Journal Pre-proof

Left Atrial Appendage Closure versus Non-Warfarin Oral Anticoagulation in Atrial

Fibrillation: 4-Year Outcomes of PRAGUE-17

Pavel Osmancik, MD, PhD, Dalibor Herman, MD, PhD, Petr Neuzil, MD, CSc, Pavel
Hala, MD, Milos Taborsky, MD, CSc, Petr Kala, MD, PhD, Martin Poloczek, MD,
Josef Stasek, MD, PhD, Ludek Haman, MD, PhD, Marian Branny, MD, PhD, Jan
Chovancik, MD, Pavel Cervinka, MD, PhD, Jiri Holy, MD, Tomas Kovarnik, MD, PhD,
David Zemanek, MD, PhD, Stepan Havranek, MD, PhD, Vlastimil Vancura, MD, PhD,
Petr Peichl, MD, PhD, Petr Tousek, MD, PhD, Veronika Lekesova, MD, Jiri Jarkovsky,
RNDr, PhD, Martina Novackova, Mgr, Klara Benesova, Mgr, Petr Widimsky, MD,
DrSc, Vivek Y. Reddy, MD, on behalf of the PRAGUE-17 Trial Investigators

PII: S0735-1097(21)07895-5
DOI: https://doi.org/10.1016/j.jacc.2021.10.023
Reference: JAC 28999

To appear in: Journal of the American College of Cardiology

Received Date: 17 September 2021

Revised Date: 5 October 2021
Accepted Date: 14 October 2021

Please cite this article as: Osmancik P, Herman D, Neuzil P, Hala P, Taborsky M, Kala P, Poloczek M,
Stasek J, Haman L, Branny M, Chovancik J, Cervinka P, Holy J, Kovarnik T, Zemanek D, Havranek
S, Vancura V, Peichl P, Tousek P, Lekesova V, Jarkovsky J, Novackova M, Benesova K, Widimsky
P, Reddy VY, on behalf of the PRAGUE-17 Trial Investigators, Left Atrial Appendage Closure versus
Non-Warfarin Oral Anticoagulation in Atrial Fibrillation: 4-Year Outcomes of PRAGUE-17, Journal of the
American College of Cardiology (2021), doi: https://doi.org/10.1016/j.jacc.2021.10.023.

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.
© 2021 Published by Elsevier on behalf of the American College of Cardiology Foundation.
GAE: Faisal Merchant, MD

GEIC: Javed Butler, MD, PhD

Left Atrial Appendage Closure versus Non-Warfarin Oral Anticoagulation in Atrial

Fibrillation: 4-Year Outcomes of PRAGUE-17

Brief title: LAAC vs NOACs: Long-Term Results of PRAGUE-17

f
oo
Pavel Osmancik, MD, PhD,1 Dalibor Herman, MD, PhD,1 Petr Neuzil, MD, CSc,2 Pavel Hala,

r
MD,2 Milos Taborsky, MD, CSc,3 Petr Kala, MD, PhD,4 Martin Poloczek, MD,4 Josef Stasek,
-p
MD, PhD,5 Ludek Haman, MD, PhD,5 Marian Branny, MD, PhD,6 Jan Chovancik, MD,6 Pavel
re
Cervinka, MD, PhD,7 Jiri Holy, MD,7 Tomas Kovarnik, MD, PhD,8 David Zemanek, MD,
lP

PhD,8 Stepan Havranek, MD, PhD,8 Vlastimil Vancura, MD, PhD,9 , Petr Peichl, MD, PhD,10
na

Petr Tousek, MD, PhD,1 Veronika Lekesova, MD,2 Jiri Jarkovsky, RNDr, PhD,11 Martina

