin K antagonist Aspirin
73.0 (65.0,78.0) 2819 (39.6)
73.0 (65.0,78.0) 2826 (39.7)
5906 (83.1) 94 (1.3) 894 (12.6) 217 (3.1)
5952 (83.5) 85 (1.2) 887 (12.4) 201 (2.8)
1052 (14.8) 2746 (38.6) 939 (13.2) 1334 (18.8) 1040 (14.6)
1052 (14.8) 2747 (38.6) 938 (13.2) 1339 (18.8) 1049 (14.7)
5771 (81.2) 1242 (17.5) 98 (1.4) 3.5 ± 0.9
5754 (80.8) 1269 (17.8) 102 (1.4) 3.5 ± 0.9
1 (0.0) 923 (13.0) 3047 (42.8) 2087 (29.3) 930 (13.1) 123 (1.7)
2 (0.0) 932 (13.1) 3156 (44.3) 1998 (28.0) 879 (12.3) 158 (2.2)
28.3 (25.2,32.1) 130.0 (120.0,140.0) 80.0 (70.0,85.0) 76.0 (67.0,86.0) 67.0 (52.0,88.0)
28.1 (25.1,31.8) 130.0 (120.0,140.0) 80.0 (70.0,85.0) 76.0 (67.0,86.0) 67.0 (52.0,86.0)
3905 (54.9) 291 (4.1) 6419 (90.3) 2869 (40.3) 1178 (16.6) 4457 (62.7) 751 (10.6)
3889 (54.6) 300/ (4.2) 6468 (90.8) 2814 (39.5) 1282 (18.0) 4437 (62.3) 742 (10.4)
4431 (62.3) 2578 (36.3%)
4458 (62.6) 2616 (36.7%)
a
Creatinine clearance was calculated using the Cockcroft–Gault formula.
3.1. Outcomes Event rates and rate differences for the efficacy and safety endpoints examined in this study are shown in Table 2. The composite efficacy endpoint of all-cause death, stroke (ischemic or hemorrhagic), myocardial infarction, or systemic embolism occurred in 471 patients in the rivaroxaban group (4.25 per 100 patient-years) and in 575 patients in the warfarin group (5.12 per 100 patient-years). The rate difference for this endpoint was − 86.8 per 10,000 patient-years (95% CI -143.6 to −30.0) favoring treatment with rivaroxaban. Thus, treatment with rivaroxaban would be expected to prevent approximately 86.8 adverse events per 10,000 patient-years compared with treatment with warfarin. The rivaroxaban group had a lower cumulative event rate over the duration of the study (Supplemental Fig. 1A, log rank P = 0.003). This finding is also shown in Supplemental Fig. 2A, which plots the cumulative difference between the number of events in the rivaroxaban group and the warfarin group over time. There was a difference in the event rates of systemic embolism favoring treatment with rivaroxaban (Table 2, rate difference − 15.0; 95% CI −24.0 to −5.9). There were numerically, but not statistically significant, lower rates of all-cause death, stroke, and myocardial infarction in patients treated with rivaroxaban.
The composite safety endpoint of major bleeding occurred in 395 patients in the rivaroxaban group (3.60 per 100 patient-years) and 386 patients in the warfarin group (3.45 per 100 patient-years). There was no statistically significant difference in the rates of major bleeding (rate difference 14.2 per 10,000 patient-years; 95% CI − 35.2 to 63.7). The cumulative event rates were similar in the rivaroxaban and warfarin arms (Supplemental Fig. 1B, log rank P = 0.576), and a plot of the cumulative rate difference (Supplemental Fig. 2A) does not clearly favor warfarin or rivaroxaban. Considering both the more and less severe forms of major bleeding, the rate difference for the endpoint of fatal or critical organ bleeding was − 41.3 per 10,000 patient-years (95% CI -68.0 to − 14.7) and favored treatment with rivaroxaban. The cumulative rate of fatal or critical organ bleeding was also lower in the rivaroxaban group (Supplemental Fig. 1C, log rank P = 0.003) along with the cumulative rate difference (Supplemental Fig. 2B). The rate difference for major bleeding other than fatal or critical organ bleeding was 55.9 per 10,000 patient-years (95% CI 14.7 to 97.2) and favored treatment with warfarin. The cumulative rate of this category of major bleeding was also lower in the warfarin group (Supplemental Fig. 1D, log rank P = 0.008). This can also be seen in the plot of the cumulative rate difference (Supplemental Fig. 2C).
