Gastrointestinal bleeding in patients with atrial fibrillation treated with Apixaban or warfarin: Insights from the Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation (ARISTOTLE) trial

Gastrointestinal bleeding in patients with atrial fibrillation treated with Apixaban or warfarin: Insights from the Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation (ARISTOTLE) trial

Journal Pre-proof Gastrointestinal bleeding in patients with atrial fibrillation treated with Apixaban or warfarin: Insights from the ARISTOTLE trial ...

1MB Sizes 0 Downloads 51 Views

Journal Pre-proof Gastrointestinal bleeding in patients with atrial fibrillation treated with Apixaban or warfarin: Insights from the ARISTOTLE trial

David A. Garcia, Deborah A. Fisher, Hillary Mulder, Lisa Wruck, Raffele De Caterina, Sigrun Halvorsen, Christopher B. Granger, Claes Held, Lars Wallentin, John H. Alexander, Renato D. Lopes PII:

S0002-8703(19)30295-9

DOI:

https://doi.org/10.1016/j.ahj.2019.10.013

Reference:

YMHJ 6004

To appear in:

American Heart Journal

Received date:

9 April 2019

Accepted date:

26 October 2019

Please cite this article as: D.A. Garcia, D.A. Fisher, H. Mulder, et al., Gastrointestinal bleeding in patients with atrial fibrillation treated with Apixaban or warfarin: Insights from the ARISTOTLE trial, American Heart Journal(2019), https://doi.org/10.1016/ j.ahj.2019.10.013

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier.

Journal Pre-proof

1

Gastrointestinal Bleeding in Patients with Atrial Fibrillation Treated with Apixaban or Warfarin: Insights from the ARISTOTLE Trial

David A. Garcia, MD,* Deborah A. Fisher, MD,†,‡ Hillary Mulder, MS,‡ Lisa Wruck, PhD,‡ Raffele De Caterina, MD, PhD,§ Sigrun Halvorsen, MD, PhD,|| Christopher B. Granger, MD,‡,¶ Claes Held, MD, PhD,** Lars Wallentin, MD, PhD,** John H. Alexander, MD, MHS, ‡,¶

of

Renato D. Lopes, MD, PhD, MHS‡

Division of Hematology, University of Washington School of Medicine, Seattle, WA; †Division of

ro

*

-p

Gastroenterology, Department of Medicine, Duke University School of Medicine, Durham, NC; ‡Duke

re

Clinical Research Institute, Durham, NC; §Cardiology Division, G d’Annunzio University, Cheiti, Italy; ||

Department of Cardiology, Oslo University Hospital, Oslo, Norway; ¶Division of Cardiology,

lP

Department of Medicine, Duke University School of Medicine, Durham, NC; **Uppsala Clinical Research

na

Center, Uppsala University, Uppsala, Sweden.

ur

Trial registration: ClinicalTrials.gov: NCT00412984

Jo

Funding: The ARISTOTLE trial was funded by Bristol-Myers Squibb and Pfizer, Inc.

Address for correspondence: Renato D. Lopes, MD, PhD, MHS, Duke Clinical Research Institute, 2400 Pratt Street, Durham, NC 27710. Phone: 919-668-8241; Fax: 919-668-7085. Email: [email protected].

Journal Pre-proof

2

Objectives: A history of gastrointestinal bleeding (GIB) in patients with atrial fibrillation (AF) may impact decisions about anticoagulation treatment. To sought to determine whether prior GIB in patients with AF taking anticoagulants was associated with an increased risk of stroke or major hemorrhage. Methods: We analyzed key efficacy and safety outcomes in patients with prior GIB in ARISTOTLE. Centrally adjudicated outcomes according to GIB history were analyzed using Cox proportional hazards models adjusted for randomized treatment and established risk factors.

of

Results: A total of 784 (4.3%) patients had prior GIB (321 [41%] lower, 463 [59%] upper); 215 (27%) occurred <1 year before study enrollment. Patients with prior GIB were older, had more comorbidities,

ro

and higher CHADS2 and HAS-BLED scores than those with no GIB. Major GIB occurred more

-p

frequently in those with prior GIB (lower: aHR 1.72, 95% CI 0.86–3.42; upper: aHR 3.13, 95% CI 1.97–

re

4.96). This association with major GIB was more pronounced in patients with GIB <1 year before randomization versus no recent GIB (recent lower: aHR 2.58, 95% CI 0.95–7.01; recent upper: aHR 5.16,

lP

95% CI 2.66–10.0). There was no association between prior GIB and risk of stroke/systemic embolism or

na

all-cause death. In those with prior GIB, the apixaban versus warfarin relative risks for stroke/systemic embolism, hemorrhagic stroke, death, or major bleeding were consistent with the results of the overall

ur

trial.

Jo

Conclusions: In patients with AF on oral anticoagulants, prior GIB was associated with an increased risk of subsequent major GIB but not stroke, intracranial bleeding, or all-cause mortality. For the key outcomes of stroke, hemorrhagic stroke, death, and major bleeding, we found no evidence that the treatment effect (apixaban vs. warfarin) was modified by a history of GIB.

