Scholarly article on topic 'Cost-Effectiveness of Apixaban Versus Other New Oral Anticoagulants for Stroke Prevention in Atrial Fibrillation'

Cost-Effectiveness of Apixaban Versus Other New Oral Anticoagulants for Stroke Prevention in Atrial Fibrillation Academic research paper on "Economics and business"

CC BY-NC-ND
0
0
Share paper
Academic journal
Clinical Therapeutics
OECD Field of science
Keywords
{"Stroke prevention" / apixaban / cost-effectiveness / "atrial fibrillation" / "new oral anticoagulant"}

Abstract of research paper on Economics and business, author of scientific article — Gregory Y.H. Lip, Thitima Kongnakorn, Hemant Phatak, Andreas Kuznik, Tereza Lanitis, et al.

Abstract Background Apixaban (5 mg BID), dabigatran (available as 150 mg and 110 mg BID in Europe), and rivaroxaban (20 mg once daily) are 3 novel oral anticoagulants (NOACs) currently approved for stroke prevention in patients with atrial fibrillation (AF). Objective The objective of this study was to evaluate the cost-effectiveness of apixaban against other NOACs from the perspective of the United Kingdom National Health Services. Methods A Markov model was developed to evaluate the pharmacoeconomic impact of apixaban versus other NOACs over a lifetime. Pair-wise indirect treatment comparisons were conducted against other NOACs by using ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation), RE-LY (Randomized Evaluation of Long-Term Anticoagulation Therapy), and ROCKET-AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) trial results for the following end points: ischemic stroke, hemorrhagic stroke, intracranial hemorrhage, other major bleeds, clinically relevant nonmajor bleeds, myocardial infarction, and treatment discontinuations. Outcomes were life-years, quality-adjusted life years gained, direct health care costs, and incremental cost-effectiveness ratios. Results Apixaban was projected to increase life expectancy versus other NOACs, including dabigatran (both doses) and rivaroxaban. A small increase in therapeutic management costs was observed with apixaban due to projected gains in life expectancy and lower discontinuation rates anticipated on apixaban versus other NOACs through lifetime. The estimated incremental cost-effectiveness ratio was £9611, £4497, and £5305 per quality-adjusted life-year gained with apixaban compared with dabigatran 150 mg BID, dabigatran 110 mg BID, and rivaroxaban 20 mg once daily, respectively. Sensitivity analyses indicated that results were robust over a wide range of inputs. Conclusions Although our analysis was limited by the absence of head-to-head trials, based on the indirect comparison data available, our model projects that apixaban may be a cost-effective alternative to dabigatran 150 mg BID, dabigatran 110 mg BID, and rivaroxaban 20 mg once daily for stroke prevention in AF patients from the perspective of the United Kingdom National Health Services.

Academic research paper on topic "Cost-Effectiveness of Apixaban Versus Other New Oral Anticoagulants for Stroke Prevention in Atrial Fibrillation"

Clinical Therapeutics/Volume 36, Number 2, 2014

Cost-Effectiveness of Apixaban Versus Other New Oral Anticoagulants for Stroke Prevention in Atrial Fibrillation^

Gregory Y.H. Lip1; Thitima Kongnakorn2; Hemant Phatak3; Andreas Kuznik4; Tereza Lanitis5; Larry Z. Liu6; Uchenna Iloeje7; Luis Hernandez8; and Paul Dorian9

''University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Birmingham, United Kingdom; 2Evidera, Bangkok, Thailand; 3Bristol-Myers Squibb, Princeton, New Jersey; 4Pfizer, New York, New York; 5Evidera, London, United Kingdom; 6Pfizer, New York, New York, Weill Medical College of Cornell University, New York, New York; 7Alexion Pharmaceuticals, Hartford, Connecticut; 8Evidera, Lexington, Massachusetts; and 9University of Toronto, St. Michael's Hospital, Toronto, Ontario, Canada

ABSTRACT

Background: Apixaban (5 mg BID), dabigatran (available as 150 mg and 110 mg BID in Europe), and rivaroxaban (20 mg once daily) are 3 novel oral anticoagulants (NOACs) currently approved for stroke prevention in patients with atrial fibrillation (AF).

Objective: The objective of this study was to evaluate the cost-effectiveness of apixaban against other NOACs from the perspective of the United Kingdom National Health Services.

Methods: A Markov model was developed to evaluate the pharmacoeconomic impact of apixaban versus other NOACs over a lifetime. Pair-wise indirect treatment comparisons were conducted against other NOACs by using ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation), RE-LY (Randomized Evaluation of Long-Term Anticoagulation Therapy), and ROCKET-AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) trial results for the following end points: ischemic stroke, hemorrhagic stroke, intracranial hemorrhage, other major bleeds, clinically relevant nonmajor bleeds, myocardial infarction, and treatment discontinuations. Outcomes were life-years, quality-adjusted life years gained, direct health care costs, and incremental cost-effectiveness ratios.

Results: Apixaban was projected to increase life expectancy versus other NOACs, including dabiga-tran (both doses) and rivaroxaban. A small increase in therapeutic management costs was observed with apixaban due to projected gains in life expectancy and lower discontinuation rates anticipated on apix-aban versus other NOACs through lifetime. The

estimated incremental cost-effectiveness ratio was £9611, £4497, and £5305 per quality-adjusted life-year gained with apixaban compared with dabigatran 150 mg BID, dabigatran 110 mg BID, and rivaroxaban 20 mg once daily, respectively. Sensitivity analyses indicated that results were robust over a wide range of inputs.

Conclusions: Although our analysis was limited by the absence of head-to-head trials, based on the indirect comparison data available, our model projects that apixaban may be a cost-effective alternative to dabigatran 150 mg BID, dabigatran 110 mg BID, and rivaroxaban 20 mg once daily for stroke prevention in AF patients from the perspective of the United Kingdom National Health Services. (Clin Ther. 2014;36:192-210) © 2014 The Authors. Published by Elsevier HS Journals, Inc. All rights reserved.

Key word: Stroke prevention, apixaban, cost-effectiveness, atrial fibrillation, new oral anticoagulant.

INTRODUCTION

Having atrial fibrillation (AF) increases a person's risk of experiencing stroke almost 5-fold.1 Traditionally, prophylactic treatment in this setting has been based

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial-No Derivative Works License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and source are credited. Current affiliation: Bristol-Myers Squibb, Princeton, NewJersey Accepted for publication December 17, 2013. http://dx.doi.org/10.1016/j.clinthera.2013.12.011 0149-2918/$-see front matter

© 2014 The Authors. Published by Elsevier HS Journals, Inc. All rights reserved.

on vitamin K antagonists (VKAs), drugs that have been in use for 60 years2 for their confirmed effectiveness in preventing thromboembolic events.3 However, the well-known challenges in managing VKA therapy, such as monitoring requirements and the risk of hemorrhages, have resulted in such therapy being underused in the treatment of AF.4 Given this context, the development of novel oral anticoagulants (NOACs), such as dabigatran, rivaroxaban, and apixaban, and their demonstrated efficacy in clinical trials have been encouraging with regard to addressing the need for improved stroke prevention treatments for patients with nonvalvular AF (NVAF).2'5 Dabigatran, a direct thrombin inhibitor, given at a dose of 110 mg BID, demonstrated non-inferiority to warfarin in the primary end point of stroke and systemic embolism coupled with a significantly lower risk of major hemorrhage.6 In addition, dabigatran 150 mg BID was superior to warfarin in the prevention of stroke and systemic embolism, with rates of major hemorrhage similar to warfarin. Rivaroxaban, an oral factor Xa inhibitor, was noninferior to warfarin in the prevention of stroke or systemic embolism, with no significant difference between the treatments in the risk of major bleeding.7 Apixaban, another oral factor Xa and the third NOAC to receive European Union marketing authorization for the prevention of stroke and systemic embolism in AF,8 is the only oral anticoagulant that has been shown to be superior to dose-adjusted warfarin in terms of reduction in the rates of stroke and systemic embolism, major bleeding, and all-cause mortality.9

This evidence on NOACs underpins current guidelines from the European Society of Cardiology, which recommend the use of these drugs as "broadly preferable to VKA in the vast majority of patients with NVAF."10 These drugs also offer the potential advantage of not requiring the anticoagulant monitoring needed for VKA therapy. The choice among NOACs, however, is not clear; this choice requires consideration of several practical issues, including patient characteristics, tolerability, and health economic outcomes.2,10,11 A key means of capturing such elements is a cost-effectiveness analysis that investigates how the differences in costs associated with therapy relate to differences in benefits. This analysis can be conducted by using modeling techniques, which are commonly accepted as valid

approaches to understanding the health economic consequences of different therapeutic alternatives.1 Of note, many such analyses have compared an individual NOAC (ie, apixaban, rivaroxaban, dabigatran) versus warfarin by using data from randomized clinical trial data,13-19 and several studies included all of the 3 NOACs from a US or Canadian perspec-tive.20-22 Crucially, however, no previous study has compared health economic outcomes between the 3 NOACs by using indirect treatment comparison data from a UK perspective conforming the drugs with their European labels.

From a health care payer's point of view, the absence of such data is a major gap in the evidence to inform decisions on resource allocation for NOACs. In particular, it is important to know whether the clinical advantages in terms of the efficacy and safety profile of apixaban over warfarin, as observed in randomized clinical trials, translate into health economic benefits, especially when compared with other NOACs, without head-to-head clinical trial data. The objective of the present study, therefore, was to assess the cost-effectiveness of apixaban (5 mg BID) versus the other NOACs (including dabigatran and rivarox-aban) approved for stroke prevention in patients with NVAF. The study was conducted from the perspective of the United Kingdom National Health Service.

METHODS

This study involved construction and use of an economic model to estimate long-term clinical and economic outcomes for patients with NVAF treated with apixaban, dabigatran, or rivaroxaban.

Model Design

The model used a Markov cohort approach. In the context of this study, such a model would conceptualize the course of AF by exploring what might happen over time to a hypothetical cohort of patients with the condition over a lifetime horizon. This analysis was performed by representing the disease course in terms of mutually exclusive health (or disease) states,12,23

such as NVAF without complications, NVAF with stroke, or NVAF with bleeding, that the patients can enter, remain in, or move ("transition") between as an approximation to potential real-life patient journeys. Time in a Markov model is represented as a recurring fixed interval, known as the model cycle.12,23 It is assumed that during each cycle, patients may remain in

their current health state (eg, NVAF in Figure 1) or experience an event (eg, ischemic stroke) that would cause them to move to at most 1 subsequent state (eg, ischemic stroke). The likelihood of each of these outcomes is known as its transition probability. These probabilities are built into the model and applied to the cohort during each cycle to calculate how the patients would be distributed between the thromboembolic and bleeding health states at the end of the cycle. This method in turn allows the model to calculate the related health care costs and benefits that will have accrued for the cohort as time has elapsed in the model.

A simplified schematic representation of the model structure for this study is shown in Figure 1. The patient cohort is assumed to start in the NVAF health state, and time continues to elapse in the model until all the patients end up in the death state,12 with calculation of the related accrued health care costs, life-years (LYs), and quality-adjusted life-years (QA-LYs) at the end of each model cycle. The cycle duration of 6 weeks was chosen deliberately to capture the possibility of events related to AF that occur within such a short time frame. The health states were either permanent, indicating that patients

No change

Apixaban

Dabigatran 110 mg

Dabigatran 150/110 mg

Systemic embolism -•

Myocardial infarction

NVAF subsequent ASA

Systemic embolism

Nonfatal Fatal

Systemic embolism

Ischemic stroke

Mild ■ 1 Mild IS

Moderate <1

Severe

' <1 Severe IS

Hemorrhagic stroke

Mild -4 1 Mild HS

Moderate ■<1 Moderate HS

Severe HS

^ Major bleed yx^, CRNM bleed Nonfatal Fatal

Myocardial infarction

Other deaths <

Nonfatal

Fatal Stay on I current AC Discontinue AC

Treatment interruption Discontinue AC

NVAF subsequent ASA

Myocardial infarction

Other AC discontinuation

NVAF subsequent ASA

Figure 1. Schematic representation of the Markov model. "M" represents a Markov process with 11 health states that are identical for each of the 3 novel oral anticoagulant treatment options. All patients remain in the nonvalvular atrial fibrillation (NVAF) state until 1 of the following events occurs: stroke, bleeding, systemic embolism, myocardial infarction (MI), treatment discontinuation, or death. The transition probabilities of these events occurring depend on the treatment. For patients on second-line aspirin ("NVAF subsequent ASA"), the events are identical; however, patients cannot experience any further discontinuation. Triangles indicate which health state the patient enters after an event. Health states colored in blue are permanent health states, with the rest being transient health states occurring for a period of 6 weeks before the patient returns to the previous state or moves to a subsequent health state. AC = anticoagulant; ASA = aspirin; CRNM = clinically relevant nonmajor; HS = hemorrhagic stroke; ICH = intracranial hemorrhage; IS = ischemic stroke.

Mild IS

Moderate IS

Severe IS

Nonfatal

Mild HS

Moderate HS

Other ICH

Severe HS

Rivaroxaban

Nonfatal

remain in them until death, or transient, suggesting that patients only spend some time in that health state (eg, clinically relevant nonmajor bleeding, other major bleeding, other intracranial hemorrhage [ICH]) before returning to their immediately previous health state.

In each cycle, the cohort is subjected to risks (ie, transition probabilities) of experiencing the following events that are health states within the model: ischemic or unspecified stroke (referred to as "ischemic stroke"); ICH (with a defined percentage of these events being assumed to be hemorrhagic stroke); other major bleeds (major bleeds that are not ICHs and that are further classified as being either gastrointestinal-or nongastrointestinal-related bleeds); clinically relevant nonmajor bleed; myocardial infarction (MI); systemic embolism; other cardiovascular hospitaliza-tion unrelated to the aforementioned events; or death.