Novackova, Mgr.11, Klara Benesova, Mgr. 11

Petr Widimsky, MD, DrSc.1,*, and Vivek Y.
ur

Reddy, MD2,12,*, on behalf of the PRAGUE-17 Trial Investigators

Jo

1. Cardiocenter, Third Faculty of Medicine, Charles University Prague and University

Hospital Kralovske Vinohrady, Prague, Czech Republic

2. Cardiocenter, Department of Cardiology, Na Homolce Hospital, Prague, Czech

Republic

3. Cardiocenter, Dept. of Cardiology, University Hospital Olomouc, Czech Republic

4. Clinic of Cardiology, Masaryk University and University Hospital Brno, Brno, Czech

Republic

1
 5. 1st Department of Internal Medicine, Faculty of Medicine, University Hospital Hradec

Kralove, Charles University Prague, Czech Republic

6. Department of Cardiology, Cardiocenter, Hospital Podlesí a.s., Trinec, Czech Republic

7. Department of Cardiology, Krajská zdravotni a.s., Masaryk hospital and UJEP, Usti nad

Labem, Czech Republic

8. Cardiocenter, 2nd internal clinic – Cardiology and Angiology, Charles University,

General Faculty Hospital, Prague Czech Republic

9. Department of Cardiology, University Hospital and Faculty of Medicine Pilsen, Pilsen,

f
oo
Czech Republic

r
10. Cardiocenter, Institute of Clinical and Experimental Medicine, Prague, Czech Republic
-p
11. Institute of Biostatistics and Analysis, Faculty of Medicine, Masaryk University, Brno,
re
Czech Republic.
lP

12. Helmsley Electrophysiology Center, Icahn School of Medicine at Mount Sinai, New
na

York, New York, USA

ur

*Senior authors
Jo

Funding: Research grant from the Ministry of Health, Czech Republic (AZV 15-29565A).

Disclosures:

P.Osmancik reports occasional speaking honoraria from Bayer and Abbott. M. Taborsky´s

disclosures include Bayer and Pfizer (both advisory board). P.Kala´s disclosures with medical

companies include Bayer (speaker bureau, advisory board), Boston Scientific (advisory board,

consultancy). M.Poloczek reports speaking honoraria from Abbott. L.Haman reports speaking

honoraria from Pfizer. D.Zemanek reports speaking honoraria from Abbott and Bayer.

2
P.Peichl reports occasional speaking honoraria from Abbott, and consultancy from Abbott,

Biotronik and Medtronic. S. Havranek reports speaking honoraria and advisory board from

Boehringer Ingelheim. P.Widimsky reports occasional honoraria from Bayer, Pfizer, and

Boehringer Ingelheim. V.Reddy reports consulting income and grant support from Abbott Inc

and Boston Scientific Inc. Unrelated to this manuscript, he has also served as a consultant for

Ablacon, Acutus Medical, Affera, Apama Medical, APN Health, Aquaheart, Atacor,

Autonomix, Axon Therapies, Backbeat, BioSig, Biosense-Webster, BioTel Heart, Biotronik,

Cardiac Implants, CardiaCare, Cardiofocus, Cardionomic, CardioNXT / AFTx, Circa

f
oo
Scientific, CoreMap, Corvia Medical, Dinova-Hangzhou DiNovA EP Technology, East End

r
Medical, EBR, EPD, Epix Therapeutics, EpiEP, Eximo, Farapulse, Fire1, Gore & Associates,
-p
HRT, Impulse Dynamics, Intershunt, Javelin, Kardium, Keystone Heart, LuxMed,
re
Medlumics, Medtronic, Middlepeak, Nuvera, Philips, Pulse Biosciences, Sirona Medical, and
lP

Valcare Vizaramed; and owns equity in Ablacon, Acutus Medical, Affera, Apama Medical,
na

APN Health, Aquaheart, Atacor, Autonomix, Axon Therapies, Backbeat, BioSig, Cardiac

Implants, CardiaCare, CardioNXT / AFTx, Circa Scientific, Corvia Medical, Dinova-

ur

Hangzhou DiNovA EP Technology, East End Medical, EPD, Epix Therapeutics, EpiEP,
Jo

Eximo, Farapulse, Fire1, HRT, Intershunt, Javelin, Kardium, Keystone Heart, LuxMed,

Manual Surgical Sciences, Medlumics, Middlepeak, Newpace, Nuvera, Pulse Biosciences,

Sirona Medical, Surecor, Valcare, and Vizaramed. All other authors report no conflict of

interests.