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
81
Table 2 Summary of efficacy and safety endpoints (on-treatment population, N = 14,236). Rivaroxaban (N = 7111) Event Efficacy Death, stroke, myocardial infarction, or systemic embolism All-cause death Strokec Ischemic stroke Myocardial infarction Systemic embolism Ischemic stroke or systemic embolism Safety Major bleeding Fatald or critical organ bleeding Fatald bleeding Critical organ bleeding Intracranial hemorrhage Non-fatal, non-critical organ bleeding a b c d
Rate difference (95% confidence interval)b
Warfarin (N = 7125)
No. events
Event ratea
No. events
Event ratea
471 208 184 149 101 5 160
4.25 1.87 1.65 1.34 0.91 0.05 1.44
575 250 221 161 126 22 194
5.12 2.21 1.96 1.43 1.12 0.19 1.72
−86.8 (−143.6, −30.0) −34.2 (−71.6, 3.1) −30.3 (−65.5, 4.8) −8.7 (−39.4, 22.1) −20.9 (−47.3, 5.5) −15.0 (−24.0, −5.9) −28.0 (−60.8, 4.9)
395 93 21 91 55 302
3.60 0.83 0.19 0.82 0.49 2.74
386 141 43 133 84 245
3.45 1.25 0.38 1.18 0.74 2.18
14.2 (−35.2, 63.7) −41.3 (−68.0, −14.7) −19.1 (−33.0, −5.2) −36.0 (−62.1, −10.0) −24.8 (−45.3, −4.3) 55.9 (14.7, 97.2)
Per 100 patient-years. Per 10,000 patient-years. Negative values favor rivaroxaban. Ischemic or hemorrhagic stroke. Narrow definition of fatal bleeding.
3.2. Net clinical benefit of rivaroxaban In Method 1, there was a trend towards NCB with rivaroxaban, although this did not reach significance (Table 3, rate difference − 35.5 per 10,000 patient-years; 95% CI − 108.4 to 37.3). In Method 2, rivaroxaban had an NCB of − 96.8 per 10,000 patient-years (95% CI −157.0 to −36.8); thus, treatment with rivaroxaban would be expected to prevent approximately 97 adverse events per 10,000 patient-years compared with warfarin. This finding was significant across almost all patient subgroups, with the exception of race marked as “other” (Supplemental Fig. 3). Of note, there was a large rate difference favoring warfarin in black patients that did not reach statistical significance (rate difference 617.4 per 10,000 patient-years; 95% CI − 85.3 to 1321.0). The event rates in this subgroup were very low (15 and 6 for rivaroxaban and warfarin, respectively). In Method 3, using the NCB method previously described by Singer et al. [4], rivaroxaban was associated with − 65.2 (95% CI − 112.3 to − 17.8; Table 3) fewer stroke equivalents per 10,000 patient-years compared with warfarin. In Method 4, using the NCB method previously described by Gangireddy et al. [11], rivaroxaban was associated with an NCB of − 54.8 (95% CI − 96.0 to − 10.2) fewer death equivalents per 10,000 patient-years (middle row, Table 4). The use of smaller weights for event types other than death reduced the NCB to − 44.5, but this was