Patients also have a defined risk of discontinuing their first-line treatment, which would cause them to transition to the "NVAF with subsequent aspirin treatment state," in which their risks of events in the following cycle are updated to those defined for their second-line aspirin treatment. Patients can enter this state after discontinuing their initial anticoagulant treatment owing to either ICH, other major bleeds, or other reasons unrelated to stroke, bleeding, MI, or systemic embolism. In this article, "other ICH" refers to ICHs that are not hemorrhagic strokes, with almost all cases being subdural hematoma.

For patients who experience nonfatal stroke (both ischemic and hemorrhagic) in the model, the severity of these events is classified into 1 of 3 categories of disability based on the modified Rankin Scale: mild, 0-2; moderate, 3 to 4; and severe, 5. Patients experiencing their first nonfatal stroke can experience 1 recurrent stroke in subsequent cycles. Those experiencing a recurrent stroke transition to the most severe health state between primary and recurrent strokes. Recurrent strokes are modeled as permanent states (ie, patients in these states are assumed to have no subsequent events until they transition to the death state). Similarly, nonfatal MI and systemic embolism are also modeled as permanent health states accumulating decrements in utility and additional costs over a lifetime.

Population

The population represented in the analyses comprised patients with AF suitable for VKA therapy. The

specific patient characteristics of the cohort (Table I) were matched to those of participants in ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation), a study that had compared apixaban with warfarin in patients with NVAF.

Comparators

The model allowed comparison of predicted outcomes for the cohort with first-line use of each of the following NOAC regimens: apixaban (5 mg BID),8 dabigatran 110 mg BID, dabigatran dose as recommended in the summary of product characteristics referred to as "dabigatran 150 mg" (ie, starting with 150 mg BID and switching to 110 mg BID at the age of 80 years), 5 and rivaroxaban (20 mg once daily).

Risk of Clinical Events

The rates of various clinical events in patients taking apixaban were obtained from the ARISTOTLE trial.9 To represent the comparative risks of such events for the other NOACs, hazard ratios (HRs) were applied to the rates for apixaban. These HRs were calculated by using an indirect treatment comparison of data from the trials of 3 drugs using the method of Bucher et al27 (see Supplemental Appendix A in the online version at http://dx.doi.org/10.1016/j.clinthera. 2013.12.011. In the absence of direct head-to-head studies, these indirect comparison analyses assumed that the treatment effect from NOACs was independent of patients' baseline characteristics in the following studies: ARISTOTLE9 (apixaban 5 mg BID vs warfarin, dose-adjusted to maintain an international normalized ratio [INR] of 2.0-3.0), the ROCKET-AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) study7 (rivaroxaban 20 mg once daily vs warfarin; INR, 2.0-3.0), and the RE-LY (Randomized Evaluation of Long-Term Anticoagulation Therapy) study6 (dabigatran 110 mg BID vs dabigatran 150 mg BID vs warfarin; INR, 2.0-3.0). Table I details the absolute event risks, the comparative HRs for dabigatran and rivaroxaban, and how patients experiencing certain events on the different NOACs would be distributed as to the severity or subtype of the event in ARISTOTLE.9

Clinical event rates for patients in the "NVAF with subsequent aspirin treatment" state (ie, those receiving

Table I. Demograph ic characteristics and clinical event rates according to treatment.

Patient Characteristics

Characteristic Source Characteristic Source

Starting age, y 70 6 CHADS2 distribution

Gender CHADS2: 0-1 34.0% 6

Male 64.7% 6 CHADS2: 2 35.8% 6

Female 35.3% 6 CHADS2: 3-6 Clinical Event Rates 30.2% 6

S ubsequent Aspirin (After Dabigatran Dabigatran Rivaroxaban

Discontinuation of (110 mg) (150 mg)

Initial Treatment)

Rate of Events per Hazard Ratio Versus Apixaban (95% CI)

Apixaban 100 Patient- years Source Source

Stroke rate 0.981 3.453 *t 1.198 (0.878-1.635) 0.823 (0.593-1.141) 0.980 (0.723-1.328) t

Intracranial hemorrhage 0.330 0.322 6t 0.733 (0.428-1.257) 1.020 (0.619-1.681) 1.731 (1.082-2.770) t

Other major bleed 1.790 0.887 6t 1.205 (0.965-1.504) 1.371 (1.102-1.705) 1.463 (1.150-1.793) t

Clinically relevant 2.083 2.936 *t 1.155 (0.986-1.354) 1.303 (1.113-1.526) 1.488 (1.261-1.755) t

nonmajor bleed

Other treatment 13.177 *t 1.452 (1.309-1.611) 1.505 (1.357-1.668) 1.184 (1.083-1.294) t

discontinuation

Myocardial infarction 0.530 1.110 t 1.474 (0.958-2.269) 1.456 (0.948-2.238) 0.935 (0.635-1.375) t

Systemic embolism 0.090 0.400 t 1.000 1.000 1.000 Assumption^

Other cardiovascular 10.460 13.571 *t 1.000 1.000 1.000 Assumption^

hospitalization

Other death rate 3.0825 *t Distributi 1.000 ons and Probabilities by Treatment 1.000 1.000 Assumption^

Stroke severity

distribution

Mild (mRS 0-2) 53% 36% *t 35% 35% 49% 10,27

Moderate (mRS 3-4) 21% 38% *t 28% 22% 18% 10,27

Severe (mRS 5) 8% 15% *t 10% 8% 6% 10,27

Fatal (mRS 6) 18% 11% *t 27% 35% 27% 10,27

Hemorrhagic stroke 77% 55% 6t 64% 41% 57% 10,27

among intracranial

hemorrhage*

(continued)

Table I (continued).

Patient Characteristics

Characteristic Source Characteristic Source

Hemorrhagic stroke 10,27

severity distribution

Mild (mRS 0-2) 23% 7% 't 35% 35 49% 10,27

Moderate (mRS 3-4) 32% 20% 't 28% 22% 18% 10,27

Severe (mRS 5) 10% 27% 't 10% 8% 6% 10,27

Fatal (mRS 6) 35% 46% "t 27% 35% 27% 10,27

Gastrointestinal bleeds 38% 39% 6t 41% 49% 45% 10,27

among other major

bleeds

Patients experiencing 1.67% 1.58% 6,Assumption^ 3.69% 3.53% 1.67% 10,27

dyspepsia

throughout

treatmenta

Patients requiring 0.00% 0.00% Assumption^ 19.40% 19.40% 0.00% 10,27

annual renal

monitoring

CHADS2 = congestive heart failure, hypertension, age >75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism; mRS = modified

Rankin scale.

*Supplemental Appendix B.

^Supplemental Appendix C.

^Supplemental Appendix A.

§Where data was unavailable for novel oral anticoagulants, we assumed the same rate as for patients treated with apixaban and varied the inputs in the deterministic

sensitivity analysis.

¡Data for dabigatran and rivaroxaban were adjusted, using warfarin as a common arm, to the respective proportions observed in the ARISTOTLE (Apixaban for

Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) trial.

^Assumed same proportion as for aspirin first-line as observed in the AVERROES (Apixaban Versus Acetylsalicylic Acid to Prevent Strokes) trial.

#Dabigatran requires annual renal monitoring for patients with moderate to severe renal impairment as per the summary of product characteristics, whereas this

monitoring is not required for apixaban and rivaroxaban.

№ №

Table II. Treatment switch on the occurrence of events.

% of Patients Switching Subsequent Impact on Future

Event Treatment Treatment Transitions Source

Ischemic stroke or systemic embolism 0 (patients treated with Warfarin No Clinical

NOACs) opinion*

100 (patients treated with

second-line aspirin)

Hemorrhagic stroke or myocardial 100 No No

infarction treatment

Other intracranial hemorrhage 56 Aspirin Yes 28

Other major bleed 25 Aspirin Yes 29

Other treatment discontinuation 100 Aspirin Yes Clinical

(unrelated to events modeled) opinion*

NOACS = novel oral anticoagulants.

*Clinical opinion to advise the treatment patterns upon the occurrence of each event was solicited from 2 cardiologists who

were part of the research team (Drs. Lip and Dorian).

aspirin after discontinuation of initial NOAC treatment) were based on the analysis of a subgroup of patients in the AVERROES (Apixaban Versus Acetylsalicylic Acid to Prevent Strokes)30 trial who had been previously prescribed, but discontinued, VKA and were then treated with aspirin instead (Table I). Assumptions about the effects of treatment discontinuation on the occurrence of events are presented in Table II. The risks of ischemic stroke, bleeding, and MI were increased progressively by a factor of 1.46,31 1.97,32 and 1.3033 per decade, respectively, to account for the increased likelihood of these events with aging.

The risk of recurrence postischemic or hemorrhagic stroke was assumed to be the same with all treatments and estimated to be 2.72 per 100 patient-years.

Mortality

On the occurrence of each event, case-fatality rates based on data from trials (Table I) were applied to represent the risk of dying as a result of the episode. The case fatality rates for MI were obtained from published literature and estimated to be 10.8% in male subjects and 15.6% in female subjects. 5 Case fatality rates for other ICHs, other major bleeds, and systemic embolism were 13%, 2%, and 9.4%, respectively. Estimates were pooled from the

AVERROES30 and ARISTOTLE9 trials and were assumed to be the same for all treatments (see Supplemental Appendix A and B in the online version at http://dx.doi.org/10.1016/jxlinthera.2013. 12.011).

All-cause mortality rates for apixaban excluding deaths attributable to stroke, bleeding, MI, and systemic embolism were derived from the ARISTOTLE trial9 and were applied to the cohort during the initial 1.8 years in the model (ie, a period matching the trial duration). These were assumed to be the same for the other NOACs. Beyond 1.8 years, mortality rates were modeled based on age- and gender-specific general mortality data in UK life tables. In addition, an HR to account for the higher mortality associated with the following conditions was applied to the general life tables and to patients in each respective health state: AF (excluding excess mortality due to the events modeled), stroke, systemic embolism, or MI events (Table III).

Utilities

Utility inputs (Table III) were obtained from a UK-based utility catalogue.37 The disutility associated with the use of NOACs was assumed to be the same as that for aspirin.38

Table III. Utility and mortality estimates for each health state.

Hazard Ratios Versus

Health State Utility (SE) Source General Population (95% CI) Source

Nonvalvular atrial fibrillation 0.7270 (0.0095) 37 1.34 (1.20-1.53) 39

Ischemic or hemorrhagic stroke

Mild 0.6151 (0.0299) 37 3.18 (1.82-4.92) 40-42

Moderate 0.5646 (0.0299) 37 5.84 (4.08-7.60) 40-42

Severe 0.5142 (0.0299) 37 15.75 (13.99-17.51) 40-42

Myocardial infarction

Females 0.6151 (0.0299) 37 4.16 (2.27-2.88) 43,44

Males 0.5646 (0.0299) 37 2.56 (3.44-5.03) 43,44

Systemic embolism 0.6265 (0.0299) 37 1.34 (1.20-1.53)

Transient Health States/ Utility Decrement Assumption*

Anticoagulation Use Utility Decrement Source Duration Source

Other intracranial hemorrhage 0.1511 (0.0401) 37 6 weeks Clinical opinion'

Other major bleeds 0.1511 (0.0401) 37 2 weeks Assumption'

Clinically relevant nonmajor bleeds 0.0582 (0.0173) 37 2 days Assumption'

Other cardiovascular hospitalization 0.1276 (0.0259) 37 6 days Clinical opinion'

Treatment with new oral anticoagulants 0.0020 (0.00-0.04) 38 While on treatment

or aspirin

*In absence of data to inform the hazard ratio of mortality for patients with systemic embolism, we assumed patients with

systemic embolism would follow a similar mortality pattern to patients with nonvalvular atrial fibrillation. This assumption

was tested in the deterministic sensitivity analysis.

'Assumptions around the duration of these utility decrements were derived upon discussion with 2 cardiologists who were

part of the research team (Drs. Lip and Dorian).

'Assumptions based on Freeman et al.13

Because the perspective adopted was that of the United Kingdom National Health Service health care payer, direct health care costs were included in the analysis. Costs in the model are reflected in 2011 British pounds and are categorized as either acute care costs relating to time spent in the hospital and rehabilitation facilities (assumed to be 2 weeks in the base case analysis described in the following discussion) or maintenance costs applied to the remainder of a patient's lifetime (Table IV). Health and cost outcomes were discounted at 3.5% per year.45

Analyses

The analyses compared apixaban with the other NOACs in patients with NVAF suitable for VKA treatment. Specifically, this analysis involved predicting clinical and economic outcomes for a cohort of

1000 such patients over their lifetime, by calculating the LYs, QALYs, and costs accumulated over this period depending on which NOAC they were started on. In the primary analysis (the base case), the various predetermined data inputs described earlier were used in the model. To assess whether potential clinical advantages of apixaban over other NOACs (as suggested by trial data) would be worth the money spent on the drug, an incremental cost-effectiveness ratio (ICER [additional cost per additional QALY gained]) was calculated for each comparison with the other 2 drugs. The ICER was then compared with the commonly accepted UK payers' willingness-to-pay threshold of £20,000 for each QALY gained45 (ie, if the ICER of apixaban vs other another NOAC was below this threshold, then apixaban was considered a cost-effective treatment alternative).

Table IV. Resource use and unit costs.

Variable Cost, £ (95% CI) Unit Source

Daily cost of apixaban 2.20 10 mg daily dose 46

Daily cost of dabigatran 110 or 150 mg 2.20 220 or 300 mg daily dose 46

Daily cost of rivaroxaban 2.10 20 mg daily dose 46

Monitoring 20.69 (17-25) Per month 47

Dyspepsia 83.19 (48-129) Per year 46,47

Annual renal monitoring test 3.00 Per test 47

Acute Cost, £ Maintenance Cost, £

Event (per Episode) (per Month) Source

Ischemic and hemorrhagic stroke

Mild 6815.00 ( 5993-7410) 145.24 (86-200) 48

Moderate 6436.88 (5793-6870) 158.31 (98-216) 48

Severe 14,107.41 (12,589-15,166) 445.82 (375-200) 48

Fatal 9063.23 (7158-12,978) — 48

Other intracranial hemorrhage 3010.00 (2190-3456) — 47

Other major bleeds

Gastrointestinal bleeds 1493.68 (1237-1825) — 47

Nongastrointestinal-related 3947.92 (2508-4554) — 47

Clinically relevant nonmajor bleed 1133.93 (751-1284) — 47

Myocardial infarction 2018.84 (1596-2554) 6.45 (4-10) 47,49

Systemic embolism 6815.00 ( 5993-7410) 145.24 (86-200) Assumed to be the

same as mild stroke48

Other cardiovascular hospitalization 1570.89 (1140-1798) — 46

The sensitivity of the base case results was evaluated by conducting 1-way sensitivity analyses. These consisted of varying the model parameters, using their 95% CIs, one by one, while keeping all others constant and re-generating results to evaluate the robustness of the model's base case predictions in relation to uncertainties in the model parameters.