Corresponding authors:

Pavel Osmancik, MD, PhD

Cardiocenter, Charles University Prague

Third Internal - Cardiology Clinic

3
Srobarova 50

10034 Prague

Czech Republic

Phone: 00420721544447

Fax: 00420267162817

Email: pavel.osmancik@gmail.com

Twitter handle: @POsmancik

f
oo
Vivek Y. Reddy, MD

r
Helmsley Electrophysiology Center

Icahn School of Medicine at Mount Sinai

-p
re
One Gustave L. Levy Place, Box 1030
lP

New York, NY 10029

na

Tel: +212-241-7114

Fax: +646-537-9691
ur

E-mail: vivek.reddy@mountsinai.org
Jo

Acknowledgements

We are grateful to the members of Data Safety and Monitoring Board, and Clinical Endpoint

Committee, as well as the whole team of IT specialists and statisticians, responsible for the

development of the randomization software and e-database, and all statistical analyses.

4
ABSTRACT

Background: The PRAGUE-17 trial demonstrated that left atrial appendage closure (LAAC)

was non-inferior to non-warfarin oral anticoagulants (NOAC) for preventing major

neurological, cardiovascular or bleeding events in high-risk patients with atrial fibrillation

(AF).

Objective: To assess the pre-specified long-term (4-year) outcomes in PRAGUE-17.

Methods: PRAGUE-17 was a randomized non-inferiority trial comparing percutaneous

LAAC (Watchman or Amulet) with NOACs (95% apixaban) in non-valvular AF patients with

f
oo
a history of cardioembolism, clinically-relevant bleeding, or both CHA2DS2-VASc > 3 and

r
HASBLED > 2. The primary endpoint was a composite of cardioembolic events (stroke,
-p
transient ischemic attack, or systemic embolism), cardiovascular death, clinically-relevant
re
bleeding, or procedure/device-related complications (LAAC group only). The primary
lP

analysis was modified intention-to-treat (mITT).

na

Results: We randomized 402 AF patients (201 per group, age 73.3±7.0 years, 65.7% male,

CHA2DS2-VASc 4.7+1.5, HASBLED 3.1+0.9). After 3.5 years median follow-up (1,354
ur

patients-years), LAAC was non-inferior to NOAC for the primary endpoint by mITT
Jo

(subdistribution hazard ratio[sHR] 0.81, 95% CI 0.56-1.18; p=0.27; p for non-

inferiority=0.006). For the components of the composite endpoint, the corresponding sHRs

(and 95% CIs) were 0.68 (0.39-1.20; p=0.19) for cardiovascular death, 1.14 (0.56-2.30;

p=0.72) for all-stroke/TIA, 0.75 (0.44-1.27; p=0.28) for clinically-relevant bleeding, and 0.55

(0.31-0.97; p=0.039) for non-procedural clinically-relevant bleeding. The primary endpoint

outcomes were similar in the per-protocol [sHR 0.80 (95% CI 0.54-1.18), p=0.25] and on-

treatment [sHR 0.82 (95% CI 0.56-1.20), p=0.30] analyses.

5
Conclusion: In long-term follow-up of PRAGUE-17, LAAC remains non-inferior to NOACs

for preventing major cardiovascular, neurological or bleeding events. Furthermore, non-

procedural bleeding was significantly reduced with LAAC.