still significant (95% CI − 82.0 to −4.8). The use of higher weights increased the NCB to −65.1 (95% CI −112.2 to −15.8). 3.3. Intention-to-treat population Similar results were achieved using the intention-to-treat population (Supplemental Tables 1–4). Rivaroxaban again had a lower rate of death, stroke, myocardial infarction, or systemic embolism, although the absolute value of the rate difference per 10,000 patient-years (RD) was numerically smaller compared with the on-treatment population (RD = −64.3, 95% CI −128.4 to −0.3). This was true as well for fatal or critical organ bleeding (RD = − 47.5, 95% CI − 75.0 to − 20.1). Rivaroxaban was favored in NCB methods 2–4, similar to the ontreatment population, although the absolute value of the rate differences were numerically smaller, and the 95% CI for method 3 barely crossed zero (Method 2: RD = −80.4, 95% CI −147.3 to −13.4; Method 3: RD = −47.2, 95% CI −100.2 to 2.0; Method 4: RD = −64.0, 95% CI −122.3 to −6.4). No difference was observed for Method 1 (RD = −14.1, 95% CI −91.6 to 63.4). 4. Discussion In this retrospective analysis, we estimated the NCB of rivaroxaban compared with warfarin using four different methods that included
Table 3 Net clinical benefit Methods 1–3 (on-treatment population, N = 14,236). Rivaroxaban (N = 7111)
Warfarin (N = 7125)
No. events
Event ratea
No. events
Event ratea
Method 1 Death + stroke + myocardial infarction + systemic embolism + major bleeding
811
7.42
863
7.78
−35.5 (−108.4, 37.3)
Method 2 Death + stroke + myocardial infarction + systemic embolism + fatal bleedingc + critical organ bleed
528
4.76
643
5.73
−96.8 (−157.0, −36.8)
Event
Method 3d based on Singer et al. [4] Net clinical benefit a
Net clinical benefit (95% confidence interval)b
−65.2 (−112.3, −17.8)
Per 100 patient-years. b Per 10,000 patient-years. Negative values favor rivaroxaban. c Narrow definition of fatal bleeding. d Net clinical benefit = (SSErivaroxaban − SSEwarfarin) + 1.5 × (ICHrivaroxaban − ICHwarfarin), where SSE = ischemic stroke or systemic embolism and ICH = intracranial hemorrhage.
82
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
Table 4 Net clinical benefit Method 4 based on Gangireddy et al. [11] (on-treatment population, N = 14,236). Net clinical benefit (95% confidence interval)a
Coefficients DE (A) SSE (B) ICH (C) MB (D) 1 1 1
0.1 0.2 0.3
0.3 0.6 0.9
0.05 0.10 0.15
−44.5 (−82.0, −4.8) −54.8 (−96.0, −10.2) −65.1 (−112.2, −15.8)
Net clinical benefit = A × (DErivaroxaban − DEwarfarin) + B×(SSErivaroxaban − SSEwarfarin) + C × (ICHrivaroxaban − ICHwarfarin) + D × (MBrivaroxaban − MBwarfarin), where DE = death, SSE = ischemic stroke or systemic embolism, ICH = intracranial hemorrhage, and MB = major bleeding excluding ICH. a Per 10,000 patient-years. Negative values favor rivaroxaban.