Keywords: gastrointestinal hemorrhage; atrial fibrillation; anticoagulants

Journal Pre-proof

3

In patients with atrial fibrillation (AF), a history of gastrointestinal (GI) bleeding may impact decisions about anticoagulation treatment. More than one study of warfarin-treated patients has found that a history of GI bleeding is independently associated with an increased risk of major hemorrhage.1,2 Clinicians must weigh the risks associated with a history of GI bleeding against the benefits of anticoagulation to prevent ischemic stroke. Apixaban is an oral, direct factor Xa inhibitor previously compared with warfarin in a large,

of

randomized trial of patients with AF. Although apixaban was superior to warfarin for 3 of the key study endpoints (stroke or systemic embolism, major bleeding, and all-cause death), the rates of major GI

ro

bleeding in the apixaban- and warfarin-treated patients were not statistically different.3

-p

We analyzed data from the Apixaban for Reduction in Stroke and Other Thromboembolic Events

re

in Atrial Fibrillation (ARISTOTLE) trial to determine whether and how a history of GI bleeding might impact the rate of several key study endpoints, including stroke or systemic embolism, major bleeding,

lP

and all-cause death. For these same endpoints, we also sought to determine whether a history of GI

na

bleeding would modify the relative treatment effects (apixaban vs. warfarin) that were seen in the overall trial. We further assessed whether the location (lower vs. upper) or the timing (recent vs. remote) of prior

METHODS

Jo

ur

GI bleeding might be associated with different GI bleeding risks.

ARISTOTLE (NCT00412984) was a double-blind, randomized controlled trial that enrolled 18,201 patients with AF and at least 1 additional risk factor for stroke; the trial design and primary results have been previously reported.3,4 Using data from ARISTOTLE, we analyzed rates of key efficacy and safety outcomes to determine whether a history of GI bleeding (prior to study entry) was associated with an increased risk of subsequent stroke/systemic embolism, major bleeding, myocardial infarction (MI), or death.

Journal Pre-proof Definitions and outcomes For the present analysis, we used information that was recorded at study entry (demographics, geographical region, clinical characteristics, medical history, bleeding risk scores, and medications). History of GI bleeding and location of the GI bleeding event (upper or lower) were extracted from the case report form. We explored whether recent (vs. more remote) GI bleeding was associated with key study outcomes (stroke or systemic embolism, major bleeding, and all-cause death). Key study outcomes

of

were centrally adjudicated and outcome definitions have been detailed elsewhere.4 For this analysis, we considered GI bleeding to be recent if it had occurred less than 1 year prior to randomization. Creatinine

ro

clearance was estimated using the Cockroft-Gault formula and stratified as normal, mildly impaired

re

-p

(>50–80 mL/min), moderately impaired (>30–50 mL/min), and severely impaired (≤30 mL/min).

Statistical analysis

lP

Baseline characteristics are presented for the randomized population with information on prior GI

na

bleeding recorded in the case report form (n=18,197). We excluded from this analysis 4 patients for whom information on prior GI bleeding was missing. Continuous baseline variables are summarized as

ur

medians and 25th and 75th percentiles; comparisons between GI bleeding groups were made using the

Jo

Kruskal-Wallis test. Categorical variables are presented as frequencies and percentages and compared using chi-square tests.

Efficacy outcomes were analyzed using the intention-to-treat population with available data on prior GI bleeding (n=18,197) through the efficacy censoring date. For safety outcomes, we used the ontreatment population (18,140 patients receiving ≥1 dose of study drug) and included all events from the time of the first dose until 2 days after the last dose of study drug was received. Centrally adjudicated outcomes according to GI bleeding history were analyzed using a Cox proportional hazards model adjusted for randomized treatment assignment and pre-specified established risk factors, including age, geographic region, and aspirin and nonsteroidal anti-inflammatory drug (NSAID) use at baseline. Additional covariates varied by outcome and were pre-specified. For stroke outcomes, adjustment

4

Journal Pre-proof

5

variables included weight, diabetes, hypertension, moderate valvular disease, prior stroke/transient ischemic attack (TIA)/systemic embolism, type of AF, and prior vitamin K antagonist (VKA) use. For death outcomes, adjustment variables included sex, weight, systolic blood pressure, diastolic blood pressure, hypertension, moderate valvular disease, left bundle branch block, prior MI, prior stroke/TIA/systemic embolism, anemia, smoking, prior VKA use, New York Heart Association (NYHA) class, CHADS2 score, and renal function. Adjustment variables for bleeding outcomes included sex,

of

coronary artery disease (CAD), prior MI, prior non-GI bleeding, anemia, CHADS2 score, and renal function. Adjustment variables for MI included diabetes, CAD, prior MI, NYHA class, and renal

ro

function. To determine whether prior GI bleeding modified the effect of randomized treatment assignment

-p

(apixaban vs. warfarin) on key study endpoints, interaction terms were tested in the Cox models. Analyses

re

were repeated for recent GI bleeding. In the primary analysis, the small number (n=32) of patients who had a history of both upper and lower GI bleeding recorded at study entry were categorized as having a

lP

history of upper GI bleeding; however, we also performed sensitivity analyses in which we categorized

na

these patients as having a history of lower GI bleeding. To compare the rates of death within 30 days after a GI bleeding event in the apixaban and

ur

warfarin treatment groups, we fit Cox models with time-dependent indicators for GI bleeding, taking a

Jo

value of 1 within 30 days after a GI bleed occurring prior to censoring for efficacy (and the value of 0 otherwise), and adjusting for the same set of covariates described earlier. To improve model fit and create an appropriate comparator, we also included time-dependent indicators for intracranial bleeds and all other bleeds (non-GI and non-intracranial). We tested the interaction between the GI bleeding indicator and randomized treatment to assess whether the short-term risk of death associated with GI bleeding differed between the treatment arms. In patients with multiple bleeding events, only the first event of each type was assessed. We also describe international normalized ratio (INR) values before and after discontinuation of study drug after GI bleeding.

Journal Pre-proof

6

Corrections for multiple testing were not performed due to the hypothesis-generating nature of the secondary analysis. All analyses were performed by the Duke Clinical Research Institute using SAS software, version 9.4 (SAS Institute, Inc., Cary, NC).