CONDENSED ABSTRACT

We report the long-term (4-year) results of PRAGUE-17 – a randomized trial comparing

percutaneous LAAC with NOAC treatment in high-risk AF patients (201 per group,

CHA2DS2-VASc 4.7+1.5, HASBLED 3.1+0.9) followed for a median 3.5 years. For the

f
oo
primary outcome (composite of stroke, TIA, systemic embolism, cardiovascular death,

r
clinically-relevant bleeding and procedural complications), LAAC was non-inferior to
-p
NOACs (sHR 0.81, 95% CI 0.56-1.18; p=0.27; p for non-inferiority=0.006). Non-procedural
re
clinically-relevant bleeding was reduced with LAAC (sHR 0.55, 95% CI 0.31-0.97; p=0.039).
lP

Thus, LAAC was non-inferior to NOAC for major AF-related cardiovascular or neurological
na

events, and was associated with less post-procedural bleeding.

ur

Key words: Atrial fibrillation, oral anticoagulation, left atrial appendage closure,
Jo

cardioembolism, non-vitamin K anticoagulant

LIST OF ABBREVIATIONS

LAAC: Left atrial appendage closure

AF: atrial fibrillation

NOAC: non-vitamin K oral anticoagulant

CHA2DS2-VASc: congestive heart failure, hypertension, age ≥ 65 or ≥ 75, diabetes, prior

cardioembolic event, female gender, or vascular disease

6
HAS-BLED: uncontrolled hypertension (i.e. systolic blood pressure > 160 mm Hg despite

antihypertensive treatment), prior bleeding, age > 75, prior cardioembolic event,

abnormal liver or renal function, and labile INR

TEE: transesophageal echocardiography

DAPT: dual antiplatelet treatment

TIA: transient ischemic attack

ISTH: International Society on Thrombosis and Hemostasis

ITT: intention-to-treat

f
oo
CIF: Cumulative incidence functions

r
sHR: subdistribution hazard ratio

CV: cardiovascular
-p
re
SE: systemic embolism
lP
na

Clinical Trial: www.clinicaltrials.gov NCT02426944

ur
Jo

7
INTRODUCTION

Left atrial appendage closure (LAAC) is a non-pharmacological option for preventing

cardioembolic events in patients with atrial fibrillation (AF) at significant stroke risk.

However, long-term results from randomized clinical trials are sparse; indeed, long-term

results are only available from two randomized studies comparing LAAC using the

Watchman device with warfarin. (1) In these reports, LAAC was associated with lower rates

of non-procedure related bleeding. However, since their introduction into clinical practice,

non-vitamin K oral anticoagulants (NOACs) have largely replaced warfarin. Since treatment

f
oo
using NOACs is associated with less bleeding (including intracranial hemorrhage) than

r
warfarin, (2) the potential benefit of LAAC relative to NOAC is unclear, prompting the

PRAGUE-17 trial.
-p
re
In PRAGUE-17, we compared LAAC with NOAC treatment for the incidence of the
lP

composite primary endpoint of cardioembolism, cardiovascular (CV) death, clinically-

na

relevant bleeding, and procedure or device associated complications. (3) In the primary

analysis, at a median follow-up of 20 months, the incidence of the primary endpoint was
ur

similar between groups. Since the trial was not powered to identify differences in the
Jo

individual components of the primary composite endpoint, there were no statistically-

significant differences in their incidence, including the incidence of non-procedural clinically-

relevant bleeding. However, the maximum benefit of LAAC, particularly in terms of

bleeding, is expected to be long-term, i.e., after patients have ceased (N)OAC and/or dual

antiplatelet treatment. Accordingly, short-term results have the potential for a Type II error –

a potential to miss significant LAAC benefits on bleeding. Conversely, if there are late LAAC

events, the relatively short follow-up in the trial also raises a potential for a Type I error, ie,

inappropriately concluding non-inferiority for the primary composite endpoint. But as

prespecified by protocol, patients in PRAGUE-17 continued follow-up beyond the timepoint

8
of the initial analysis. Herein, we report the clinical outcomes after 4 years of follow-up of the

PRAGUE-17 trial population.