In addition to 1-way sensitivity analysis assessing the statistical uncertainty around the identified parameters, further scenario analysis was conducted to test uncertainties in structural assumptions as well as the plausibility and relevance of certain input data. These scenarios were determined by the 2 cardiologists that were part of the research team (Drs. Lip and Dorian) and specifically involved testing assumptions around the following: (1) treatment discontinuation; (2) stroke severity; (3) bleed severity; (4) costs and utilities of MI in current practice; and (5) utility decrement for ICH.

In addition to these analyses, probabilistic sensitivity analyses were performed by running 2000 iterations (or "simulations") of a 1000-patient cohort

entering the model, with the values for key model inputs being varied between each iteration. The specific inputs for each iteration were obtained by random sampling from probability distributions of the parameters concerned. The fact that these inputs varied between iterations meant that every iteration would generate its own prediction of incremental costs and effectiveness and therefore its own ICER. The results of the probabilistic analysis were plotted on scatter-diagrams depicting the additional gains in QALYs, with apixaban compared with the other NOACs (on the x-axis) against the additional costs of the drug (on the y-axis). The probabilistic results were also used to generate a cost-effectiveness acceptability curve (CEAC), representing the proportion of simulations for which each treatment was the optimal treatment option at a given willingness-to-pay threshold. Warfarin and aspirin were included as treatment alternatives in the probabilistic analysis to assess the probability of apixaban being the most cost-effective option among all drugs used for preventing stroke in

routine practice for patients with AF. Data for warfarin and aspirin are displayed in in the online version at http://dx.doi.org/10.1016/jj.clinthera.2013. 12.011.

RESULTS

Base Case Analysis

For a cohort of 1000 patients considered over their lifetime, starting treatment with apixaban rather than another NOAC was predicted to result in fewer strokes (first and recurrent ischemic and hemorrhagic strokes), systemic embolisms, and cardiovascular-related deaths (Table V). Patients treated with apixaban were also predicted to experience fewer major bleeds compared with dabigatran 150 mg and

rivaroxaban. However, compared with dabigatran 110 mg, apixaban increased the number of major bleeds (first and recurrent hemorrhagic strokes, other ICHs, and other major bleeds) (15 episodes over the lifetime horizon).

In terms of therapeutic management costs, compared with dabigatran 110 mg, dabigatran 150 mg, and rivaroxaban, respectively, apixaban yielded additional anticoagulant (drug) and management costs of £713, £794, and £515 with average cost-offsets in monitoring and clinical event-related costs (lifetime reduction) of £249, £140, and £269; this led to a net increment in total cost over a projected lifetime of £464, £654, and £246. Apixaban's additional benefit in reducing the number of various clinical events led to

Table V. Base case results over lifetime: clinical events per 1000 adjusted life-years (QALYs), and costs per patient. patients, mean life-years (LYs), quality-

Variable Apixaban Dabigatran 110 mg Dabigatran 150 mg Rivaroxaban

No. of events (per cohort of 1000)

Ischemic stroke 229 250 240 235

Recurrent ischemic stroke 18 19 18 18

Hemorrhagic stroke 26 18 17 28

Recurrent hemorrhagic stroke 1 1 1 2

Systemic embolism 24 26 27 26

Myocardial infarction 84 96 97 84

Other intracranial hemorrhage 12 13 18 22

Other major bleed 165 157 170 197

Clinically relevant nonmajor bleed 287 287 303 338

Other cardiovascular hospitalization 1186 1178 1195 1187

Other treatment discontinuation 635 715 729 668

Event related 370 396 388 381

Other 631 603 612 619

Health outcomes (per patient)

LYs (undiscounted) 11.14 10.96 11.02 11.06

QALYs (discounted) 6.26 6.16 6.19 6.21

Costs (£ discounted per patient)

Anticoagulant and management 3555 2842 2761 3040

Monitoring 106 128 132 117

Clinical events 5417 5644 5531 5675

Total 9078 8614 8424 8832

Incremental cost-effectiveness ratio

(apixaban vs comparator)

£ per QALY gained 4497 9611 5305

Table VI. Scenario analysis: incremental cost-effectiveness ratio (ICER) of apixaban versus the comparator (percent deviation from base case). Unless otherwise noted, values are 2011 British pounds.

Variable Dabigatran 110 mg Dabigatran 150 mg Rivaroxaban

Base 4497 9611 5305

Stroke severity independent of treatment 4824 (7.3%) 11,721 (22.0%) 5366 (1.2%)

Bleed severity independent of treatment 4157 (-7.6%) 7135 (-25.8%) 2835 (-46.6%)

Stroke and bleed severity independent of treatment 4388 (-2.4°%) 8052 (-16.2%) 2501 (-52.9%)

Treatment discontinuation rates set to be the same 2417 (-46.3°%) 9039 (-6.0%) 3299 (-37.8%)

as apixaban after trial period (1.9 years)

Treatment discontinuation rates set to be 0 after 3229 (-28.2%) 8956 (-6.8%) 3690 (-30.4%)

trial period (1.9 years)

No treatment discontinuation unrelated to stroke, 412 (-90.8%) 8596 (-10.6%) 1308 (-75.3%)

MI, SE, and bleeding events

Assume no difference in relative efficacy of MI 5050 (12.3%) 12,829 (33.5%) 5110 (-3.7%)

MI utility increased (assumed to be the same as 4811 (7.0%) 10,708 (11.4%) 5279 (-0.5%)

that for patients with AF)

ICH disutility increased to 0.3 4497 (0%) 9600 (-0.1%) 5290 (-0.3 %)

HR of mortality for AF set to 1 4178 (-7.1%) 8515 (-11.4%) 4908 (-7.5%)

AF = atrial fibrillation; MI = myocardial infarction; SE hazard ratio. = systemic embolism; ICH = intracranial hemorrhage; HR =

an additional 0.18, 0.12, and 0.08 undiscounted LY and an additional 0.10, 0.07, and 0.05 discounted QALY per patient, compared with dabigatran 110 mg, dabigatran 150 mg, and rivaroxaban, respectively. The result was the respective ICERs of £4497, £9611, and £5305 per QALY gained (Table VI).

One-Way Sensitivity Analyses and Scenario Analyses

Figures 2A through 2C present the results from the sensitivity analyses for the top 15 parameters that had the largest effect on the ICERs, in the order of their respective influence. The statistical uncertainty around model inputs not included on these tornado diagrams had negligible impact on the ICER results. Table VI presents the results of the scenario analysis around structural uncertainties as well as the plausibility and relevance of certain input data.

Compared with dabigatran 110 mg, the ICERs from all sensitivity analyses and scenarios varied from apixaban being dominant (ie, providing a higher number of QALYs at a lower cost) to the drug being associated with additional costs of £11,307 per QALY

gained (Figure 2A; Table VI). The comparison against dabigatran 150 mg showed that the ICERs from all scenarios varied from apixaban being dominant to its incurring an additional £32,717 per QALY gained (Figure 2B). In the comparisons between apixaban and rivaroxaban, the ICERs from all scenarios varied between apixaban dominating and being dominated (Figure 2C), with the latter scenario occurring when greater disutility was assigned to use of apixaban, despite the drug's anticipated advantage in relation to bleeding and adverse events (Table I; see also Supplemental Appendix A in the online version at http://dx.doi.org/10.1016/j.clinthera.2013.12.011.

Compared with the other NOACs, the 4 scenarios tested resulted in an ICER for apixaban above the commonly accepted threshold of £20,00045 per QALY: (1) increasing the disutility of treatment with apixaban to >5 times that associated with other NOACs (thus equaling the disutility for patients treated with warfarin [ie, 0.013]); (2) decreasing the ischemic stroke rate for aspirin used as second-line therapy by 75% to 1.97 per 100 patient-years, thereby benefiting treatments with higher

Figure 2

Daily cost apixaban Daily cost dabigatran 110 mg Stroke HR dabigatran 110 mg Stroke rate aspirin second-line Disutility associated with apixaban HR of mortality for dabigatran 110-mg trial period ICH rate apixaban ICH HR dabigatran 110 mg Routine care cost per month Disutility associated with dabigatran 110 mg Treatment discontinuation HR dabigatran 110 mg CV hospitalization rate apixaban CV hospitalization rate aspirin second-line Other ma|or bleeds HR dabigatran 110 mg Myocardial Infarction HR dabigatran 110 mg

-10,000 -5000 0 5000 10,000 15,000 20,000 25,000 Incremental Cost per QALY (£)

Disutility associated with apixaban Daily cost apixaban Stroke rate aspirin second-line Daily cost dabigatran 150 mg Stroke HR dabigatran 150 mg HR of mortality for dabigatran 150-mg trial period Stroke rate apixaban Myocardial infarction HR dabigatran 150 mg ICH rate apixahan ICH HR dabigatran 150 mg Disutility associated with dabigatran 150 mg Stroke risk adjustment factor per decade Disutility associated with aspirin second-line % Hemorrhagic strokes among ICH dabigatran 150 mg CV hospitalization risk aspirin second-line

-10,000 0 10,000 20,000 30,000 40,000 Incremental Cost per QALY {£)

Base Case 4497per QALY

,7, Dominant 1 11,307

Dominant 9898

,«4 1860 I 9760

5.34 2536 ■ I 7858

4146 | I 8423 <Ш1

ш 3737 | □ 5985

0.IS 3629 | | 5624 «Й

3506 | | 5268 ОЛ

4497 □ 6112 „3

1 ™ 1 3284 | | 4821 ■

3705 | | 5093

и 3911 | | 5241

12a8a 1 3779 | | 5066 7.7S

и 3872 | | 5039 <И7

2*7 3967 | 9 5114 0»

Base Case 9611 per QALY

Disutility associated with apixaban Stroke HR rivaroxaban Daily cost apixaban Daily cost rivaroxaban HR of mortality for rivaroxaban during the trial period Stroke HR rivaroxaban Stroke rate aspirin second-line Disutility associated with rivaroxaban Other major bleed HR rivaroxaban Stroke rate apixaban % Hemorrhagic strokes among ICH rivaroxaban CV hospitalization HR rivaroxaban Myocardial infarction HR rivaroxaban Treatment discontinuation HR rivaroxaban Stroke risk adjustment factor per decade

-15,000 5000 25,000 45,000

Incremental Cost per QALY (£)

Base Case 5305 per QALY

2.«4 ] ,«] [„,] H

,.97 j

[j-« ]

Results of the sensitivity analyses: (A) apixaban versus dabigatran 110 mg; (B) apixaban versus dabigatran 150 mg; and (C) apixaban versus rivaroxaban. Rates are displayed per 100 patient-years.The solid vertical line represents the base case incremental costs per quality-adjusted life year (QALY) for apixaban compared with the respective novel oral anticoagulant. Horizontal bars indicate the range of incremental costs per QALY obtained by setting each variable to the values shown in the boxes while holding all other values constant. HR = hazard ratio; CV = cardiovascular; ICH = intracranial hemorrhage.

Incremental QALY Gained (Over Lifetime)

Willingness to Pay (per QALY; £)

Figure 3. Results of the probabilistic sensitivity analyses: (A) apixaban versus dabigatran 110 mg; (B) apixaban versus dabigatran 150 mg; (C) apixaban versus rivaroxaban; and (D) cost-effectiveness acceptability curves. Each line (A-C) represents a cost-effectiveness threshold representing the maximum amount society is willing to pay for a quality-adjusted life-year (QALY) gain. Apixaban is a cost-effective alternative in cases that fall to the right of this line; apixaban is not a cost-effective alternative in cases that fall to the left of this line.

discontinuation rates; (3) decreasing the HR of stroke for rivaroxaban in reference to apixaban to 0.72 (from 0.98 in the base case); and (4) increasing the daily cost of apixaban to £2.64 at a 26% premium to other NOACs.

Probabilistic Sensitivity Analysis

Probabilistic sensitivity analysis demonstrated that apixaban was more effective at a small additional cost versus other NOACs over a lifetime horizon (Figures 3A-3C). The ICER was below the commonly assumed threshold of £20,000 per QALY gained45 in 98% of trials comparing apixaban with dabigatran 110 mg, in 83% of trials compared with dabigatran 150 mg, and in 85% of trials compared with rivaroxaban. The results of the probabilistic analysis when all treatment comparators were included are shown as multi-way CEACs (Figure 3D). The CEAC, when generated by using the data on outcomes with NOACs from the indirect treatment comparisons and the data from direct head-to-head comparisons against warfarin and aspirin (see Supplemental Appendices B and C in the online version at http://dx.doi.org/10.1016/j.clinthera. 2013.12.011, indicated that apixaban was an optimal treatment choice representing the maximum net benefit over aspirin, warfarin, and the other NOACs, assuming payers are willing to pay £15,000 per QALY gained.