RESULTS Baseline Characteristics

of

The 784 (4.3%) enrolled patients with a history of GI bleeding (321 lower, 463 upper) were older, disproportionately male, and more likely to have impaired renal function. Compared with the 17,413

ro

patients without a history of GI bleeding, patients with a history of GI bleeding were more likely to have a

-p

history of 1 or more comorbidities at study entry, including anemia, dyspepsia, gastroesophageal reflux,

re

GI surgery, peptic ulcer disease, chronic liver disease, and either peripheral or coronary artery disease

lP

(Table 1).

na

Clinical Outcomes According to GI Bleeding History, Location, and Timing After adjustment for covariates, prior lower (but not upper) GI bleeding was associated with an increased

ur

risk of MI (adjusted hazard ratio [HR] 1.96; 95% confidence interval [CI] 1.02–3.77). We did not find

Jo

statistically significant associations between prior GI bleeding (upper or lower) and other key efficacy endpoints (stroke or systemic embolism, ischemic/uncertain-cause stroke, all-cause death, cardiovascular death) (Table 2).

Patients with a history of lower GI bleeding at study entry were more likely to experience major lower GI bleeding during the study (adjusted HR 5.47; 95% CI 2.53–11.80). Several other safety endpoints were associated with a history of lower GI bleeding, but the associations were less strong. Among patients with a history of lower GI bleeding, we did not observe a statistically significant association with overall major bleeding, major upper GI bleeding, or intracranial bleeding (Table 2). For patients with a history of upper GI bleeding at study entry, we did not find a statistically significant association with any efficacy endpoint. However, a history of upper GI bleeding at study entry

Journal Pre-proof

7

was associated with overall major GI bleeding (adjusted HR 3.13; 95% CI 1.97–4.96), major upper GI bleeding (adjusted HR 3.42; 95% CI 2.02–5.81), and overall major bleeding (HR 2.06; 95% CI 1.52– 2.79). Weaker associations with non-major bleeding outcomes are shown in Table 2. A history of upper GI bleeding was not associated with an increased risk for intracranial hemorrhage. Of patients with prior GI bleeding, 60% stopped study drug at some time during the trial (vs. 50% among those with no history of GI bleeding). Of patients in the safety population with a prior history of

of

GI bleeding, 58% in the apixaban arm stopped anticoagulation at some point during the trial and 62% in the warfarin arm stopped anticoagulation at some point during the trial (p=0.23).

ro

At study entry, 215 patients had experienced recent GI bleeding (<1 year prior to randomization).

-p

Of these, 97 were lower GI bleeding events and 118 upper GI bleeding events. No major differences from

re

the associations with upper or lower prior GI bleeding were observed, although some of the associations

lP

lost statistical significance (Table 3).

na

Treatment Effect of Apixaban versus Warfarin According to Gastrointestinal Bleeding A history of GI bleeding did not change the treatment effect (apixaban vs. warfarin) that was reported

ur

from the overall trial for stroke/systemic embolism, hemorrhagic stroke, death, or major bleeding (Table

Jo

4). Of the randomized population, 218 patients experienced major GI bleeding during the trial, prior to the efficacy censoring date. Among them, 20 (8.8%) died within 30 days of the hemorrhage, including 11/102 (10.8%) in the apixaban arm and 9/116 (7.8%) in the warfarin arm. There was an increased risk of death during the 30 days after a GI bleeding event in both the apixaban and warfarin arms (adjusted HR 29.15; 95% CI 15.96–53.25 and HR 12.25; 95% CI 6.13–24.89, respectively). After multivariable adjustment, the difference between the randomized treatment arms in the increased risk of death within 30 days was not statistically significant (p=0.07). Kaplan-Meier curves for the 30-day incidence of death among patients with GI bleeding, separated by treatment arm, are presented in Figure 1. The apixaban and warfarin groups were similar with respect to not only the proportion of patients not on study medication at the time of GI bleeding (approximately 13%), but also the proportion of

Journal Pre-proof

8

patients who interrupted study medication for at least 1 day either on the day of or within the 3 days following the GI bleeding event (approximately 75%). Of those who interrupted study medication, approximately 50% eventually resumed. The median value for the most recent INR prior to the GI bleeding event was 1.2 (25th, 75th: 1.1, 1.4) in the apixaban arm and 2.5 (2.1, 3.0) in the warfarin arm (Table 5).

of

DISCUSSION In this analysis of patients with AF and prior GI bleeding who were randomly assigned to receive either

ro

apixaban or warfarin in the ARISTOLE trial, we found that a previous history of lower GI bleeding was

-p

strongly associated with an increased risk of major lower GI bleeding, and a history of prior upper GI

re

bleeding was associated with an increased risk of major upper GI bleeding. These anatomy-specific associations build upon a prior analysis that looked at the relationship between a history of bleeding and

lP

the risk of major hemorrhage while on anticoagulation.5 For example, prior upper GI bleeding, but not

na

prior lower GI bleeding, was associated with a greater than 3-fold risk of any major GI bleeding during the study period. This means that if a patient with AF and a history of upper GI bleeding decides to begin

ur

anticoagulation, they can expect that their annual risk for major GI bleeding will be approximately 3 to 6 percentage points higher than it would be in a similar patient without prior GI bleeding, for whom the