both ischemic and bleeding events. Rivaroxaban was associated with a favorable NCB in three out of the four methods. The overall NCB of rivaroxaban was attributable to both a reduced risk of ischemic events and a lower risk of fatal or critical organ bleeding compared with warfarin. However, rivaroxaban was associated with a higher risk of less clinically significant bleeding. Accordingly, an NCB was not observed in the one method that counted extracranial bleeding as equivalent to other hemorrhagic and ischemic event types (Method 1). Overall, these results support the use of rivaroxaban, which is associated with a decreased risk of ischemic and fatal or critical organ bleeding events, over warfarin in patients with AF. Antithrombotic agents are highly effective in preventing thromboembolism in patients with AF. However, this benefit can be partially or completely offset by the increased risk of bleeding. Therefore, to assess the overall benefit of oral anticoagulants, one must consider both the reduced risk of thromboembolism and the increased risk of bleeding. In this analysis, we used several different methods to estimate the NCB of rivaroxaban compared with warfarin in patients with AF at increased risk of thromboembolism, taking into account the clinical significance of the events involved. First, we compared the annualized rates of the composite of death, stroke, myocardial infarction, systemic embolism, and major bleeding in the two treatment groups (Method 1). Using this method, treatment with rivaroxaban was associated with a lower composite event rate, although this finding did not reach statistical significance. This method of calculating NCB weights major extracranial hemorrhages as equivalent to intracranial events, although the latter are far more likely to be fatal or result in long-term disability. Rivaroxaban had a higher rate of extracranial hemorrhages but a lower rate of intracranial events. We then conducted an analysis of NCB focused on events that can cause irreversible harm by substituting fatal or critical organ bleeding for major bleeding in the composite (Method 2). Extracranial bleeding events that did not lead to death, stroke, myocardial infarction, or embolism were excluded. This analysis favored rivaroxaban, with an estimated 96.8 events prevented per 10,000 patient-years compared with warfarin. This result is attributable to a reduced risk of both ischemic events and fatal or critical organ bleeding events in the rivaroxaban group (Table 2). Importantly, this finding was consistent across almost all patient subgroups, with the exception of race categorized as “other”, a relatively small group (N = 418, Supplemental Fig. 3). The strongest NCB with rivaroxaban was observed in patients with newly diagnosed AF (RD = −876.1 per 10,000 patient-years, 95% CI −1582.0 to −170.0; Supplemental Fig. 3), although this group of patients was relatively small (N = 202). Another way to estimate the NCB of rivaroxaban is to compare the risk of events that cause irreversible harm to the risk of events that are generally reversible. With rivaroxaban, 97 fewer patients per 10,000 patient-years would experience an event that is generally fatal or would cause irreversible harm (i.e., death, stroke, myocardial infarction, systemic embolism, or fatal or critical organ bleeding). In contrast,
56 more patients treated with rivaroxaban per 10,000 patient-years would experience transient transfusions and reductions in hemoglobin (i.e., major bleeding other than fatal or critical organ bleeding). This gives a ratio of 97/56 generally fatal or irreversible harm events prevented versus transient bleeding events caused, favoring rivaroxaban. While simply comparing event rates of ischemic and bleeding composite outcomes is informative, it does not account for the greater relative clinical impact of certain outcomes. For example, ICH is associated with a higher mortality rate relative to ischemic stroke [13,14]. Therefore, this outcome is sometimes multiplied by a weight N1 to reflect the greater clinical significance of these events [4]. In that case, a therapy must prevent more than one thromboembolic event to offset a single ICH in order to achieve a favorable NCB. We conducted two additional analyses using methods that apply weights to different outcomes to account for different relative clinical impacts. The first method (Method 3), previously published by Singer et al. in an analysis of the NCB of warfarin, applies a weight of 1.5 to the rate difference for ICH [4]. This method favored rivaroxaban, with an estimated 65.2 SSE equivalents prevented per 10,000 patient-years compared with warfarin. In the second analysis (Method 4), we calculated NCB as a sum of the rate differences of death, SSE, ICH, and major bleeding, each multiplied by a different weight [11]. Three different sets of weights were used, and all favored rivaroxaban. Using the baseline set of weights with SSE assigned to 0.2, rivaroxaban use resulted in an estimated 54.8 fewer death equivalents per 10,000 patient-years.