DISCUSSION

This study assessed the cost-effectiveness of apixaban as a first-line treatment in the prevention of stroke for patients with AF eligible for treatment with VKA, compared with other NOACs. Patients starting treatment with apixaban were predicted to experience fewer strokes and cardiovascular-related deaths compared with those taking the other NOACs. Although a cost-effectiveness analysis including comparisons of NOACs versus warfarin for AF has recently been published,20-22 our study seems to be the first to be conducted from a UK perspective, also incorporating use of dabigatran as suggested by its European

label.25

Interestingly, an assessment limited to consideration of the indirect comparison data would give a favorable impression of dabigatran 150 mg compared with apixaban in terms of reduction of stroke risk (ie, HR versus apixaban = 0.823); however, a broader overall consideration of efficacy, major bleeding, and

tolerability profile extrapolated over a lifetime suggests that apixaban therapy would result in fewer strokes. This potential reduction in events (compared with other NOACs) is attributable to: (1) a switch to dabigatran 110 mg when patients reached age 80 years (as recommended by the summary of product characteristics),25 where its protective effect was poorer than apixaban (HR VS apixaban = 1.198); and (2) fewer patients discontinuing first-line treatment and starting second-line treatment (ie, aspirin) in the apixaban arm because of fewer major bleeds and, therefore, anticipated lower discontinuation rates. Consequently, lower discontinuation rates and assumed parity in drug acquisition costs resulted in apixaban-treated patients gaining a stroke prevention benefit over a longer period in the model at a small increase in pharmacologic treatment costs over the lifetime horizon.

Bleeding outcomes were predicted to be less likely with apixaban in general, except for a slightly higher number of expected hemorrhagic strokes in the comparisons with dabigatran 110 mg and dabigatran 150 mg (27 vs 19 and 18, respectively [in the cohort of 1000 patients]), as well as a slight increase in the number of other major bleeds compared with dabiga-tran 110 mg (165 vs 157 [in the cohort of 1000 patients]). However, the overall benefit of apixaban outweighed these effects, as shown by an increase in LYs and QALYs in comparisons with both dabigatran and rivaroxaban in the analysis.

Our model was similar to those in earlier cost-effectiveness studies assessing NOACs versus warfarin or among each other that used a Markov approach to represent potential stroke and bleeding events related to AF over a patient's lifetime.13,18-21,29 As with our study, some of these other trials examined the impact of treatment discontinuation.18-22,29,13 However, caution is needed when comparing their results with those of our study, given the key differences in the study designs. In the most recent published studies, which compared all NOACs against warfarin and among each other,20,21 dabigatran 150 mg was not modeled according to the European label (ie, switching to 110 mg at age 80 years).25 Had we adopted a similar approach, assuming patients remain on dabigatran 150 mg through their lifetime, apixaban would still be expected to be cost-effective with an ICER of £9913 under the base case assumptions. In addition, treatment discontinuation was not explicitly modeled.20,21

This limitation, in addition to the others, could account for the fact that these studies predicted (at odds with ours) a slightly higher number of QALYs gained in patients treated with dabigatran 150 mg compared with rivaroxaban (the result of the study excluding explicit modeling of treatment discontinuation, which was lower in patients treated with rivar-oxaban than those treated with dabigatran 150 mg). By contrast, treatment discontinuation rates used in our study were those observed in the clinical trials, and the risk of further events in patients receiving second-line treatment with aspirin were based on secondary analysis of the subgroup of participants in the AVERROES30 trial who had been "demonstrated" to be unsuitable for VKA (rather than "expected") at randomization (42% of the trial population). This approach allows more accurate modeling of treatment patterns and the clinical event risks in these patients, and is a key strength of our study. Furthermore, results were robust to changes around the assumptions of future treatment discontinuation rates. When setting treatment discontinuation rates for rivaroxaban, dabigatran 110 mg, and dabigatran 150 mg to equal those of apixaban beyond the duration of the trial (ie, 1.8 years), the ICERs decreased relative to the base case to £3296, £2417, and £9614, respectively, due to increase in costs among the NOACs arms; this outcome thus demonstrates that the base case results presented are conservative.

Additional important differentiators of our analysis from previous work include its more detailed modeling of mortality. This analysis was conducted with the use of mortality rates (for an initial period in the model equivalent to the duration of the ARISTOTLE9 trial) and incorporation of an HR for increased mortality (beyond this initial period) for patients with AF compared with rates in the general population (in both cases, excluding mortality due to strokes, MI, and systemic embolism). Other models have assumed that patients with AF who had not experienced any events would follow mortality

patterns similar to the general population,3,18,21,29

despite clear published evidence indicating that mortality of patients with AF is higher than that in the general population, even after adjusting for mortality rates due to stroke and MI.39,50 Our study's incorporation of increased background mortality estimates represents a more cautious modeling approach, in that

it reduced the predicted number of additional QALYs gained from treatment, therefore leading to higher estimates for the ICER (Table VI). Furthermore, we included additional granularity and detail surrounding the severity of stroke and bleeding events, allowing these to be dependent on anticoagulant treatment based on data from the trials. Although ARISTOTLE9 highlighted apixaban' s favourable impact in reducing the severity of stroke events, potential differences in the severity of stroke events between the NOACs is much less certain. The assumption that ischemic stroke severity is dependent on treatment may have favoured apixaban. Had we assumed the same distribution of mRS in patients experiencing stroke events, regardless of treatment, the ICERs of apixaban versus the other NOACs would increase however remain below the £20,000 per QALY threshold.

As in previously published models,18'19'21 our study included recurrent stroke events. However, in contrast to these earlier studies' our analysis assumed that NOACs would offer no additional prophylaxis for secondary stroke prevention because published trials of these treatments did not assess their efficacy in preventing recurrence of stroke. Consequently, recurrent stroke rates used in this study were assumed to be the same for all treatments.34 Had we assumed that the efficacy of NOACs for secondary prevention is equal to the efficacy observed in primary prevention similarly to earlier cost-effectiveness studies, the results are likely to have been favorable for apixaban, given the lower number of strokes predicted in our model using primary prevention efficacy from the trials.

Various limitations also apply to our analysis. Although the utility estimates used in our study for all events were obtained from the same source (a standard EuroQol 5-Dimension catalogue) and validated by the 2 cardiologists who were part of the research team (Drs. Lip and Dorian), allowing consistent estimates between the various health states represented in the model, several utilities may be outdated with current practice. For example, the evolution of MI would suggest a higher baseline utility for these patients than those with a mild stroke; however, setting the utility of patients with MI to equal those with AF (ie, assuming no utility decrement associated with MI) or setting the HR of MI for all NOACs versus apixaban to 1 did not alter the model

conclusions (Table VI). Similarly, doubling the utility decrement for ICH had a negligible impact on results.

Importantly, event rates for each treatment were derived from clinical trial settings, and, consequently, the efficacy and safety observed may not reflect real-world outcomes. In the absence of head-to-head trials, an indirect treatment comparison based on the published NOAC studies was used to estimate the clinical event rates for other NOACS. Although several earlier indirect treatment comparisons have been conducted,51-54 we performed further analysis to conform to model definitions and requirements. However, similar to previously conducted indirect treatment comparisons, this analysis did not control for the differences in patient baseline characteristics, CHADS2 (congestive heart failure, hypertension, age >75 years, diabetes mellitus, prior stroke or transient ischemic attack, or thromboembolism) risk profile, or time in therapeutic range. Also, the comparison did not correct for key differences between the designs of the trials, which was open-label for RE-LY6 and double-blind for ROCKET-AF7 and ARISTOTLE.9 This presents a fundamental challenge in view of the heterogeneity of patient populations studied in each study (eg, ROCKET-AF studied a higher-risk population,7 and there was an open-label design in RE-LY6 vs double-blinded comparisons in ROCKET-AF7 and ARISTOTLE9). However, these variations seem more likely to have favored the other drugs rather than apixaban.

Lack of adjustment of baseline population characteristics can be considered conservative favoring rivar-oxaban. In an indirect comparison assessing treatment effects among patients with a CHADS2 score >3, apixaban was found to reduce the risk of the primary efficacy outcome of stroke or systemic embolism by 23% compared with rivaroxaban (HR = 0.77 [0.56, 1.06, 95% CI]).54 However, when including all trial patients,52 a smaller reduction of these risks compared with rivaroxaban has been shown (ie, HR of apixaban versus rivaroxaban = 0.90 [0.71-1.32]). Similarly, differences in trial design, in which ARISTOTLE was double-blinded and RE-LY was an open-label study, may also be conservative favoring dabigatran, if suggestions regarding the bias and overestimation of treatment effects of open-label trials are valid.55-58

Although we acknowledge the limitations associated with the methods of adjustment used in the indirect comparisons, we consider that these analyses

provide a reasonable comparison of treatment effects and are of appropriate use in our model in the absence of head-to-head trials. Furthermore, results generated from this analysis are highly consistent with earlier indirect comparisons,51-54 with minor deviations attributed to use of odds ratios rather than HRs and use of updated RE-LY6 data.

Overall, our predictions of the cost-effectiveness of apixaban could be considered cautious, given that the assumptions made in our study around input parameters and the differences between our model and previous models were likely to favor the comparator drugs.

CONCLUSIONS

The comprehensive assessment of the long-term efficacy, safety, and tolerability profile of apixaban in this study, generated through means of an economic model, predicted that the drug would provide an attractive alternative to other NOACs in the prevention of thromboembolic events in patients with AF. Specifically, it could offer favorable health benefits for a marginal increase in costs. In an economic environment of constrained health care resources, we believe that the findings of this study may help UK payers in making informed decisions that are in the best interests of patients who have NVAF.

ACKNOWLEDGEMENTS

The authors thank Jack Mardekian, Pfizer, for providing secondary analyses of the ARISTOTLE and AVERROES trial data. The authors also thank Pamela Pei, Dale Rublee (Pfizer), and Gethin Griffith (Pfizer) for their support in the model design and identification of model inputs.

CONFLICTS OF INTEREST

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

Dr. Lip has has served as a consultant for Bayer, Astellas, Merck, Sanofi, BMS/Pfizer, Daiichi-Sankyo, Biotronik, Portola, and Boehringer Ingelheim and has been on the speakers' bureau for Bayer, BMS/Pfizer, Boehringer Ingelheim, and Sanofi Aventis. Dr. Phatak is an employee of Bristol-Myers Squibb with ownership of stocks in Bristol-Myers Squibb. Dr. Iloeje was an employee of Bristol-Myers Squibb with ownership of stocks in Bristol-Myers Squibb at the time the study was conducted. Drs. Kuznik and Liu are full-time employees of Pfizer Inc with ownership of stocks in Pfizer Inc. Dr.

Kongnakorn, Ms. Lanitis and Mr. Hernandez are employees of Evidera who were paid consultants to Bristol-Myers Squibb and Pfizer in connection with conducting this study. Dr. Dorian has received consulting fees and research support from Bristol-Myers Squibb, Pfizer, Bayer, and Boehringer Ingelheim, and he served on the steering committee of the ARISTOTLE trial. The authors have indicated that they have no other conflicts of interest regarding the content of this article.

SUPPLEMENTAL MATERIAL

Supplemental materials accompanying this article can be found in the online version at http://dx.doi.org/10. 1016/j.clinthera .2013.12.011

REFERENCES

1. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham Study. Stroke. 1991;22:983-988.

2. Riva N, Lip GY. A new era for anticoagulation in atrial fibrillation. Which anticoagulant should we choose for long term prevention ofthromboembolic complications in patients with atrial fibrillation? Pol Arch Med Wewn. 2012;122:45-53.

3. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antith-rombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation. Ann Intern Med. 2007; 14б:857-8б7.

4. Ogilvie IM, Newton N, Welner SA, et al. Underuse of oral anticoagulants in atrial fibrillation: a systematic review. Am J Med. 2010;123:638-645.e4.

5. Prasad V, Kaplan RM, Passman RS. New frontiers for stroke prevention in atrial fibrillation. Cerebrovasc Dis. 2012;33:199-208.

6. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med. 2009;361:1139-1151.

7. Patel MR, Mahaffey KW, Garg J, et al. ROCKET AF Investigators. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med. 2011;365:883-891.

8. European Medicines Agency (EMA). Eliquis: EPAR-Product information - EMEA/H/C/002148 -PSUV/0012. 2013. http://www.ema.europa.eu/ema/. Accessed August 6, 2013.

9. Granger CB, Alexander JH, McMurrayJJ, et al, Aristotle Committees and Investigators. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med. 2011;365:981-992.

10. Camm AJ, Lip GY, De Caterina R, et al. 2012 Focused dpdate of the ESC guidelines for the management of atrial fibrillation: an update of the 2010 ESC guidelines for the

management of atrial fibrillation. Developed with the special contribution of the European Heart Rhythm Association. Eur Heart J. 2012;33:2719-2747.

11. De Caterina R, Husted S, Wallentin L, et al, Coordinating Committee. New oral anticoagulants in atrial fibrillation and acute coronary syndromes: ESC Working Group on Thrombosis-Task Force on Anticoagulants in Heart Disease Position Paper. J Am Coll Cardiol. 2012;59:1413-1425.

12. Briggs A, Sculpher M. An introduction to Markov modelling for economic evaluation. Pharmacoeconomics. 1998;13:397-409.

13. Freeman JV, Zhu RP, Owens DK, et al. Cost-effectiveness of dabigatran compared with warfarin for stroke prevention in atrial fibrillation. Ann Intern Med. 2011;154:1-11.

14. Kamel H, Easton JD, Johnston SC, Kim AS. Cost-effectiveness of apixaban vs warfarin for secondary stroke prevention in atrial fibrillation. Neurology. 2012;79:1428-1434.

15. Kamel H, Johnston SC, Easton JD, Kim AS. Cost-effectiveness of dabigatran compared with warfarin for stroke prevention in patients with atrial fibrillation and prior stroke or transient ischemic attack. Stroke. 2012;43:881-883.

16. Lee S, Anglade MW, Pham D, et al. Cost-effectiveness of rivaroxaban compared to warfarin for stroke prevention in atrial fibrillation. Am J Cardiol. 2012;110:845-851.

17. Lee S, Mullin R, Blazawski J, Coleman CI. Cost-effectiveness of apixaban compared with warfarin for stroke prevention in atrial fibrillation. PLoS One. 2012;7: e47473.

18. Shah SV, Gage BF. Cost-effectiveness of dabigatran for stroke prophylaxis in atrial fibrillation. Circulation. 2011;123:2562-2570.