Jo

annual risks of upper and lower GI bleeding are 1% and 1.5%, respectively.6 Our finding that neither a history of upper GI bleeding nor a history of lower GI bleeding was associated with intracranial hemorrhage, the most feared complication of anticoagulant therapy, may be relevant to clinicians who are weighing the risks and benefits of anticoagulation in a patient with AF. While this observation might lead clinicians to pause before prescribing an anticoagulant to a patient with both AF and a history of GI bleeding, the risk-benefit tradeoffs will favor anticoagulation treatment for most such patients. Our finding that the treatment effect (apixaban vs. warfarin) is not modified by prior GI bleeding means that, for a patient with AF who chooses to take anticoagulant therapy, previous GI bleeding itself would not favor warfarin over apixaban or vice versa. Our results

Journal Pre-proof

9

support guideline recommendations that apixaban would—along with warfarin, dabigatran 110 mg twice daily, and edoxaban 30 mg once daily—be a preferred anticoagulant strategy for patients with AF and a history of GI bleeding.7 Among the warfarin-treated patients who experienced GI bleeding, our observations that the last INR, performed a median of 11 days prior to the GI bleeding event, was between 2.1 and 3.0 in 50% are consistent with similar findings from an analysis of 174 patients with intracranial hemorrhage in

of

ARISTOTLE. Nearly 80% of these patients’ pre-intracranial hemorrhage INR values were in the target range.8 Other previously published cohort studies further highlight the imperfect nature of the INR

ro

measurement as a tool to predict major warfarin-associated bleeding.9,10

-p

The small number (11 vs. 9) of deaths within 30 days of major GI bleeding, along with the

re

numerous covariates that could have affected the rates we observed, prevent us from determining whether there is a difference in the case-fatality rates following major GI bleeding for either apixaban or warfarin.

lP

A previous meta-analysis suggested that, overall, direct oral anticoagulant-associated major bleeding is

na

not more likely to cause death within 30 days than warfarin-associated major bleeding.11 The main limitation of our analysis is that many of the comparisons (e.g., history of prior GI

ur

bleeding vs. none) were not randomized. Thus, although we attempted to adjust for potentially

Jo

confounding variables, our findings should be interpreted with caution. It is possible that the low number of patients with prior GI bleeding prevented us from detecting a difference in the risk of one or more of the outcomes listed in Table 2.

In conclusion, a history of GI bleeding—especially recent GI bleeding—was associated with an increased risk of subsequent anticoagulant-associated major GI bleeding among anticoagulant-treated patients with AF. Although this risk increase may change calculations regarding the net benefit of anticoagulation therapy in general, we found no evidence that a history of GI bleeding would substantially modify what is already known about the superiority of apixaban as compared with warfarin in the relative risks for important clinical outcomes such as stroke or systemic embolism, major bleeding, and all-cause death.

Journal Pre-proof

10

Disclosures Garcia: Research grants from BMS/Pfizer and Daiichi Sankyo; consulting fees from BMS/Pfizer and Janssen. Fisher, Mulder, Wruck: None. De Caterina: Research grants from Pfizer; Bayer HealthCare Pharmaceuticals, and Boehringer Ingelheim; consulting fees from Pfizer, BMS, Bayer HealthCare Pharmaceuticals, Boehringer Ingelheim, Daiichi Sankyo, Eli Lilly, Novartis, Roche. Halvorsen: Consulting fees from BMS, Pfizer, AstraZeneca, Bayer, Boehringer Ingelheim, Merck, and Sanofi.

of

Granger: Research grants from Bristol-Myers Squibb, Pfizer, Inc., Apple s, AstraZeneca, Bayer, Boehringer Ingelheim, US FDA, GlaxoSmithKline, Janssen, Medtronic Foundation, Novartis; consulting

ro

fees from Bristol-Myers Squibb, Pfizer, Inc., Abbvie, AstraZeneca, Bayer, Boehringer Ingelheim, Boston

-p

Scientific, GlaxoSmithKline, Janssen, Medscape, Medtronic, Inc., Merck, NIH, Novo Nordisk, Novartis,

re

Roche Diagnostics, Rho Pharmaceuticals, Sirtex, Verseon. Held: Institutional research grants from AstraZeneca, Bristol-Myers Squibb/Pfizer, Merck & Co, GlaxoSmithKline, and Roche Diagnostics;

lP

advisory board member for AstraZeneca, Bayer, and Boehringer Ingelheim. Wallentin: Institutional

na

research grants from AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb/Pfizer, GlaxoSmithKline, Roche Diagnostics, and Merck & Co; Consulting fees from Abbott; Holds 2 patents involving GDF-15

ur

licensed to Roche Diagnostics. Alexander: Research grants from Bristol-Myers Squibb, Boehringer

Jo

Ingelheim, CryoLife, CSL Behring, US FDA, NIH, Tenax Therapeutics, VoluMetrix; consulting fees from Bristol-Myers Squibb, Pfizer, Inc., AbbVie Pharmaceuticals, CSL Behring, Janssen, Novo Nordisk, Portola Pharmaceuticals, Teikoku Pharmaceuticals, VA Cooperative Studies, Zafgen. Lopes: Research grants from Bristol-Myers Squibb, Pfizer, Inc., Amgen, GlaxoSmithKline, Medtronic PLC, Sanofi Aventis; consulting fees from Pfizer, Inc., Boehringer Ingelheim.

Journal Pre-proof

11

REFERENCES 1

Beyth RJ, Quinn LM, Landefeld CS. Prospective evaluation of an index for predicting the risk of major bleeding in outpatients treated with warfarin. Am J Med. 1998;105: 91-9.

2

Shireman TI, Mahnken JD, Howard PA, Kresowik TF, Hou Q, Ellerbeck EF. Development of a contemporary bleeding risk model for elderly warfarin recipients. Chest. 2006;130: 1390-6.