METHODS

Study design

The protocol and results of the PRAGUE-17 Trial has been described previously. (4)

(3) In brief, the Left Atrial Appendage Closure versus non-vitamin K Anticoagulants in High

Risk Patients with Atrial Fibrillation trial (PRAGUE-17; NCT02426944) was an investigator-

initiated, multicenter, prospective, open-label, randomized, non-inferiority trial conducted at

f
oo
10 cardiac centers in the Czech Republic. Patient enrollment began in October 2015 and

r
concluded in January 2019. The trial was approved by the multicenter ethics committee at
-p
University Hospital Kralovske Vinohrady (Approval No. EK-VP/29/0/2014) and by the local
re
ethics committees at participating centers.
lP

Criteria for eligibility

na

Patients with non-valvular AF, and at moderate or high risk for stroke or bleeding

were eligible. Inclusion criteria were AF and either: a) history of bleeding requiring
ur

intervention or hospitalization, b) history of cardioembolism while taking anticoagulation, or

Jo

c) a moderate to high risk profile, defined as CHA2DS2-VASc ≥ 3 plus HAS-BLED ≥ 2. Key

exclusion criteria include mechanical valve prosthesis, mitral stenosis, comorbidities other

than AF mandating anticoagulation, patent foramen ovale with large atrial septal aneurysm,

mobile aortic plaque, symptomatic carotid arterial atherosclerosis, clinically significant

bleeding within 30 days, cardioembolic event within 30 days, and creatinine clearance < 30

ml/min. If randomized to LAAC, transesophageal echocardiography (TEE) was performed to

exclude left atrial (LA) thrombi. Consistent with clinical practice, the protocol only mandated

TEE in the LAAC group, and not before NOAC initiation. The presence of a thrombus in the

left atrial appendage (LAA) or left atrium was a pre-specified additional exclusion criterion.

9
Patients were randomly assigned to LAAC or NOAC in a 1:1 ratio using a centralized

computer system, and stratified by center to ensure comparable CHA2DS2-VASc scores

between groups.

Treatments and follow-up

Patients randomized to NOAC therapy could receive either rivaroxaban, apixaban, or

dabigatran (the approved NOACs at the time of the trial) at the manufacturer-recommended

dose. Investigators were instructed to reserve crossover from NOAC to LAAC for cases of

bleeding while taking the prescribed NOAC. Patients randomized to LAAC underwent

f
oo
implantation using a commercially-available Amulet (Abbott Inc, St. Paul, MN) or Watchman

r
/ Watchman-FLX (Boston Scientific Inc, St. Paul, MN) device. Device selection was at the
-p
discretion of the implanting center. After LAAC, the recommended anti-thrombotic regimen
re
was aspirin 100 mg/day plus clopidogrel 75 mg/day for three months. If a TEE then
lP

demonstrated no device-related thrombus or leak > 5 mm, clopidogrel was withdrawn; aspirin
na

was continued indefinitely. Based on patient characteristics and device type, this post-implant

anti-thrombotic regimen could be individualized, and was ultimately left to physician

ur

discretion. In patients at high risk for bleeding, dual antiplatelet treatment (DAPT) could be
Jo

shortened to 6 weeks. Conversely, in patients with a very high thrombotic risk, alternative

regimens included NOAC substitution for DAPT for up to three months, or NOACs for 6

weeks followed by DAPT for 6 weeks. For both groups, outpatient follow-up occurred at 6

weeks and 3, 6, 9, and 12 months, and every 6 months thereafter.

Study Outcomes

Because the risks associated with each treatment strategy are significantly different,

the primary endpoint was a composite of safety and efficacy characteristics of both strategies:

(1) stroke (ischemic or hemorrhagic) or transient ischemic attack (TIA), (2) systemic

embolism (SE), (3) clinically-significant bleeding, 4) CV death, or (5) a significant peri-

10
procedural or device-related complications. Clinically-significant bleeding was a composite of

major and non-major clinically-relevant bleeding according to the International Society on