4.1. Limitations This analysis has several limitations. While we believe that these findings apply to most patients with AF, it is unclear whether they can be extrapolated to AF patients that are dissimilar from patients enrolled in ROCKET AF, such as patients at low risk of stroke. We used the ontreatment population for all analyses, which differed slightly from the intention-to-treat population and therefore may have been subject to residual confounding. However, we repeated all analyses using the intention-to-treat population and achieved similar results. We did not stratify warfarin patients by time in therapeutic range. It is possible that the NCB of rivaroxaban may be different if compared with warfarin populations with a higher or lower time in therapeutic range than the overall ROCKET AF population. However, in the ROCKET AF trial, the effect of rivaroxaban did not differ across quartiles of the duration of time that international normalized ratio values were within the therapeutic range.
5. Conclusions In conclusion, taking into account both ischemic and bleeding risk, rivaroxaban was associated with a favorable NCB in three out of four different methods used in this study. The NCB of rivaroxaban was attributable to lower rates of ischemic events and fatal or critical organ bleeding. There was an increased risk of lower severity bleeding with rivaroxaban but a decreased risk of fatal or critical organ bleeding. These results support the use of rivaroxaban over warfarin in patients with AF who are clinically similar to the cohort enrolled in ROCKET AF. In this patient population, treatment with rivaroxaban confers a lower risk of ischemic events and fatal or critical organ bleeding but may increase the risk of non-fatal or non-critical organ bleeding.
Funding ROCKET AF was supported by Johnson & Johnson Pharmaceutical Research and Development and Bayer HealthCare.
A.S. Barnett et al. / International Journal of Cardiology 257 (2018) 78–83
Disclosures JPP receives research funding from AHRQ, ARCA Biopharma, Boston Scientific, Gilead, Janssen Pharmaceuticals, ResMed, Spectranetics, and St Jude Medical and provides consulting to BMS/Pfizer, GSK, Johnson & Johnson, Medtronic, and Spectranetics. ZY and BL are salaried employees of Janssen Research & Development. SDB is employed as a clinical research physician at Bayer HealthCare Pharmaceuticals, Inc., Parsippany, NJ. SGG has received research grant support and speaker/ consulting honoraria from Bayer, Johnson & Johnson, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi-Sankyo, and Pfizer. DES provides consulting to Boehringer Ingelheim, Bristol-Myers Squibb, Merck, Johnson and Johnson and Medtronic and has received research grants from Boehringer Ingelheim, Bristol-Myers Squibb, and Medtronic. The remaining authors have nothing to disclose.
[3]
[4]
[5]
[6]
[7]
Conflict of interest The authors report no relationships that could be construed as a conflict of interest Appendix A. Supplementary data
[8]
[9] [10]
Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ijcard.2017.06.110.
[11]
References
[12]
[1] C.T. January, L.S. Wann, J.S. Alpert, H. Calkins, J.E. Cigarroa, J.C. Cleveland, J.B. Conti, P.T. Ellinor, M.D. Ezekowitz, M.E. Field, K.T. Murray, R.L. Sacco, W.G. Stevenson, P.J. Tchou, C.M. Tracy, C.W. Yancy, ACC/AHA Task Force Members, 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: executive summary: a report of the American College of Cardiology/American Heart Association Task Force on practice guidelines and the Heart Rhythm Society, Circulation 130 (2014) 2071–2104. [2] P. Kirchhof, S. Benussi, D. Kotecha, A. Ahlsson, D. Atar, B. Casadei, M. Castella, H.-C. Diener, H. Heidbuchel, J. Hendriks, G. Hindricks, A.S. Manolis, J. Oldgren, B.A. Popescu, U. Schotten, B. Van Putte, P. Vardas, Authors/Task Force Members, Document reviewers: 2016 ESC guidelines for the management of atrial fibrillation