3

Granger CB, Alexander JH, McMurray JJ, Lopes RD, Hylek EM, Hanna M, Al-Khalidi HR,

of

Ansell J, Atar D, Avezum A, Bahit MC, Diaz R, Easton JD, Ezekowitz JA, Flaker G, Garcia D, Geraldes M, Gersh BJ, Golitsyn S, Goto S, Hermosillo AG, Hohnloser SH, Horowitz J, Mohan P,

ro

Jansky P, Lewis BS, Lopez-Sendon JL, Pais P, Parkhomenko A, Verheugt FW, Zhu J, Wallentin

-p

L. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med. 2011;365: 981-92.

4

re

10.1056/NEJMoa1107039.

Lopes RD, Alexander JH, Al-Khatib SM, Ansell J, Diaz R, Easton JD, Gersh BJ, Granger CB,

lP

Hanna M, Horowitz J, Hylek EM, McMurray JJ, Verheugt FW, Wallentin L, Investigators A.

na

Apixaban for reduction in stroke and other ThromboemboLic events in atrial fibrillation (ARISTOTLE) trial: design and rationale. Am Heart J. 2010;159: 331-9.

De Caterina R, Andersson U, Alexander JH, Al-Khatib SM, Bahit MC, Goto S, Hanna M, Held

Jo

5

ur

10.1016/j.ahj.2009.07.035.

C, Hohnloser S, Hylek EM, Lanas F, Lopes RD, Lopez-Sendon J, Renda G, Horowitz J, Granger CB, Wallentin L, Investigators A. History of bleeding and outcomes with apixaban versus warfarin in patients with atrial fibrillation in the Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation trial. Am Heart J. 2016;175: 175-83. 10.1016/j.ahj.2016.01.005. 6

Miller CS, Dorreen A, Martel M, Huynh T, Barkun AN. Risk of Gastrointestinal Bleeding in Patients Taking Non-Vitamin K Antagonist Oral Anticoagulants: A Systematic Review and Meta-analysis. Clin Gastroenterol Hepatol. 2017;15: 1674-83.

Journal Pre-proof 7

12

Kirchhof P, Benussi S, Kotecha D, Ahlsson A, Atar D, Casadei B, Castella M, Diener HC, Heidbuchel H, Hendriks J, Hindricks G, Manolis AS, Oldgren J, Popescu BA, Schotten U, Van Putte B, Vardas P, Agewall S, Camm J, Baron Esquivias G, Budts W, Carerj S, Casselman F, Coca A, De Caterina R, Deftereos S, Dobrev D, Ferro JM, Filippatos G, Fitzsimons D, Gorenek B, Guenoun M, Hohnloser SH, Kolh P, Lip GY, Manolis A, McMurray J, Ponikowski P, Rosenhek R, Ruschitzka F, Savelieva I, Sharma S, Suwalski P, Tamargo JL, Taylor CJ, Van

of

Gelder IC, Voors AA, Windecker S, Zamorano JL, Zeppenfeld K. 2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS. Eur Heart J. 2016;37:

Lopes RD, Guimaraes PO, Kolls BJ, Wojdyla DM, Bushnell CD, Hanna M, Easton JD, Thomas

-p

8

ro

2893-962. 10.1093/eurheartj/ehw210.

re

L, Wallentin L, Al-Khatib SM, Held C, Gabriel Melo de Barros ESP, Alexander JH, Granger CB, Diener HC. Intracranial hemorrhage in patients with atrial fibrillation receiving anticoagulation

Kucher N, Connolly S, Beckman JA, Cheng LH, Tsilimingras KV, Fanikos J, Goldhaber SZ.

na

9

lP

therapy. Blood. 2017;129: 2980-7. 10.1182/blood-2016-08-731638.

International normalized ratio increase before warfarin-associated hemorrhage: brief and subtle.

Palareti G, Leali N, Coccheri S, Poggi M, Manotti C, D'Angelo A, Pengo V, Erba N, Moia M,

Jo

10

ur

Arch Intern Med. 2004;164: 2176-9. 10.1001/archinte.164.19.2176.

Ciavarella N, Devoto G, Berrettini M, Musolesi S. Bleeding complications of oral anticoagulant treatment: an inception-cohort, prospective collaborative study (ISCOAT). Italian Study on Complications of Oral Anticoagulant Therapy. Lancet. 1996;348: 423-8. 11

Chai-Adisaksopha C, Hillis C, Isayama T, Lim W, Iorio A, Crowther M. Mortality outcomes in patients receiving direct oral anticoagulants: a systematic review and meta-analysis of randomized controlled trials. J Thromb Haemost. 2015;13: 2012-20.

Journal Pre-proof Figure Legends Figure 1. 30-day rate of death for apixaban and warfarin groups following major GI bleed (dashed lines

Jo

ur

na

lP

re

-p

ro

of

define 95% confidence bands)

13

Journal Pre-proof

14

Table 1. Patient characteristics at study entry Prior GI Bleed Any (n=784)

Lower GI (n=321)

Upper GI (n=463)

73.5 (67.0, 78.0)

74.0 (67.0, 79.0)

73.0 (67.0, 78.0)

70.0 (62.0, 76.0)

<.0001

205 (26.1%)

91(28.3%)

114(24.6%)

6211(35.7%)

<.0001

Characteristic Age, yrs Female sex, no. (%)

313 (39.9%)

151 (47.0%)

Latin America

134 (17.1%)

41 (12.8%)

Europe

240 (30.6%)

86 (26.8%)

Asia Pacific

97 (12.4%)

Systolic BP, mm Hg

l a

43 (13.4%)

162 (35.0%)

4160 (23.9%)

93 (20.1%)

3332 (19.1%)

154 (33.3%)

7102 (40.8%)

54 (11.7%)