Thrombosis and Hemostasis (ISTH) criteria. Major bleeding includes either a decrease in

hemoglobin ≥ 2.0 g/dL over a 24-hour period, transfusion of ≥ 2 units of packed red cells,

bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, intraarticular,

intramuscular with compartment syndrome or retroperitoneal), or fatal bleeding. Non-major

clinically-relevant bleeding is defined as bleeding requiring hospitalization or an invasive

procedure, but not meeting ISTH major criteria.

f
oo
Statistical design and analysis

r
The primary hypothesis was that LAAC would be noninferior to NOACs for the
-p
primary endpoint. The primary analysis was prespecified to be performed on a modified
re
intention-to-treat (mITT) basis, including all randomized patients without an LAA thrombus
lP

by TEE. As previously described, (3), 13% and 10% of the NOAC and LAAC cohorts,
na

respectively, were estimated to experience the primary endpoint annually. A minimum of 396

study subjects provided 80% power at an alpha level of 0.05, for a non-inferiority margin, of
ur

5% (or 1.47, expressed as a hazard ratio). This margin is concordant with the FDA
Jo

guidance. The power analysis was computed using PASS 13 software (2014, NCSS, LLC.

Kaysville, Utah; ncss.com/software/pass). Since ITT outcomes can potentially bias non-

inferiority trials towards the null hypothesis, post hoc secondary per-protocol and on-

treatment analyses were also performed. Cumulative incidence functions (CIFs) and Fine-

Gray competing risk regression models were adopted for data visualization and

description. As previously described, primary analyses were conducted after adjusting for the

competing risk of mortality. For other data, standard descriptive statistical methods were

used: absolute and relative frequencies for categorical data and the median with interquartile

range or mean with standard deviation for continuous data. The influence of patient

11
characteristics on the occurrence of endpoints was calculated using the Fine-Gray regression

models with the study group as a covariate, and is displayed as subdistribution hazard ratio

(sHR). The sHR for the primary endpoint was compared to the non-inferiority margin of 5%

evaluated for 4-year cumulative incidence using Wald statistic. Statistical analyses were

performed using SPSS v. 25.0 software (IBM Corporation, 2013) and cmprsk (2.2-10)

package in R (version 4.0.0).

RESULTS

Between October 2015 and January 2019, 415 patients were enrolled at 10 centers.

f
oo
(Online Figure S1) Thirteen patients were excluded, 8 withdrew informed consent, and 5 had

r
LAA thrombus on pre-procedure TEE imaging. Ultimately, 201 patients were randomized per
-p
group, and followed for a median of 3.5 years (interquartile range 2.6 – 4.3) in the LAAC
re
group, and 3.5 years (interquartile range, 2.6 – 4.2) in the NOAC group, for an aggregate of
lP

1,354 patients-years. Baseline clinical characteristics are shown in the Table 1 and Online
na

Table S1. The mean age was 73.3 ± 7.0 years, 34.3% were women, mean CHA2DS2-VASc =

4.7±1.5 (over 25% being CHA2DS2-VASc > 6), prior cardioembolism in 35.3%, and prior
ur

clinically-relevant bleeding in 47.8% of patients.

Jo

As previously published, (3) 7.0% (14 of 201) did not undergo the procedure because

of either patient refusal (n=9) or anatomical considerations. All 14 patients continued follow-

up, and 12 crossed over to NOAC. Ultimately, 187 patients underwent LAAC, with the LAA

being successfully occluded in 96.8% of procedure attempts (181 of 187). Significant

complications occurred in 9 patients including one device embolization with a need for

surgical exctraction and surgical LAA closure. Five patients with unsuccessful closure

(including the aforementioned patient) also crossed over to NOAC. The implanted devices

were either Amulet, Watchman or Watchman-FLX in 61.3%, 35.9% or 2.8%, respectively.