19. Sorensen SV, Kansal AR, Connolly S, et al. Cost-effectiveness of dabigatran etexilate for the prevention of stroke and systemic embolism in atrial fibrillation: a Canadian payer perspective. Thromb Haemost. 2011; 105:908-919.

20. Coyle D, Coyle K, Cameron C, et al. Cost-effectiveness of new oral anticoagulants compared with warfarin in preventing stroke and other cardiovascular events in patients with atrial fibrillation. Value Health. 2013; 16:498-506.

21. Harrington AR, Armstrong EP, Nolan PEJr, Malone DC. Cost-effectiveness of apixaban, dabigatran, rivaroxaban, and warfarin for stroke prevention in atrial fibrillation. Stroke. 2013;44:1676-1681.

22. Pink J, Pirmohamed M, Hughes DA. Comparative effectiveness of dabigatran, rivaroxaban, apixaban, and warfarin in the management of patients with nonvalvular atrial fibrillation. Clin Pharmacol Ther. 2013;94:269-276.

23. Siebert U, Alagoz O, Bayoumi AM, et al, Modeling Good Research Practices Task Force. State-transition modeling:

a report of the ISPOR-SMDM Modeling Good Research Practices Task Force—3. Value Health. 2012;15:812-820.

24. BamfordJ, Sandercock P, Dennis M, et al. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project—1981-86. 2. Incidence, case fatality rates and overall outcome at one year of cerebral infarction, primary intracerebral and subarachnoid haemorrhage. J Neurol Neurosurg Psychiatry. 1990;53: 16-22.

25. European Medicines Agency (EMA). Pradaxa: EPAR-Product information -EMEA/H/C/000829 -PSU/0034. 2013. http://www.ema.europa.eu/ ema/. Accessed August 6, 2013.

26. European Medicines Agency (EMA). Xarelto: EPAR-Product information -EMEA/H/C/000944 -II/0023. 2013. http://www.ema.europa.eu/ema/. Accessed August 6, 2013.

27. Bucher HC, Guyatt GH, Griffith LE, Walter SD. The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials. J Clin Epidemiol. 1997; 50:683-691.

28. Claassen DO, Kazemi N, ZubkovAY, et al. Restarting anticoagulation therapy after warfarin-associated in-tracerebral hemorrhage. Arch Neurol. 2008;65:1313-1318.

29. Sorensen SV, Dewilde S, Singer DE, et al. Cost-effectiveness of warfarin: trial versus "real-world" stroke prevention in atrial fibrillation. Am Heart J. 2009;157:1064-1073.

30. Connolly SJ, Eikelboom J, Joyner C, et al, AVERROES Steering Committee and Investigators. Apixaban in patients with atrial fibrillation. N Engl J Med. 2011;364:806-817.

31. Pisters R, Lane DA, Marin F, et al. Stroke and thromboembolism in atrial fibrillation. Circ J. 2012;76: 2289-2304.

32. Ariesen MJ, Claus SP, Rinkel GJ, Algra A. Risk factors for intracere-bral hemorrhage in the general

population: a systematic review. Stroke. 2003;34:2060-2065.

33. Executive Summary of the Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). JAMA. 2001;285:2486-2497.

34. Easton JD, Lopes RD, Bahit MC, et al, ARISTOTLE Committees and Investigators. Apixaban compared with warfarin in patients with atrial fibrillation and previous stroke or transient ischaemic attack: a subgroup analysis of the Aristotle trial. Lancet Neurol. 2012;11:503-511.

35. Scarborough P, Bhatnagar P, Wick-ramasinghe K, et al. Coronary Heart Disease Statistics Edition. 2010. London, UK: British Heart Foundation.

36. Human Mortality Database. 2009 UK life tables. http://www.mortality. org/cgi-bin/hmd/country.php?cntr= GBR&level=2. Accessed July 29, 2013.

37. Sullivan PW, Slejko JF, Sculpher MF, Ghushchyan V. Catalogue of EQ-5D scores for the United Kingdom. Med Decis Making. 2011;31:800-804.

38. Gage BF, Cardinalli AB, Owens DK. The effect of stroke and stroke prophylaxis with aspirin or warfarin on quality of life. Arch Intern Med. 1996;156:1829-1836.

39. Friberg L, HammarN, Pettersson H, et al. Increased mortality in 28. Claassen DO, Kazemi N, Zubkov AY, et al. Restarting anticoagulation therapy after warfarin-associated in-tracerebral hemorrhage. Arch Neurol. 2008;65:1313-1318.

40. Friberg L, HammarN, Pettersson H, et al. Increased mortality in paroxysmal atrial fibrillation: report from the Stockholm Cohort-Study of At-rial Fibrillation (SCAF). Eur Heart J. 2007;28:2346-2353.

41. Huybrechts KF, Caro JJ, Xenakis J. The prognostic value of the modified rankin scale score for long-term survival after first-ever stroke. Cere-brovasc Dis. 2008;26:381-387.

42. Henriksson KM, Farahmand B, Johansson S, et al. Survival after stroke-the impact of CHADS2 score and atrial fibrillation. Int J Cardiol. 2010;141:18-23.

43. Bronnum-Hansen H, Davidsen M, Thorvaldsen P. Long-term survival and causes of death after stroke. Stroke. J Cerebral Circ. 2001;32:2131-2136.

44. Bronnum-Hansen H, Jorgensen T, Davidsen M, etal. Survival and cause of death after myocardial infarction: the Danish MONICA study. J Clin Epidemiol. 2001;54:1244-1250.

45. Stewart S, Hart CL, Hole DJ, et al. A population-based study of the long-term risks associated with atrial fibrillation: 20-year follow-up of the Renfrew/Paisley study. Am J Med. 2002;113:359-364.

46. National Institute for Health and Clinical Excellence (January 2009) The guidelines manual.London: National Institute for Health and Clinical Excellence. www.nice.org.uk. Accessed January 21, 2014.

47. National Health Service (NHS). Electronic Drug Tariff. NHS Business Services Authority, NHS Prescription Services; 2011 http://www.ppa.org. uk/ppa/edt_intro.htm. Accessed July 29, 2013.

48. Department of Health, National Schedule of Reference Costs Year : '2009-10' - NHS Trusts and PCTs combined. http://www.dh.gov.uk/ en/Publicationsandstatistics/Publica tions/PublicationsPolicyAndGuidance/ DH_131140. Accessed February 11, 2013.

49. Youman P, Wilson K, Harraf F, et al. The economic burden of stroke in the United Kingdom. PharmacoEco-nomics. 2003;21(Suppl 1):43-50.

50. Beswick AD, Rees K, Griebsch I, et al. Provision, uptake and cost of cardiac rehabilitation programmes: improving services to under-represented groups. Health technology assessment. 2004:1-152. 8:iii -iv, ix-x.

51. Boggon R, Lip GY, Gallagher AM, van Staa TP. Resource utilization and outcomes in patients with atrial

fibrillation: a case control study. Appl Health Econ Health Policy. 2012; 10:249-259.

52. Baker WL, Phung OJ. Systematic review and adjusted indirect comparison meta-analysis of oral anticoagulants in atrial fibrillation. Circ Cardiovasc Qual Outcomes. 2012;5:711-719.

53. Lip GY, Larsen TB, Skjoth F, Rasmus-sen LH. Indirect comparisons of new oral anticoagulant drugs for efficacy and safety when used for stroke prevention in atrial fibrillation. J Am Coll Cardiol. 2012;60:738-746.

54. Mantha S, Ansell J. An indirect comparison of dabigatran, rivarox-aban and apixaban for atrial fibrillation. Thromb Haemost. 2012;108: 476-484.

55. Schneeweiss S, GagneJJ, Patrick AR, et al. Comparative efficacy and safety of new oral anticoagulants in patients with atrial fibrillation. Circ Cardiovasc Qual Outcomes. 2012;5: 480-486.

56. Psaty BM, Prentice RL. Minimizing bias in randomized trials: the importance of blinding. JAMA. 2010;304:793-794.

57. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. JAMA. 1995;273:408-412.

58. Welton NJ, Ades AE, Carlin JB, et al. Models for potentially biased evidence in meta-analysis using empirically based priors. J Roy Statistic Soc Series A. 2009;172:119-136.

Address correspondence to: Gregory Y.H. Lip, University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Birmingham, B18 7QH, England, United Kingdom. E-mail: g.y.h.lip@bham.ac.uk

SUPPLEMENTAL APPENDIX A

Cost-Effectiveness of Apixaban Versus Other New Oral Anticoagulants for Stroke Prevention in Patients with Atrial Fibrillation: Indirect Comparison Analysis (Using Event-Rate Data) to Assess Relative Efficacy and Safety of the New Oral Anticoagulants

OBJECTIVE

The following report details results of the indirect comparison analysis conducted to examine the relative efficacy and safety of the new oral anticoagulants (apixaban, dabigatran 110 mg, dabigatran 150 mg, and rivaroxaban). This report is primarily focused on reporting the hazard ratios (HRs) used as model inputs. Analyses were conducted by using event-rate data reported in the primary publications and related reports.

METHODS

The analysis is restricted to randomized controlled trials of pharmacologic prophylaxis of stroke prevention in patients with atrial fibrillation.

Population

The population covered in the analysis included: (1) patients with a risk of stroke and diagnosed with mild-to-moderate nonvalvular atrial fibrillation; and (2) patients who were warfarin eligible.

Interventions for the Analysis

The following pharmacologic methods of prophylaxis were included: (1) apixaban; (2) rivaroxaban; and (3) dabigatran etexilate.

Outcomes Outcomes Reported

The current analysis analyzed event-rate data for the following outcomes: ischemic or unspecified stroke; myocardial infarction; intracranial hemorrhage; other major bleeds; clinically relevant nonmajor (CRNM) bleed; and total discontinuations.

Studies Included in the Analyses

Three randomized controlled trials were included in the analyses:

1. ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fi-brillation)1: apixaban 5 mg BID versus warfarin; international normalized ratio (INR), 2.0-3.0;

2. ROCKET-AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) 2: rivaroxaban, 20 mg once daily versus warfarin; INR, 2.0-3.0; and

3. RE-LY (Randomized Evaluation of Long-Term Anticoagulation Therapy)3: dabigatran, 110 mg BID versus dabigatran 150 mg BID versus warfarin; INR, 2.0-3.0.

Statistical Analysis

Direct and Indirect Meta-Analysis

Direct meta-analysis was conducted in Microsoft Excel. The HRs and their SEs were calculated by using standard formulae:

eventratea eventsa/PtTimeatRiska

eventrateb eventsb / PtTimeatRiskb

SE[ln(HR)] — \J 1/eventsa + 1/eventsb

Indirect comparisons between apixaban and other treatments of interest via a common comparator were made by using the method of Bucher et al4 and the HRs produced from the direct meta-analysis. The (indirect) HR between apixaban and the treatment of interest is given by:

log(HRAVB) — log(HRAvc)~ log(HRBva)

with SE given by:

SE[ log(HRAVB)] SE[ log(HRAVC)]2 + SE[ log(HRBVC)]2

For some studies, rates were estimated where the rate events were not reported in the publication. The rates were estimated from the number of patients experiencing an outcome in cases in which the patient numbers were reported but the rate of first events was not reported. We calculated the probability of experiencing an outcome at the median follow-up point by dividing the number of first events by the number of patients randomized and converted this probability into an annual rate as: rate = -ln(1-probability)/median follow-up. This approximation accurately predicted the event rate for studies in which both the rate and number of patients with events were reported; this value is particularly approximate for studies in which patients were followed up for a fixed period of time.

The analysis of event-rate outcomes was conducted on an intent-to-treat (ITT) basis for efficacy outcomes and modified ITT (mITT) for safety outcomes. The analysis was conducted for efficacy end points in ITT

populations (or number randomized and received treatment, if ITT figures are not reported). For adverse events, the safety population used in the analysis was the mITT population (number randomized and on treatment, with ITT data used if mITT data not reported). Similar analyses were performed for safety outcomes: ITT as the first analysis and per protocol as the second analysis.

Connolly et al5 published an update to the original RE-LY publication that reported several additional primary efficacy and safety outcome events noted during routine clinical site closure visits after the database was locked. "Data from the update of the original RE-LY publication i.e. data reported in the updated 2010 publication was used in the primary analysis."5 All the newly identified events were adjudicated in a blinded fashion and in accordance with the study protocol.

All efficacy and safety analyses were reported on the ITT population for RE-LY. In the present analysis,

efficacy results for ARISTOTLE are based on the ITT population. For the ROCKET-AF study, apart from primary efficacy outcome (stroke or systemic embolism), all secondary efficacy outcomes were reported on an on-treatment basis; although 14,264 patients were randomized in the ROCKET-AF study, 93 patients (rivaroxaban, n = 50; warfarin, n = 43) were excluded from all efficacy analyses before unblinding due to violations in Good Clinical Practice guidelines at 1 site. Therefore, the denominators used in the efficacy analysis reported here are as follows: rivaroxaban, n = 7081; warfarin, n = 7090.

The safety analyses for ARISTOTLE were based on a mITT population and included all patients who received at least 1 dose of the study drug and included all events from the time the first dose of a study drug was received until 2 days after the last dose was received. To align the reporting of safety results from ROCKET-AF with those from ARISTOTLE, the safety, as-treated population was used for the denominator (all patients who received at

Supplemental Table I. Data used in analysis of ischemic + unspecified stroke outcome (calculated for ROCKET-AF [Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation] publication) in warfarin-eligible patients.

Study Treatment Arm No. of Patients Event Rate (%/Year)

Dabigatran 110 mg BID 6015* 1.34*

Warfarin; INR, 2.0-3.0 6022* 1.21+

Dabigatran 150 mg BID 6076* 0.92*

ROCKET-AF

Rivaroxaban 20 mg once daily 7081* 1.15§

Warfarin, INR 2.0-3.0 7090* 1.27§

ARISTOTLE

Apixaban 5 mg BID 9120* 0.97+

Warfarin; INR, 2.0-3.0 9081* 1.05+

RE-LY = Randomized Evaluation of Long-Term Anticoagulation Therapy; INR = international normalized ratio; ARISTOTLE = Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation.