2819 (16.2%)

e

r P

f o

o r p

Region, no. (%) North America

No Prior GI Bleed (n=17,413) P-Value*

130 (120.0, 140.0) 130 (118.0, 140.0) 130 (120.0, 140.0) 130 (120.0, 140.0)

rn

<.0001

0.102

Diastolic BP, mm HG

80.0 (70.0, 85.0)

78.0 (70.0, 84.0)

80.0 (70.0, 86.0)

80.0 (70.0, 87.0)

<.0001

Weight, kg

83.0 (70.5, 97.0)

84.0 (72.0, 97.3)

82.0 (70.0, 95.5)

82.0 (70.0, 95.3)

0.126

68.8 (51.7, 88.2)

68.2 (52.3, 88.8)

74.2 (56.8, 95.4)

<.0001

Calculated CrCL, mL/min Renal function, no. (%)

u o

J

68.4 (51.9, 88.8)

<.0001

Normal

268 (34.2%)

109 (34.0%)

159 (34.3%)

7250 (41.6%)

Mild impairment (>50–80 mL/min)

334 (42.6%)

137 (42.7%)

197 (42.5%)

7251 (41.6%)

Moderate impairment (>30–50 mL/min)

166 (21.2%)

68 (21.2%)

98 (21.2%)

2581 (14.8%)

15 (1.9%)

7 (2.2%)

8 (1.7%)

255 (1.5%)

Severe impairment (≤30 mL/min)

Journal Pre-proof

15

Prior GI Bleed Any (n=784)

Lower GI (n=321)

Upper GI (n=463)

History of dyspepsia, no. (%)

130 (16.6%)

48 (15.0%)

82 (17.7%)

1244 (7.1%)

<.0001

History of gastroesophageal reflux, no. (%)

183 (23.3%)

75 (23.4%)

108 (23.3%)

1733 (10.0%)

<.0001

History of peptic ulcer disease, no. (%)

326 (41.6%)

40 (12.5%)

286 (61.8%)

881 (5.1%)

<.0001

History of gastrointestinal surgery, no. (%)

203 (25.9%)

82 (25.5%)

121 (26.1%)

1495 (8.6%)

<.0001

30 (3.8%)

8 (2.5%)

22 (4.8%)

481 (2.8%)

0.036

Prior stroke, TIA, or SE, no. (%)

163 (20.8%)

56 (17.4%)

107 (23.1%)

3374 (19.4%)

0.089

Diabetes, no. (%)

218 (27.8%)

92 (28.7%)

126 (27.2%)

4328 (24.9%)

0.157

HF or reduced LVEF, no. (%)

277 (35.3%)

111 (34.6%)

166 (35.9%)

6173 (35.5%)

0.933

CAD, no. (%)

310 (39.5%)

133 (41.4%)

177 (38.2%)

5731 (32.9%)

0.000

PAD, no. (%)

66 (8.4%)

26 (8.1%)

40 (8.6%)

818 (4.7%)

<.0001

71 (22.1%)

137 (29.6%)

1037 (6.0%)

<.0001

93 (29.0%)

128 (27.6%)

5410 (31.1%)

0.216

Characteristic

History of chronic liver disease, no (%)

History of anemia, no. (%) Aspirin use, no. (%) Randomized Treatment

l a

rn

u o

208 (26.5%)

J

221 (28.2%)

f o

o r p

e

r P

No Prior GI Bleed (n=17,413) P-Value*

0.803

Warfarin

394 (50.3%)

157 (48.9%)

237 (51.2%)

8684 (49.9%)

Apixaban

390 (49.7%)

164 (51.1%)

226 (48.8%)

8729 (50.1%)

Values are median (25th, 75th percentile), unless otherwise indicated. *P-value compares history of GI bleed and location groups (lower GI vs. upper GI vs. no GI bleed). BP indicates blood pressure; CAD, coronary artery disease; CrCl, creatinine clearance; GI, gastrointestinal; HF, heart failure; LVEF, left ventricular ejection fraction; PAD, peripheral artery disease; SE, systemic embolism; TIA, transient ischemic attack.

Journal Pre-proof

16

Table 2. Association between history of upper or lower GI bleeding and efficacy or safety endpoints Prior Lower GI Bleed

Prior Upper GI Bleed

No Prior GI Bleed

Event Rate*

Event Rate*

Event Rate*

(Count)

(Count)

(Count)

(n=321)

HR (95% CI)†

(n=463)

HR (95% CI)†

Stroke or SE

0.50 (3)

0.34 (0.11–1.06)‡

1.60 (13)

1.01 (0.58–1.76)‡

Ischemic/uncertain type stroke

0.33 (2)

0.31 (0.08–1.26)‡

0.98 (8)

0.87 (0.43–1.76)‡

Hemorrhagic stroke

0.17 (1)

0.54 (0.08–3.91)‡

Death from any cause

5.23 (32)

1.07 (0.75–1.53)§

Cardiovascular death

3.10 (19)

1.37 (0.86–2.20)§

Myocardial infarction

1.67 (10)

Efficacy Endpoints

Jo

-p

e r P

1.45 (461) 1.02 (327)

0.49 (4)

1.50 (0.55–4.10)‡

0.35 (113)

5.64 (47)

1.11 (0.82–1.51)§

3.65 (1193)

2.76 (23)

1.06 (0.68–1.67)§

1.87 (610)

1.96 (1.02–3.77)||

0.61 (5)

0.85 (0.35–2.07)||

0.55 (177)

1.72 (0.86–3.42)¶

3.23 (24)

3.13 (1.97–4.96)¶

0.65 (190)

l a

n r u

Safety Endpoints

ro

f o

(n=17,413)

Major GI bleeding

1.66 (9)

Upper

0.00 (0)