Most patients (148, 81.8%) received DAPT upon discharge, 25 patients (13.8%) received

12
apixaban for three months followed by aspirin, and 8 patients (4.4%) received apixaban for

six weeks followed by DAPT for 6 weeks. In the NOAC group, the most frequently used

anticoagulant was apixaban, in 192 patients (95.5%): 5 and 2.5 mg twice daily in 159 (79.1%)

and 33 (16.4%) patients, respectively. Dabigatran and rivaroxaban were used in 8 (4.0%) and

1 (0.5%) patients, respectively. In 17 LAAC patients, long-term NOAC was initiated at some

point during the study period (in 5 patients because of ischemic stroke/TIA). Conversely,

NOAC treatment was permanently stopped in 26 NOAC patients during follow-up (in 15 due

to clinically-relevant bleeding) with (n=13) or without subsequent crossover to the LAAC (for

f
oo
details see the Online Appendix).

r
Primary and secondary endpoints in the mITT analysis
-p
The primary outcome occurred in 49 patients (a total of 58 events) with LAAC (8.6
re
events per 100 patient-years) compared to 60 patients (81 events) with NOAC (11.9 events
lP

per 100 patient-years; sHR = 0.81, 95% CI 0.56 – 1.18, p=0.27; Central Illustration
na

&Figure 1, Table 2). Also in the present long-term analysis, LAAC remained non-inferior to

NOAC treatment (one-sided p value for non-inferiority = 0.006). The result was consistent
ur

across all subgroups with no statistically-significant interactions. (Figure 2)

Jo

Death from cardiovascular reasons occurred in 20 patients in the LAAC Arm (3.0

events per 100 patient-years) and 30 patients in the NOAC Arm (4.4 events per 100 patient-

years; sHR 0.68, 95% CI 0.39 – 1.20, p=0.19; Figure 3, Table 2). Non-cardiovascular

mortality (sHR = 0.99, 95% CI 0.55 – 1.77, p=0.96) and all-cause mortality (HR = 0.81, 95%

CI 0.54 – 1.22, p=0.31) were similar in both groups (Online Figure S2 and S3).

All-stroke/TIA occurred in 16 patients with LAAC (16 events, 15 ischemic and 1

hemorrhagic events; 14 of them were strokes) and in 15 patients with NOAC (18 events, 16

ischemic and 2 hemorrhagic events; 12 strokes). The corresponding annualized rate of all-

stroke/TIA was 2.4% with LAAC and 2.7% with NOAC (sHR = 1.14, 95% CI 0.56 – 2.30,

13
p=0.72; Figure 3 & Central Illustration, Table 2). When TIA were excluded, the

annualized rate for stroke was 2.1% with LAAC, and 1.8% with NOAC (sHR 1.38, 95% CI

0.63-3.03; p=0.42). Systemic embolism occurred in only 1 patient, who was in the NOAC

Arm.

Clinically-relevant bleeding occurred in 24 patients with LAAC (29 events), and in 32

patients with NOAC (40 events). The corresponding annualized rate of clinically-relevant

bleeding was 4.3% with LAAC and 5.9% with NOAC (sHR 0.75, 95% CI 0.44 – 1.27, p =

0.28, Figure 3, Table 2). However, six bleeding events in the LAAC Arm were procedure-

f
oo
related. Accordingly, the annualized incidence of non-procedural clinically-relevant bleeding

r
was significantly different between the groups: 3.4% with LAAC and 5.9% with NOAC (sHR
-p
0.55, 95% CI 0.31-0.97, p = 0.039; Figure 3 & Central Illustration, Table 2).
re
Per Protocol Analysis
lP

In the post hoc per protocol analysis, 181 and 199 patients were included in the LAAC
na

and NOAC groups, respectively (details of patient assignment and censoring are noted in the

Online Appendix). LAAC was non-inferior to NOAC for the primary endpoint outcome
ur

(sHR = 0.80, 95%CI 0.54-1.18, p=0.25; one-sided p value for non-inferiority = 0.020, Figure
Jo

4). There were also no significant differences between groups for all-stroke/TIA (sHR 1.01,

95% CI 0.48-2.11, p=0.99), all-stroke (sHR 1.30, 95% CI 0.57-2.93, p=0.53), or

cardiovascular mortality (sHR 0.70, 95% CI 0.38-1.31, p=0.27; Online Figure S4, Table S2).