[13]
[14]
83
developed in collaboration with EACTS: the task force for the management of atrial fibrillation of the European Society of Cardiology (ESC) developed with the special contribution of the Europea, Eur. Heart J. 37 (38) (2016) 2893–2962. L. Macle, J. Cairns, K. Leblanc, T. Tsang, A. Skanes, J.L. Cox, J.S. Healey, A. Bell, L. Pilote, J.G. Andrade, L.B. Mitchell, C. Atzema, D. Gladstone, M. Sharma, S. Verma, S. Connolly, P. Dorian, R. Parkash, M. Talajic, S. Nattel, A. Verma, CCS Atrial Fibrillation Guidelines Committee, 2016 focused update of the Canadian Cardiovascular Society guidelines for the management of atrial fibrillation, Can. J. Cardiol. 32 (2016) 1170–1185. D.E. Singer, Y. Chang, M.C. Fang, L.H. Borowsky, N.K. Pomernacki, N. Udaltsova, A.S. Go, The net clinical benefit of warfarin anticoagulation in atrial fibrillation, Ann. Intern. Med. 151 (2009) 297–305. P.S. Wells, M.H. Prins, B. Levitan, J. Beyer-Westendorf, T.A. Brighton, H. Bounameaux, A.T. Cohen, B.L. Davidson, P. Prandoni, G.E. Raskob, Z. Yuan, E.G. Katz, M. Gebel, A.W. Lensing, Long-term anticoagulation with rivaroxaban for preventing recurrent VTE: a benefit-risk analysis of EINSTEIN Extension, Chest 150 (5) (2016) 1059–1068. M.R. Patel, K.W. Mahaffey, J. Garg, G. Pan, D.E. Singer, W. Hacke, G. Breithardt, J.L. Halperin, G.J. Hankey, J.P. Piccini, R.C. Becker, C.C. Nessel, J.F. Paolini, S.D. Berkowitz, Fox KAA, R.M. Califf, ROCKET AF Investigators, Rivaroxaban versus warfarin in nonvalvular atrial fibrillation, N. Engl. J. Med. 365 (2011) 883–891. G. Renda, M. Di Nicola, R. De Caterina, Net clinical benefit of non-vitamin K antagonist oral anticoagulants versus warfarin in phase III atrial fibrillation trials, Am. J. Med. 128 (2015) 1007–1014.e2. ROCKET AF Study Investigators, Rivaroxaban-once daily, oral, direct factor Xa inhibition compared with vitamin K antagonism for prevention of stroke and embolism trial in atrial fibrillation: rationale and design of the ROCKET AF study, Am. Heart J. 159 (2010) 340–347.e1. B.N. Beasley, E.F. Unger, R. Temple, Anticoagulant options—why the FDA approved a higher but not a lower dose of dabigatran, N. Engl. J. Med. 364 (2011) 1788–1790. E.F. Unger, Weighing benefits and risks—the FDA's review of prasugrel, N. Engl. J. Med. 361 (2009) 942–945. S.R. Gangireddy, J.L. Halperin, V. Fuster, V.Y. Reddy, Percutaneous left atrial appendage closure for stroke prevention in patients with atrial fibrillation: an assessment of net clinical benefit, Eur. Heart J. 33 (2012) 2700–2708. S.J. Connolly, J.W. Eikelboom, J. Ng, J. Hirsh, S. Yusuf, J. Pogue, R. de Caterina, S. Hohnloser, R.G. Hart, ACTIVE (Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events) Steering Committee and Investigators, Net clinical benefit of adding clopidogrel to aspirin therapy in patients with atrial fibrillation for whom vitamin K antagonists are unsuitable, Ann. Intern. Med. 155 (2011) 579–586. E.M. Hylek, A.S. Go, Y. Chang, N.G. Jensvold, L.E. Henault, J.V. Selby, D.E. Singer, Effect of intensity of oral anticoagulation on stroke severity and mortality in atrial fibrillation, N. Engl. J. Med. 349 (2003) 1019–1026. M.C. Fang, A.S. Go, Y. Chang, E.M. Hylek, L.E. Henault, N.G. Jensvold, D.E. Singer, Death and disability from warfarin-associated intracranial and extracranial hemorrhages, Am. J. Med. 120 (2007) 700–705.