*Intention-to-treat population (for ARISTOTLE, defined as all patients who underwent randomization and included all

events from the time of randomization until the cutoff date for efficacy outcomes). * Events occurring through the cutoff date (includes follow-up period).

^Reported in the publication as the intention-to-treat population (number of patients randomized: rivaroxaban, n = 7131; warfarin, n = 7133). Patients excluded due to violations in Good Clinical Practice guidelines at 1 site that made the data unreliable. ^Calculated event rates.

least 1 dose of study drug and were followed up for events regardless of adherence to the protocol, while they were receiving the assigned study drug or within 2 days after discontinuation).

Calculated events from dichotomous data were used in cases in which event rates were not reported in publications, and the calculated event-rate data were used in the analyses for up to 2 decimal places wherever possible.

RESULTS

Indirect Comparison Analysis (Warfarin-Eligible Patients)

The results reported in this section are based on the groups of warfarin-eligible patients for indirect comparison analysis.

Ischemic + Unspecified Stroke (Calculated for ROCKET Publication)

Data for RE-LY are that reported in the updated 2010 publication. Calculated HRs (95% [CIs]) for the event ischemic or unspecified stroke are shown in Supplemental Table I.

Supplemental Table II presents a comparison of efficacy between treatments for warfarin-eligible patients: ischemic + unspecified stroke (calculated for ROCKET Publication).

Myocardial Infarction

Data for RE-LY are that reported in the updated 2010 publication. Calculated HRs (95% CIs) for the event ischemic or unspecified stroke are shown in Supplemental Table III. Supplemental Table IV

presets a comparison of efficacy between treatments for warfarin-eligible patients: myocardial infarction.

Intracranial Hemorrhage

Data for RE-LY are that reported in the updated 2010 publication. Calculated HRs (95% CIs) for the event ischemic or unspecified stroke are shown in Supplemental Table V. Supplemental Table VI presents a comparison of efficacy between treatments for warfarin-eligible patients: intracranial hemorrhage.

Other Major Bleeds (Major Bleeding Minus Intracranial Hemorrhage: On-Treatment Analysis)

Data for RE-LY are that reported in the updated 2010 publication. Calculated HRs (95% CIs) for the event ischemic or unspecified stroke are shown in Supplemental Table VII. Supplemental Table VIII

presents a comparison of efficacy between treatments for warfarin-eligible patients: other major bleeds (calculated event rates).

Supplemental Table II. Comparison of efficacy between treatments for warfarin-eligible patients: ischemic + unspecified stroke (calculated for ROCKET-AF [Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation] publication).

Indirect Comparison:

HR Versus Warfarin HR Versus Warfarin Comparator Versus

Treatment (95% CI) Comparator (95% CI) Treatment HR (95% CI)

Apixaban 0.923 (0.746-1.143) Rivaroxaban 0.905 (0.729-1.124) 0.980 (0.723-1.328)

Dabigatran 110 mg BD 1.107 (0.883-1.388) 1.198 (0.878-1.635)

Dabigatran 150 mg BD 0.760 (0.593-0.974) 0.823 (0.593-1.141)

Rivaroxaban 0.905 (0.729-1.124) Dabigatran 110 mg BD 1.107 (0.883-1.388) 1.222 (0.894-1.672)

Dabigatran 150 mg BD 0.760 (0.593-0.974) 0.839 (0.604-1.167)

Dabigatran 1.107 (0.883-1.388) Dabigatran 150 mg BID 0.760 (0.593-0.974) 0.686 (0.490-0.960) 110 mg BID

HR = hazard ratio.

*Significant differences are shown in bold.

Supplemental Table I I. Data used in analysis of myocardial infarction outcome in warfarin-eligible patients.

Study Treatment Arm No. of Patients Event Rate (%/Year)

Dabigatran 110 mg BID 6015* 0.82'

Warfarin; INR, 2.0-3.0 6022* 0.64'

Dabigatran 150 mg BID 6076* 0.81'

ROCKET-AF

Rivaroxaban 20 mg once daily 7081' 0.91§

Warfarin; INR, 2.0-3.0 7090' 1.12§

ARISTOTLE

Apixaban 5 mg BID 9120* 0.53'

Warfarin; INR, 2.0-3.0 9081* 0.61'

RE-LY = Randomized Evaluation of Long-Term Anticoagulation Therapy; INR = international normalized ratio; ROCKET-AF = Rivaroxaban Once Daily Oral Direct FactorXa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation; ARISTOTLE = Apixaban for Reduction in Stroke and OtherThromboembolic Events in Atrial Fibrillation.

*Intention-to-treat population (for ARISTOTLE, defined as all patients who underwent randomization and included all

events from the time of randomization until the cutoff date for efficacy outcomes). 'Events occurring through the cutoff date (includes follow-up period).

'Reported in the publication as the intention-to-treat population (number of patients randomized: rivaroxaban, n = 7131; warfarin, n = 7133). Patients excluded due to violations in Good Clinical Practice guidelines at 1 site that made the data unreliable.

§Safety-on-treatment population, including patients who received >1 dose of study drug and were followed up for events, regardless of adherence to the protocol, while they were receiving the assigned study drug or within 2 days after discontinuation.

Supplemental Table IV. Comparison infarction. of efficacy between treatments for warfarin -eligible patients: myocardial

Treatment HR Versus Warfarin (95% CI) Comparator HR Versus Warfarin (95% CI) Indirect Comparison: Comparator Versus Treatment HR (95% CI)

Apixaban Rivaroxaban Dabigatran 110 mg BID 0.868 (0.654-1.153) 0.812 (0.625-1.055) 1.281 (0.925-1.773) Rivaroxaban Dabigatran 110 mg BiD Dabigatran 150 mg BiD Dabigatran 110 mg BiD Dabigatran 150 mg BiD Dabigatran 150 mg BiD 0.812 (0.625-1.055) 1.281 (0.925-1.773) 1.265 (0.916-1.747) 1.281 (0.925-1.773) 1.265 (0.916-1.747) 1.265 (0.916-1.747) 0.935 (0.635-1.375) 1.474 (0.958-2.269) 1.456 (0.948-2.238) 1.576 (1.038-2.393) 1.557 (1.028-2.360) 0.987 (0.624-1.561)

HR = hazard ratio. *Significant differences are shown in bold

Supplemental Table V. Data used in analysis of intracranial hemorrhage in warfari n-eligible patients.

Study Treatment Arm No. of Patients Event Rate (%/Year)

Dabigatran 110 mg BID 6015* 0.23*

Warfarin; INR, 2.0-3.0 6022* 0.76*

Dabigatran 150 mg BID 6076* 0.32*

ROCKET-AF

Rivaroxaban 20 mg OD 7061 * 0.5§

Warfarin; INR, 2.0-3.0 7082* 0.7§

ARISTOTLE

Apixaban 5 mg BID 9088" 0.33§

Warfarin; INR, 2.0-3.0 9052" 0.8§

RE-LY = Randomized Evaluation of Long-Term Anticoagulation Therapy; INR = international normalized ratio; ROCKET-AF =

Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and

Embolism Trial in Atrial Fibrillation; ARISTOTLE = Apixaban for Reduction in Stroke and Othe r Thromboembolic Events in

Atrial Fibrillation.

*Intention-to-treat population.

'Events occurring through the cutoff date (includes follow-up period).

^Patients who received at least 1 dose of study drug and were followed up for events regardless of adherence to protocol,

while they were receiving the drug or within 2 day s after discontinuation.

§Events occurring during the treatment period or within 2 days after discontinuation (excluding follow-up).

Patients who received at least 1 dose of study drug and events that occurred from the ti me the patients received the first dose

of the study drug through 2 days after they received the last dose.

Supplemental Table VI. Comparison hemorrhage. of efficacy between treatments for warfarin * -eligible patients: intracranial

Treatment HR Versus Warfarin (95% CI) Comparator HR Versus Warfarin (95% CI) Indirect Comparison: Comparator Versus Treatment HR (95% CI)

Apixaban Rivaroxaban Dabigatran 110 mg BID 0.412 (0.298-0.570) 0.714 (0.508-1.003) 0.302 (0.196-0.465) Rivaroxaban Dabigatran 110 mg BID Dabigatran 150 mg BID Dabigatran 110 mg BID Dabigatran 150 mg BID Dabigatran 150 mg BID 0.714 (0.508-1.003) 0.302 (0.196-0.465) 0.421 (0.288-0.615) 0.302 (0.196-0.465) 0.421 (0.288-0.615) 0.421 (0.288-0.615) 1.731 (1.082-2.770) 0.733 (0.428-1.257) 1.020 (0.619-1.681) 0.423 (0.244-0.733) 0.589 (0.354-0.980) 1.391 (0.784-2.468)

HR = hazard ratio. *Significant differences are shown in bole

Supplemental Table VII. Data used in analysis of other major bleeds (calculated event rates) in warfarin-

eligible patients.

Study Treatment Arm No. of Patients Event Rate (%/Year)

Dabigatran 110 mg BID 6015* 2.69'

Warfarin; INR, 2.0-3.0 6022* 2.83'

Dabigatran 150 mg BID 6076* 3.06'

ROCKET-AF

Rivaroxaban 20 mg once daily 7061' 3.07§

Warfarin; INR, 2.0-3.0 7082' 2.71 §

ARISTOTLE

Apixaban 5 mg BID 908811 1.79

Warfarin; INR, 2.0-3.0 905211 2.27

RE-LY = Randomized Evaluation of Long-Term Anticoagulation Therapy; INR = international normalized ratio; ROCKET-AF

= Rivaroxaban Once Daily Oral Direct FactorXa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke

and Embolism Trial n Atrial Fibrillation; ARISTOTLE = Apixaban for Reduction in Stroke and OtherThromboembolic Events

in Atrial Fibrillation

*Intention-to-treat numbers, randomized.

'Calculated event rates.

'Patients who received at least 1 dose of study drug and events that occurred from the time the patients received the first

dose of the study drug through 2 days after they received the last dose.

^Calculated event rates with median on-treatment follow-up.

Patients who received at least 1 dose of study drug and were followed up for events regardless of adherence to protocol,

while they were receiving the drug or within 2 days after discontinuation.

CRNM Bleed (ARISTOTLE Clinical Study Report Data) (Including Minor Bleed as Reported in RE-LY for Dabigatran)

Calculated HRs (95% CIs) for the event ischemic or unspecified stroke are shown in Supplemental Table IX. Supplemental Table X presents a comparison of efficacy between treatments for

warfarin-eligible patients: CRNM bleed (minor bleed data used for RE-LY).

Total Discontinuations (Calculated Event Rates)

Calculated HRs (95% CIs) for the event ischemic or unspecified stroke are shown in Supplemental Table XI. Supplemental Table XII presents a comparison of

Supplemental Table VIII. Comparison of efficacy between treatments for warfarin major bleeds (calculated event rates). -eligible patients: other

Treatment HR Versus Warfarin (95% CI) Comparator HR Versus Warfarin (95% CI) Indirect Comparison: Comparator Versus Treatment HR (95% CI)

Apixaban Rivaroxaban Dabigatran 110 mg BID 0.788 (0.672-0.924) 1.132 (0.970-1.322) 0.950 (0.814-1.109) Rivaroxaban Dabigatran 110 mg BID Dabigatran 150 mg BID Dabigatran 110 mg BID Dabigatran 150 mg BID Dabigatran 150 mg BID 1.132 (0.970-1.322) 0.950 (0.814-1.109) 1.081 (0.931-1.255) 0.950 (0.814-1.109) 1.081 (0.931-1.255) 1.081 (0.931-1.255) 1.436 (1.150-1.793) 1.205 (0.965-1.504) 1.371 (1.102-1.705) 0.839 ( 0.674-1.044) 0.954 (0.769-1.183) 1.137 (0.917-1.409)

HR = hazard ratio. *Significant differences are shown in bold.

Supplemental Table IX. Data used in analysis of clinically relevant nonmajor bleed (minor bleed data used for RE-LY [Randomized Evaluation of Long-Term Anticoagulation Therapy]) in warfarin-eligible patients.

Study Treatment Arm No. of Patients Event Rate (%/Year)

RE-LY minor bleed*

Dabigatran 110 mg BID 6015* 13.16*

Warfarin; INR, 2.0-3.0 6022* 16.37*

Dabigatran 150 mg BID 6076* 14.84*

ROCKET-AF

Rivaroxaban 20 mg once daily 7061 § 11.8"

Warfarin; INR, 2.0-3.0 7082§ 11.4"

ARISTOTLE

Apixaban 5 mg BID 9088^ 2.083"

Warfarin; INR, 2.0-3.0 9052^ 2.995"

INR = international normalized ratio; ROCKET-AF = Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation; ARISTOTLE = Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation.

*Defined as all bleeds that did not meet the definition of major bleed or clinically relevant nonmajor bleed. *Intention-to-treat population.

^Events occurring through the cutoff date (includes follow-up period).

§Patients who received at least 1 dose of study drug and were followed up for events regardless of adherence to protocol, while they were receiving the drug or within 2 days after discontinuation.

Events during the treatment period or within 2 days after discontinuation (as reported in the clinical study report). ^Patients who received at least 1 dose of study drug and events that occurred from the time the patients received the first dose of the study drug through 2 days after they received the last dose.

Supplemental Table X. Comparison of efficacy between treatments for warfarin-eligible patients: clinically relevant nonmajor bleed (minor bleed data used for RE-LY LY [Randomized Evaluation of Long-Term Anticoagulation Therapy]).*

Treatment HR Versus Warfarin (95% CI) Comparator HR Versus Warfarin (95% CI)

Apixaban Rivaroxaban Dabigatran 110 mg BD 0.695 (0.602-0.803) Rivaroxaban Dabigatran 110 mg BID Dabigatran 150 mg BID 1.035 (0.954-1.122) Dabigatran 110 mg BID Dabigatran 150 mg BID 0.803 (0.752-0.859) Dabigatran 150 mg BID 1.035 (0.954-1.122) 0.803 (0.752-0.859) 0.906 (0.850-0.966) 0.803 (0.752-0.859) 0.906 (0.850-0.966) 0.906 (0.850-0.966)

HR = hazard ratio. *Significant differences are shown in bold.