---

2.41 (18)

3.42 (2.02–5.81)¶

0.45 (132)

Lower

1.66 (9)

5.47 (2.53–11.80)¶

0.80 (6)

2.42 (0.93–6.33)¶

0.20 (60)

Major bleeding

4.17 (22)

1.21 (0.78–1.88)¶

6.55 (48)

2.06 (1.52–2.79)¶

2.47 (718)

Major or CRNM bleeding

8.92 (45)

1.49 (1.10–2.02)¶

10.16 (72)

1.77 (1.38–2.26)¶

4.82 (1372)

GUSTO severe bleeding

0.73 (4)

0.79 (0.29–2.14)¶

1.59 (12)

1.66 (0.92–3.01)¶

0.80 (236)

GUSTO moderate/severe bleeding

2.41 (13)

1.01 (0.57–1.76)¶

4.44 (33)

1.924 (1.33–2.76)¶

1.64 (480)

Journal Pre-proof Prior Lower GI Bleed

17

Prior Upper GI Bleed

No Prior GI Bleed

Event Rate*

Event Rate*

Event Rate*

(Count)

(Count)

(Count)

(n=321)

HR (95% CI)†

(n=463)

HR (95% CI)†

GUSTO mild bleeding

14.26 (68)

1.55 (1.21–1.98)¶

12.70 (86)

1.408 (1.13–1.76)¶

ISTH minor bleeding

9.28 (47)

1.51 (1.12–2.04)¶

7.62 (54)

1.276 (0.97–1.68)¶

Intracranial bleeding

0.18 (1)

0.31 (0.04–2.25)¶

0.93 (7)

1.555 (0.72–3.36)¶

(n=17,413)

p e

ro

f o

8.53 (2315) 5.71 (1588) 0.56 (166)

*Rates are per 100 patient-years. † Hazard ratios are versus no GI bleed. ‡ Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, weight, diabetes, hypertension, moderate valvular disease, prior stroke/TIA/SE, type of AF, prior VKA. § Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, sex, weight, systolic blood pressure, diastolic blood pressure, weight, hypertension, moderate valvular disease, left bundle branch block, prior MI, prior stroke/TIA/SE, anemia, smoking, prior VKA, NYHA class, CHADS2 score, renal function. || Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, diabetes, CAD, prior MI, NYHA class, renal function. ¶ Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, sex, CAD, prior MI, prior non-GI bleeding, anemia, CHADS2 score, renal function. CI indicates confidence interval; CRNM, clinically relevant non-major; GI, gastrointestinal; GUSTO, Global Use of Strategies to Open Occluded Arteries; HR, hazard ratio; ISTH, International Society on Thrombosis and Haemostasis; SE, systemic embolism.

l a

Jo

n r u

r P

Journal Pre-proof

18

Table 3. Association between history of recent* upper or lower GI bleeding and efficacy or safety endpoints No Recent Prior GI Prior Recent Lower GI Bleed

Prior Recent Upper GI Bleed

Event Rate†

Event Rate† HR (95% CI)‡

(Count)

(n=118)

0.56 (1)

0.35 (0.05–2.47)§

Ischemic/uncertain type stroke

0.56 (1)

0.48 (0.07–3.41)§

Hemorrhagic stroke

0.00 (0)

---

Death from any cause

5.02 (9)

0.99 (0.51–1.92)||

Cardiovascular death

2.79 (5)

Myocardial infarction

0.00 (0)

Safety Endpoints

e

HR (95% CI)‡

(Count) (n=17,982)

1.48 (3)

0.92 (0.30–2.86)§

1.44 (473)

0.98 (2)

0.87 (0.22–3.49)§

1.01 (334)

0.49 (1)

1.68 (0.23–12.17)§

0.35 (117)

6.80 (14)

1.51 (0.87–2.63)||

3.70 (1249)

1.11 (0.46–2.70)||

2.43 (5)

0.95 (0.35–2.56)||

1.90 (642)

---

0.98 (2)

1.39 (0.34–5.61)¶

0.57 (190)

l a

rn

u o

f o

o r p

Efficacy Endpoints Stroke or SE

Event Rate†

(Count)

(n=97)

Bleed

r P

Major GI bleeding

2.58 (4)

6.25 (11)

5.16 (2.66–10.00)#

0.69 (208)

Upper GI bleeding

J

2.58 (0.95–7.01)#

0.00 (0)

---

5.10 (9)

6.65 (3.26–13.57)#

0.47 (141)

Lower GI bleeding

2.58 (4)

7.73 (2.73–21.87)#

1.09 (2)

1.68 (0.23–12.47)#

0.23 (69)

Major bleeding

4.60 (7)

1.42 (0.68–3.02)#

8.56 (15)

2.51 (1.47–4.30)#

2.56 (766)

Major or CRNM bleeding

11.96 (17)

2.04 (1.26–3.30)#

11.76 (20)

1.87 (1.18–2.97)#

4.95 (1452)

GUSTO severe bleeding

1.27 (2)

1.44 (0.36–5.81)#

2.75 (5)

2.90 (1.18–7.17)#

0.81 (245)

Journal Pre-proof

19

No Recent Prior GI Prior Recent Lower GI Bleed

Prior Recent Upper GI Bleed

Event Rate†

Event Rate† HR (95% CI)‡

(Count)

GUSTO moderate/severe bleeding

Event Rate† HR (95% CI)‡

(Count)

(n=97)

f o

(n=118) 1.13 (0.42–3.03)#

2.57 (4)

6.81 (12)

Bleed

o r p

(Count) (n=17,982)

2.80 (1.56–5.03)#

1.69 (510)