Table XI. Data used in analysis of total discontinuations outcome (calculated event rates) in warfarin-eligible

patients.

Study Treatment Arm No. of Patients Event Rate (%/Year)

RE-LY*

Dabigatran 110 mg BID 6015' 10.72'§

Warfarin; INR, 2.0-3.0 6022' 8.11 '§

Dabigatran 150 mg BID 6076' 11.11 '§

ROCKET-AF

Rivaroxaban 20 mg once daily 7131^ 14.02'#

Warfarin, INR 2.0-3.0 7133^ 13.01 '#

ARISTOTLE**

Apixaban 5 mg BID 9120' 16.23'''

Warfarin; INR, 2.0-3.0 9081 ' 17.83'''

RE-LY = Randomized Evaluation of Long-Term Anticoagulation Therapy; INR = international normalized ratio; ROCKET-AF = Rivaroxaban Once Daily Oral Direct FactorXa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation; ARISTOTLE = Apixaban for Reduction in Stroke and OtherThromboembolic Events in Atrial Fibrillation. *Median follow-up is 2 years. 'Randomized numbers.

'Calculated event rates from the dichotomous data.

umber of patients who discontinued at 2 years. Median follow-up is 707 days or 1.93 years.

^Number of patients randomized: rivaroxaban, n = 7131; warfarin, n = 7133.

#Patients lost to follow-up, experiencing primary end point, death, Good clinical practice violating site/closed site patients, and those not receiving any study drug excluded. **Median follow-up is 1.8 years.

''Accounted for "end of treatment period" randomized patients (as reported in the clinical study report).

Supplemental Table XII. Comparison of efficacy between discontinuations (calculated event treatments for warfari rates).* n-eligible patients: total

Treatment HR Versus Warfarin (95% CI) Comparator HR Versus Warfarin (95% CI) Indirect Comparison: Comparator Versus Treatment HR (95% CI)

Apixaban Rivaroxaban Dabigatran 110 mg BID 0.910 (0.860-0.963) 1.077 (1.006-1.154) 1.321 (1.211-1.441) Rivaroxaban Dabigatran 110 mg BID Dabigatran 150 mg BID Dabigatran 110 mg BID Dabigatran 150 mg BID Dabigatran 150 mg BID 1.077 (1.006-1.154) 1.321 (1.211-1.441) 1.369 (1.256-1.493) 1.321 (1.211-1.441) 1.369 (1.256-1.493) 1.369 (1.256-1.493) 1.184 (1.083-1.294) 1.452 (1.309-1.611) 1.505 (1.357-1.668) 1.226 (1.097-1.369) 1.270 (1.138-1.418) 1.036 (0.917-1.171)

HR = hazard ratio. *Significant differences are shown in bold.

efficacy between treatments for warfarin-eligible patients: total discontinuations (calculated event rates).

REFERENCES

1. Granger CB, Alexander JH, McMurray JJ, et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med. 2011;365:981-992.

2. Patel MR, Mahaffey KW, GargJ, et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med. 2011;10:883-891.

3. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med. 2009;361:12:1139-1151.

4. Bucher HC, Guyatt GH, Griffith LE, Walter SD. The results of direct and indirect treatment comparisons in metaanalysis of randomized controlled trials. J Clin Epidemiol. 1997;50:683-691.

5. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Newly identified events in the RE-LY trial. N Engl J Med. 2010;363: 1875-1876.

SUPPLEMENTAL APPENDIX B

Cost-Effectiveness of Apixaban versus Other New Oral Anticoagulants for Stroke Prevention in Patients with Atrial Fibrillation: Secondary Analysis of ARISTOTLE (CV185-030) to Support Apixaban Cost Effectiveness Modelling for the Indication of Stroke Prevention in Atrial Fibrillation

STUDY DESIGN

Study design and data collection associated with ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) have been described previously.1 Seminal results, published by Granger et al,2 indicated that apixaban was superior to warfarin in terms of prevention of stroke and systemic embolism, reduction in major bleeding, and reduction in all-cause mortality.

METHODS

ARISTOTLE data were analyzed on a post hoc basis. Annualized event rates and related hazard ratios were computed for the relevant end points. The intention-to-treat (ITT) principle was used for the analysis of all end points except the safety end points (which were based on the modified ITT population). Stratified analyses were conducted across CHADS2 (congestive heart failure, hypertension, age >75 years, diabetes mellitus, prior stroke or transient ischemic attack, or thromboembolism) categories.3 All risks were assessed in absolute terms or relative risk was computed by using hazard ratios for the event associated with warfarin versus apixaban. Drug-specific (ie, separately for apixaban as well as for warfarin) analysis of the center's time in therapeutic range (cTTR) was conducted by estimating the relative hazard of experiencing an event in a cTTR quartile versus that observed in the reference cTTR quartile of 52.38% < cTTR <66.02%.4 This quartile was chosen as a reference quartile because it encompassed the median TTR of 66%.

Mortality risk associated with reasons other than stroke, systemic embolism, bleeding, and myocardial infarction was assessed by deleting mortality associated with the aforementioned causes from all-cause mortality. Absolute mortality risk as well as relative hazard of mortality with warfarin versus apixaban for causes other than stroke, bleeding, and myocardial infarction was assessed.

Data were pooled from the treatment arms of ARISTOTLE and AVERROES (Apixaban Versus Ace-tylsalicylic Acid to Prevent Strokes) to assess fatality rates associated with intracranial hemorrhages, excluding hemorrhagic stroke and other major bleeds excluding any type of intracranial hemorrhages.

All analyses were conducted by using SAS version 9.2 (SAS Institute, Inc, Cary, North Carolina). Because this was a secondary analysis of ARISTOTLE analyses sets performed in a post hoc manner, no statistical significance was assessed.

Risk of an Event by cTTR Quartiles Ischemic and Unspecified Stroke

cTTR quartiles (as described in the ARISTOTLE clinical study report Table 7.1.1.4) were used for the purpose of this analysis.4 Supplemental Table I describes the risk of ischemic or unspecified stroke according to cTTR quartiles associated with apixaban. Supplemental Table II describes the risk of ischemic or unspecified stroke according to cTTR quartiles associated with warfarin.

Intracranial Hemorrhages

Similar to ischemic or unspecified stroke, risk of intracranial hemorrhages was assessed across cTTR quartiles, as described earlier. 4 Supplemental Tables III and IV describe the event rate and relative hazard ratios (HRs) for intracranial hemorrhage associated with apixaban and warfarin, respectively.

Other Major Bleeds Excluding Intracranial Hemorrhages

Other major bleeds were defined as major bleeds excluding intracranial hemorrhages. Supplemental Tables V and VI describe the event rate and relative HRs for other major bleeds associated with apixaban and warfarin, respectively.

Clinically Relevant Nonmajor Bleeding

Event rates and associated HRs associated with clinically relevant nonmajor bleeding are described in Supplemental Tables VII and VIII for apixaban and warfarin, respectively.

Risk of Ischemic or Unspecified Stroke by CHADS2 Score

Risk of ischemic or unspecified stroke was assessed at prespecified CHADS2 categories of 0 to 1, 2, and

Supplemental Table I. Ischemic or unspecified stroke risk according to (cTTR) quartiles for apixaban. center's time in therapeutic range

Stroke No. Event Rate HR (95% CI)

cTTR <52.38% 23 1.091 0.922 (0.578-1.472)

52.38% < cTTR <66.02% 74 1.178 1.000 (Ref.)

66.02% < cTTR <76.51% 51 0.806 0.690 (0.483-0.986)

cTTR >76.51% 13 0.656 0.561 (0.311-1.011)

HR = hazard ratio.

Supplemental Table II. Ischemic or unspecified stroke risk according to (cTTR) quartiles for warfarin. center's time in therapeutic range

Stroke No. Event Rate HR (95% CI)

cTTR <52.38% 34 1.678 1.542 (1.022-2.328)

52.38% < cTTR <66.02% 68 1.078 1.000 (Ref.)

66.02% < cTTR <76.51% 58 0.912 0.836 (0.588-1.188)

cTTR >76.51% 15 0.778 0.717 (0.410-1.254)

HR = hazard ratio.

Supplemental Table III. Intracranial hemorrhage risk by center's time in therapeutic range (cTTR) quartiles for apixaban.

Stroke No. Event Rate HR (95% CI)

cTTR <52.38% 5 0.259 0.581 (0.223-1.514)

52.38% < cTTR < 66.02% 26 0.447 1.000 (Ref.)

66.02% < cTTR < 76.51% 18 0.307 0.688 (0.377-1.256)

cTTR >76.51% 3 0.161 0.361 (0.109-1.194)

HR = hazard ratio.

>3 as specified in the seminal ARISTOTLE publication. As demonstrated in Supplemental Table IX, this risk was assessed separately for each treatment arm. Weighted average risk of ischemic or unspecified stroke was 0.981 and 1.077 for apixaban and warfarin, respectively, with weights computed based on number of patients in different CHADS2 categories for each treatment arm within ARISTOTLE.

February 2014 210.e11

Supplemental Table IV. Intracranial hemorrhage risk according to center's time in therapeutic range (cTTR)

quartiles for warfarin.

Stroke No. Event Rate HR (95% CI)

cTTRo 52.38% 18 0.997 1.052 (0.617-1.793)

52.38% < cTTR <66.02°% 54 0.941 1.000 (Ref.)

66.02%% < cTTR <76.51% 37 0.633 0.681 (0.448-1.034)

cTTR >76.51% 15 0.721 0.777 (0.424-1.424)

HR = hazard ratio.

Supplemental Table V. Other major bleeding risk quartiles for apixaban. according to center's time in therapeutic range (cTTR)

Stroke No. Event Rate HR (95% CI)

cTTR <52.38% 19 0.991 0.722 (0.438-1.191)

52.38% < cTTR < 66.02% 79 1.369 1.000 (Ref.)

66.02% < cTTR < 76.51% 132 2.297 1.688 (1.277-2.231)

cTTR >76.51% 44 2.402 1.765 (1.221-2.552)

HR = hazard ratio.

Supplemental Table VI. Other major bleeding risk according quartiles for warfarin. to center's time in therapeutic range (cTTR)

Stroke No. Event Rate HR (95% CI)

cTTR <52.38% 52.38% < cTTR <66.02% 66.02% < cTTR <76.51% cTTR >76.51% 32 120 137 51 1.796 2.120 2.383 2.876 0.843 (0.571-1.245) 1.000 (Ref.) 1.130 (0.884-1.443) 1.365 (0.984-1.895)

HR = hazard ratio.

Supplemental Table VII. Clinically relevant nonmajor bleeding risk according to center's time in therapeutic

range (cTTR) quartiles for apixaban.

Stroke No. Event Rate HR (95%% CI)

cTTR <52.38% 25 1.307 0.711 (0.460-1.100)

52.38% < cTTR <66.02%% 105 1.829 1.000 (Ref.)

66.02%% < cTTR <76.51%% 130 2.265 1.247 (0.964-1.613)

cTTR >76.51% 56 3.094 1.702 (1.230-2.354)

HR = hazard ratio.

Supplemental Table VIII. Clinically relevant nonmajor bleeding risk according to center's time in therapeutic range (cTTR) quartiles for warfarin.

Stroke No. Event Rate HR (95% CI)

cTTR < 52.38% 47 52.38%% < cTTR <66.02%% 150 66.02%% < cTTR <76.51%% 189 cTTR >76.51% 58 2.659 2.667 3.316 3.341 0.987 (0.711-1.369) 1.000 (Ref.) 1.256 (1.014-1.556) 1.266 (0.935-1.714)

HR = hazard ratio.

Supplemental Table IX. Ischemic or unspecified stroke risk according to CHADS2 (congestive heart failure, hypertension, age >75 years, diabetes mellitus, prior stroke or transient ischemic attack, or thromboembolism) score.

CHADS2 Score Apixaban Warfarin

0-1 0.521 0.458

2 0.950 0.934

>3 1.534 1.944

Weighted average 0.981 1.077

AVERAGE RISK ASSOCIATED WITH BLEEDING EVENTS

The average risk of events, as described in Supplemental Table X, was observed in patients using apixaban versus warfarin in ARISTOTLE patients.

Hemorrhagic Stroke as a Part of Intracranial Hemorrhages

Hemorrhagic strokes comprised 77% of intracranial hemorrhages observed in apixaban-treated patients

(40 of 52) versus 64% of intracranial hemorrhages observed in warfarin-treated patients (78 of 122). As stated earlier, the absolute event risk of experiencing intracranial hemorrhages was higher with warfarin versus apixaban.

Severity Distribution Associated with Strokes as Assessed by Modified Rankin Score

The distribution of stroke severity was observed by using modified Rankin Scale scores at 30-day follow-up postevent for apixaban

Supplemental Table X. Average risk of bleeding events as observed in ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) in apixaban and warfarin arms.

Intracranial hemorrhage

Other major bleeds excluding intracranial hemorrhages Clinically relevant nonmajor bleeding

(Supplemental Table XI) versus warfarin (Supplemental Table XII). Patients with a modified Rankin Scale score of 6 represented fatality associated with the stroke event.

Apixaban Warfarin

(per 100 Patient-Years) (per 100 Patient-Years)

0.330 0.800

1.790 2.270

2.083 2.995

Ischemic or Unspecified Stroke

Supplemental Table XI presents the stroke severity classification associated with ischemic or unspecified stroke.

Supplemental Table XI. Stroke severity classification associated with ischemic or unspecified stroke. Values are given as number (%).

mRS classification Apixaban Warfarin

0-2 57 (53) 49 (45)

3-4 23 (21) 32 (30)

5 9 (8) 11 (10)

6 19 (18) 16 (15)

^Information about modified Rankin Scale (mRS) classification was not available for 54 ischemic or unspecified stroke events in the apixaban arm and 67 ischemic or unspecified events in the warfarin arm of ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation).

Supplemental Table XII. Stroke severity classification associated with hemorrhagic stroke. Values are given as number (%).

mRS Classification Apixaban Warfarin

0-2 7 (23°%) 13 (20°%)

3-4 10 (32%) 10 (15%)

5 3 (10%) 8 (12%)

6 11 (35%) 34 (53%)

^Information about modified Rankin Scale (mRS) classification was not available for 9 hemorrhagic stroke events in the apixaban arm and 13 hemorrhagic stroke events in the warfarin arm of ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation).

Supplemental Table XIII. Treatment discontinuation rates not related to absorbing # states.*

Treatment No. Event Rate HR (95% CI)

Warfarin 2182 14.405 1.089 (1.025-1.157)

Apixaban 2047 13.177 1.000 (Ref.)

HR = hazard ratio.

*Excluding stroke, major bleeding, myocardial infarction, or systemic embolism.

Supplemental Table XIV. Pooled analysis of fatality rates associated with apixaban versus warfarin or aspirin

using ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) and AVERROES (Apixaban Versus Acetylsalicylic Acid to Prevent Strokes) data.

Trial Treatment Other ICH Fatal Other ICH OMB Fatal OMB

AVERROES Apixaban 5 1 (20%%) 34 1 (3%%)

Aspirin 2 1 (50%) 18 1 (6%%)

ARISTOTLE Apixaban 12 2 (17%%) 275 6 (2%%)

Warfarin 44 4 (9%%) 340 7 (2%%)

Overall average 63 8 (13%%) 667 15 (2%%)

ICH = intracranial hemorrhage; OMB = other major bleed.

Hemorrhagic Stroke

Supplemental Table XII presents the stroke severity classification associated with hemorrhagic stroke.

Treatment Discontinuation Rates Not Related to Absorbing States

Treatment discontinuation rates for reasons other than stroke or major bleeding were higher for warfarin versus those observed for apixaban. This resulted in greater HRs for warfarin versus apixaban both for treatment discontinuations related to: (1) excluding stroke or major bleeding; and (2) excluding stroke, major bleeding, myocardial infarction, or systemic embolism (Supplemental Table XIII).

Fatality Rates Associated with Bleeding Episodes

Pooled data analyses, as described in Supplemental Table XIV, indicated 13% mortality associated with intracranial hemorrhage that did not manifest into hemorrhagic stroke. Pooled mortality rate for other

major bleeds that were not intracranial in nature was 2%.

Mortality Risk Associated With Causes Other Than Stroke, Systemic Embolism, Major Bleeding, or Myocardial Infarction

Over the duration of the ARISTOTLE trial, apixaban had a nonstroke, non-systemic embolism, non-bleeding, and nonmyocardial infarction mortality rate of 3.0825 per 100 patient-years (n = 528) versus warfarin, which had mortality rate of 3.3404 per 100 patient-years (n = 568). The resulting hazard ratio was 1.0836 (95% CI, 0.962-1.220).

REFERENCES

1. Lopes RD, Alexander JH, Al-Khatib S et al. Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation (ARISTOTLE) trial: design and rationale. Am Heart J. 2010;159:331-339.

2. Granger CB, Alexander JH, McMur-ray JJ, et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med. 2011; 365:981-992.

3. 2011 Writing Group members. 2011 ACCF/AHA/HRS focused updated on the management of patients with atrial fibrillation (updating the 2006 guidelines). Circulation. 2011;123:104-123.

4. National Institute for Health and Care Excellence (NICE) TA287. Apixaban (Eliquis) for the prevention of stroke and systemic embolism in people with non-valvular atrial fibrillation. A single technology appraisal; Manufacturer submission; 17 August 2012.

SUPPLEMENTAL APPENDIX C

Cost-Effectiveness of Apixaban Versus Other New Oral Anticoagulants for Stroke Prevention in Patients with Atrial Fibrillation: Secondary Analysis of AVERROES (CV185-048) to Support Apixaban Cost-Effectiveness Modelling for the Indication of Stroke Prevention in Atrial Fibrillation

STUDY DESIGN

Study design and data collection associated with the AVERROES (ApixabanVersus Acetylsalicylic Acid to Prevent Strokes) trial have been described previously.1 Seminal results, published by Connolly et al,2 indicated that apixaban reduced the risk of stroke or systemic embolism without significantly increasing the risk of major bleeding or intracranial hemorrhage.

METHODS

AVERROES data were analyzed on a post hoc basis. Annualized event rates and related hazard ratios (HRs) were computed for the relevant end points. The intention-to-treat (ITT) principle was used for the analysis of all end points except the safety end points (which were based on the modified ITT population). Stratified analyses were conducted across CHADS2 (congestive heart failure, hypertension, age >75 years, diabetes mellitus, prior stroke or transient ischemic attack, or thromboembolism) categories.3 All risks were assessed in absolute terms or relative risk was computed by using HRs for the events associated with aspirin versus apixaban. Subgroup analyses were conducted in AVERROES participants who had previously used warfarin (vitamin K antagonists [VKAs]) but were not able to stay on the VKA due to their inability to maintain international normalized ratio control while on warfarin, because they experienced either bleeding or nonbleeding events while on warfarin, or for other similar reasons described in the AVERROES publication.2 For the purposes of these analyses, the patients were labeled as the warfarin-unsuitable population (ie, proven unsuitable after experiencing warfarin use).

Mortality risk associated with reasons other than stroke, systemic embolism, bleeding, and myocardial infarction was assessed by deleting mortality associated with the aforementioned causes from all-cause mortality. Absolute mortality risk as well as relative hazard of mortality with aspirin versus apixaban for causes other than stroke, bleeding, and myocardial

infarction was assessed. Apixaban was used as a reference category for computing HRs associated with these events.

Data were pooled from ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) and AVERROES treatment arms to assess percentage of intracranial hemorrhages manifesting into hemorrhagic stroke, as well as fatality rates associated with intracranial hemorrhages (excluding hemorrhagic stroke and other major bleeds [excluding any type of intracranial hemorrhage]). For assessing distribution of hemorrhagic stroke across different levels of stroke severity, events were pooled across the apixaban and aspirin arms due to low absolute event rates in both arms.

All analyses were conducted by using SAS version 9.2 (SAS Institute, Inc, Cary, North Carolina). Because this was a secondary analysis of AVERROES analyses sets conducted in a post hoc manner, please refer to the article by Connolly et al for efficacy and safety results obtained based on prespecified primary and secondary hypotheses.

Risk of Ischemic or Unspecified Stroke According to CHADS2 Score

Risk of ischemic or unspecified stroke was assessed at prespecified CHADS2 categories of 0 to 1, 2, and > 3, as specified in the AVERROES publication.2 As demonstrated in Supplemental Table I, this risk was assessed separately for each treatment arm. Weighted average risk of ischemic or unspecified stroke was 1.374 and 3.103 for apixaban and aspirin, respectively, with weights computed based on number of patients in different CHADS2 categories for each treatment arm within AVERROES. Compared with apixaban, patients treated with aspirin had a greater risk of experiencing ischemic or unspecified stroke (HR = 2.270 [95% CI, 1.590-3.230]).

Risk of Experiencing Bleeding or Other Events

Supplemental Table II describes the absolute event rates (per 100 patient-years) and relative risk (as HRs) of experiencing an event when treated with aspirin versus apixaban.

Hemorrhagic Stroke as a Part of Intracranial Hemorrhages

Hemorrhagic strokes comprised 55% of intracra-nial hemorrhages observed in patients treated with

Supplemental Table I. Ischemic or unspecified stroke risk according to hypertension, age >75 years, diabetes mellitus, attack, or thromboembolism) score. CHADS2 (congestive heart failure, prior stroke or transient ischemic

CHADS2 Score Apixaban Aspirin

0-1 0.831 1.411

2 1.526 3.363

>3 1.957 5.196

Weighted average 1.374 3.103

Supplemental Table II. Absolute and relative hazard of experiencing an event with apixaban versus aspirin in the AVERROES (Apixaban Versus Acetylsalicylic Acid to Prevent Strokes) trial population.

Apixaban Absolute Event Event Rate Aspirin Absolute Event Rate Hazard Ratio (95% CI)

Intracranial hemorrhages 0.344 Other major bleeds excluding intracranial 1.066 hemorrhages Clinically relevant nonmajor bleeding 3.113 CV hospitalizations unrelated to stroke or MI 10.460 0.348 0.571 2.371 12.087 1.013 (0.439-2.337) 0.535 (0.302-0.947) 0.762 (0.563-1.030) 1.155 (0.992-1.345)

CV = cardiovascular; MI = myocardial infarction.

apixaban and aspirin. As described in the AVERROES publication,2 patients taking apixaban had slightly lower risk of experiencing intracranial hemorrhages (HR = 0.85 [0.38-1.90]) well as hemorrhagic stroke (HR = 0.37 [0.24-1.88]) compared with patients taking aspirin.

Severity Distribution Associated With Strokes as Assessed by Modified Rankin Scale

The distribution of stroke severity was observed by using the modified Rankin Scale at 30-day follow-up postevent for apixaban versus aspirin. Patients with a modified Rankin Scale score of 6 represented fatality associated with the stroke event. Stroke severity distribution associated with ischemic or unspecified stroke according to treatment is described in Supplemental Table III. For hemorrhagic stroke, events were pooled across the apixaban and aspirin

arms due to low absolute event rate in both arms for this analysis (Supplemental Table IV).

Gastrointestinal Bleeding Rates as Part of Other Major Bleeding (Excluding Intracranial Hemorrhages)

Gastrointestinal bleeding comprised 35% (12 of 34) and 39% (7 of 18) of other major bleeding events (excluding intracranial hemorrhages) for apixaban and aspirin, respectively.

Treatment Discontinuation Rates Not Related to Absorbing States

Treatment discontinuation rate for reasons other than stroke or major bleeding were higher for aspirin versus that observed for apixaban, resulting into greater HRs for aspirin versus apixaban for treatment discontinuations not related to: (1)

Supplemental Table II I. Stroke severity classification associated with are given as number (%). ischemic or unspecified stroke. Values

mRS Classification Apixaban Aspirin

0-2 17 (40) 35 (36)

3-4 12 (28) 37 (38)

5 5 (12) 15 (15)

6 9 (20) 10 (11)

mRS = modified Rankin Scale.

Supplemental Table IV. Stroke severity classification associated with hemorrhagic stroke.

mRS Apixaban Aspirin Pooled Sample

Classification (No.) (No.) (No. [%])

0-2 1 0 1 (7)

3-4 1 2 3 (20)

5 0 4 4(27)

6 4 3 7 (46)

mRS = modified Rankin Scale.

Supplemental Table V. Treatment discontinuation rates not related to absorbing states.

Drug No. Event Rate (per 100 Patient Years) Hazard Ratio (95% CI)

Aspirin 537 19.012 1.099 (0.972-1.241)

Apixaban 495 17.310 1.000 (Ref.)

*Excluding stroke, major bleeding, myocardial infarction, and systemic embolism.

stroke or major bleeding; or (2) stroke, major bleeding, myocardial infarction, or systemic embolism (Supplemental Table V).

Fatality Rates Associated With Bleeding Episodes

Pooled data analyses, as described in Supplemental Table VI, indicated 13% mortality associated with intracranial hemorrhage that did not manifest into

hemorrhagic stroke. The pooled morality rate for other major bleeds that were not intracranial in nature was 2%.

Mortality Risk Associated With Causes Other Than Stroke, Systemic Embolism, Major Bleeding, or Myocardial Infarction

Over the duration of the AVERROES trial, the aspirin arm exhibited higher nonstroke, nonsystemic

Supplemental Table VI. Pooled analysis of fatality rates associated with apixaban Supplemental versus

warfarin or aspirin by using the ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) and AVERROES (Apixaban Versus Acetylsalicylic Acid to Prevent Strokes) data.

Trial Treatment Other ICH Fatal Other ICH OMB Fatal OMB

AVERROES Apixaban 5 1 (2000) 34 1 (300)

Aspirin 2 1 (5000) 18 1 (600)

ARISTOTLE Apixaban 12 2 (1700) 275 6 (200)

Warfarin 44 4 (90) 340 7(25)

Average 63 8 (1300) 667 15 (200)

ICH = intracranial hemorrhages; OMB = other major bleeds.

Supplemental Table VII. Absolute event rates in warfarin-unsuitable patients randomized to receive aspirin.

Event Annualized Event Rate (per 100 Patient-Years)

Ischemic or unspecified stroke 3.453

Intracranial hemorrhages 0.322

Other major bleeds excluding intracranial hemorrhages 0.887

Clinically relevant nonmajor bleeds 2.936

Myocardial infarction 1.110

CV hospitalizations unrelated to stroke or MI 13.571

CV = cardiovascular; MI = myocardial infarction.

embolism, nonbleeding, and nonmyocardial infarction mortality rates of 3.5935 per 100 patient-years (n = 114) versus apixaban, which had a mortality rate of 2.9668 per 100 patient-years (n = 94). This yielded a slightly greater HR of 1.212 (95% CI, 0.9221.593).

Absolute Annualized Event Rates in Warfarin-Unsuitable Patients Randomized to Aspirin Arm

Supplemental Table VII describes annualized event rates for various events of interest in warfarin-unsuitable patients randomized to the aspirin arm.

REFERENCES

1. Eikelboom JW, O'Donnell M, YusufS, et al. Rationale and design of AVERROES: apixaban versus acetylsalicylic acid to prevent stroke in atrial fibrillation patients who have failed or are unsuitable for vitamin K antagonist treatment. Am Heart J. 2010;159:348-353.

2. Connolly S, Eikelboom J, Joyner C et al. Apixaban in patients with atrial fibrillation. N Engl J Med. 2011;364:806-817.

3. 2011 Writing Group members. 2011 ACCF/AHA/HRS Focused Updated on the Management of Patients with Atrial Fibrillation (Updating the 2006 Guidelines). Circulation. 2011;123:104-123.