*Recent is defined as less than 1 year prior to randomization. † Rates are per 100 patient-years. ‡Hazard ratios are versus no GI bleed. § Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, weight, diabetes, hypertension, moderate valvular disease, prior stroke/TIA/SE, type of AF, prior VKA. || Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, sex, weight, systolic blood pressure, diastolic blood pressure, weight, hypertension, moderate valvular disease, left bundle branch block, prior MI, prior stroke/TIA/SE, anemia, smoking, prior VKA, NYHA class, CHADS2 score, renal function. ¶ Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, diabetes, CAD, prior MI, NYHA class, renal function. # Adjusted for randomized treatment assignment, age, geographic region, aspirin and NSAID use at baseline, sex, CAD, prior MI, prior non-GI bleeding, anemia, CHADS2 score, renal function. CI indicates confidence interval; CRNM, clinically relevant non-major; GI, gastrointestinal; GUSTO, Global Use of Strategies to Open Occluded Arteries; HR, hazard ratio; SE, systemic embolism.

l a

Jo

n r u

r P

e

Journal Pre-proof

20

Table 4. Association between treatment effect and efficacy and safety endpoints by history of GI bleed/location Prior Lower GI Bleed

Prior Upper GI Bleed

No Prior GI Bleed

P-Value†

Apixaban

Warfarin

Apixaban

Warfarin

Apixaban

Warfarin

Rate*

Rate*

Rate*

Rate*

Rate*

Rate*

(n=164)

(n=157)

HR (95% CI)

(n=226)

(n=237)

(n=8729)

(n=8684)

Stroke or SE

0.66 (2)

0.33 (1)

2.01 (0.18 - 22.24)

1.02 (4)

2.15 (9)

0.47 (0.15 - 1.54)

1.28 (206)

1.61 (255)

0.80 (0.67 - 0.96) 0.5150

Death from any

4.22 (13)

6.25 (19)

0.66 (0.33 - 1.34)

5.43 (22)

5.84 (25)

o r p

0.93 (0.53 - 1.66)

3.46 (568)

3.84 (625)

0.90 (0.80 - 1.01) 0.7298

CV death

2.27 (7)

3.95 (12)

0.57 (0.22 - 1.44)

0.68 (0.29 - 1.57)

1.78 (292)

1.96 (318)

0.91 (0.78 - 1.07) 0.5165

MI

1.32 (4)

2.02 (6)

0.70 (0.20 - 2.51)

l a

3.27 (14) 0.71 (3)

0.71 (0.12 - 4.27)

0.52 (84)

0.58 (93)

0.89 (0.67 - 1.20) 0.8758

Major GI bleeding

1.79 (5)

1.51 (4)

1.17 (0.31 - 4.36)

3.04 (11)

3.41 (13)

0.88 (0.39 - 1.97)

0.59 (88)

0.70 (102)

0.84 (0.63 - 1.12) 0.8829

Major bleeding

4.45 (12)

3.87 (10)

1.13 (0.49 - 2.63)

6.13 (22)

6.96 (26)

0.88 (0.50 - 1.54)

1.99 (292)

2.97 (426)

0.67 (0.58 - 0.78) 0.3268

Major or CRNM

8.16 (21)

9.71 (24)

0.83 (0.46 - 1.49)

8.85 (31)

11.45 (41)

0.77 (0.48 - 1.22)

3.87 (560)

5.81 (812)

0.67 (0.60 - 0.75) 0.6683

0.36 (1)

0.00 (0)

---

1.08 (4)

0.77 (3)

1.39 (0.31 - 6.21)

0.32 (47)

0.82 (119)

0.39 (0.28 - 0.54) 0.2657

HR (95% CI)

f o

Efficacy Endpoints

cause

n r u

Safety Endpoints

Jo

2.22 (9) 0.51 (2)

e

r P

HR (95% CI)

bleeding Intracranial bleeding *Rates are per 100 patient-years. † P-value for the randomized treatment by history of GI bleed and location interaction. CI indicates confidence interval; CRNM, clinically relevant non-major; CV, cardiovascular; GI, gastrointestinal; HR, hazard ratio; MI, myocardial infarction; SE, systemic embolism.

Journal Pre-proof

21

Table 5. INR before and study drug discontinuation after first major GI bleed Apixaban

Warfarin

Overall

91 (86.7%)

104 (87.4%)

195 (87.0%)

105

119

224

104

119

223

1.3 (0.6)

2.8 (1.5)

2.1 (1.4)

1.2 (1.1, 1.4)

2.5 (2.1, 3.0)

1.7 (1.2, 2.7)

13 (5, 24)

11 (4, 22)

12 (4, 22)

68 (74.7%)

76 (73.1%)

144 (73.8%)

34 (50%)

35 (46.0%)

69 (47.9%)

Time from bleed to resuming study drug, median (25th, 75th), days

16 (6, 41)

17 (7, 51)

17 (7, 43)

Time at follow-up when bleed occurred, median (25th, 75th), days

352 (123, 582)

240 (101, 534)

280 (121, 562)

On study drug day before bleed Events (GI bleeds), no. INR, most recent before bleed N Mean (SD)

of

Median (25th, 75th) Time from INR to bleed, median (25th, 75th), days

ro

Interruption of randomized study drug*

re

-p

Study drug resumed

Jo

ur

na

lP

*Interruption was defined as discontinuation of the study drug for at least 1 day, either on the day of the bleed or within the 3 days following the bleed.  For this analysis, all GI bleeds were counted, including the 6 that occurred after efficacy censoring. GI indicates gastrointestinal; INR, international normalized ratio; SD, standard deviation.

Journal Pre-proof

Jo

ur

na

lP

re

-p

ro

of

Figure 1.

22

Figure 1