The rate of clinically-relevant bleeding was not significantly different between groups (sHR

0.76, 95% CI 0.44-1.31, p=0.32); however, for non-procedural- and device-related events,

LAAC was again associated with a significantly reduced rate of clinically-relevant bleeding

(sHR 0.52, 95% CI 0.29-0.97, p=0.039).

On-Treatment Analysis

14
 The post hoc on-treatment analysis ultimately included 184 and 216 patients in the

LAAC and NOAC groups, respectively (details of patient assignment and censoring are noted

in the Online Appendix). Again, LAAC was non-inferior to NOAC for the primary endpoint

outcome (sHR = 0.82, 95% CI 0.56-1.20, p=0.30; one-sided p value for non-inferiority =

0.025), Figure 4). There were also no significant differences between groups for all-

stroke/TIA (sHR 0.86, 95% CI 0.42-1.75, p=0.68), all-stroke (sHR 1.12, 95% CI 0.52-2.45,

p=0.77), or cardiovascular mortality (sHR 0.70, 95% CI 0.38-1.30, p=0.26; Online Figure

S5, Table S3). And again, the rate of clinically-relevant bleeding was similar between groups

f
oo
(sHR = 0.78, 95% CI 0.45-1.35, p=0.38), but non-procedural clinically-relevant bleeding was

r
significantly reduced with LAAC (sHR 0.54, 95% CI 0.30-1.00, p=0.049).

DISCUSSION
-p
re
In this long-term analysis of the PRAGUE-17 trial, non-inferiority of LAAC vs
lP

NOAC was maintained for the primary composite endpoint after ~4 years of follow-up. The
na

results were similar in both the pre-specified mITT analysis, and the post hoc per protocol &

on-treatment analyses. There were no significant differences in the rates of all-stroke/TIA or

ur

CV death between groups. On the other hand, in this extended follow-up, the incidence of
Jo

non-procedural clinically-relevant bleeding was significantly reduced with LAAC in all

presented analyses.

In our study, the benefit of LAAC on clinically-relevant bleeding was derived from a

reduction in late events. Events during the 1st year post-implant are primarily driven by

procedural complications and postimplant antithrombotic medications, so the benefit of

LAAC was best appreciated over longer periods of follow-up. These findings are in

agreement with other randomized trials of LAAC in AF patients. The 5-year outcomes of the

PROTECT-AF and PREVAIL trials showed that the composite of stroke, SE, and CV death

occurred at similar frequencies in the LAAC and warfarin groups (HR 0.82, 95% CI 0.58–

15
1.17); similarly, the rate of all-stroke or SE was also not different between groups (HR 0.96,

95% CI 0.60–1.54). However, LAAC patients experienced substantially fewer non-procedural

major bleeding events than warfarin patients (HR 0.48, 95% CI 0.32–0.71).(1)

After the initial publication of the PRAGUE-17 trial, several non-randomized registries

and studies with propensity score-matched controls have been published – all of which further

corroborate the embolic role of the LAA in AF patients. The largest of these propensity-

matching studies compared the 2-year clinical follow-up of 1,088 patients after LAAC with a

propensity score-matched cohort of 1,184 AF patients treated using NOAC. The risk of

f
oo
ischemic stroke was similar between groups (HR 1.11), while the risk of major bleeding (HR

r
0.62, 95% CI 0.49–0.79) and even all-cause mortality (HR 0.53, 95% CI 0.43–0.64) were
-p
significantly reduced with LAAC. (5) In the APPLY study, 500 patients after LAAC were
re
compared with 500 medically treated AF patients (78.8% on OAC), also using propensity
lP

score-matching. (6) After a mean follow-up of 2.7 years, the incidence of the primary
na

composite endpoint of stroke and CV death were significantly lower in the LAAC group (HR

0.7, 95% CI 0.53–0.95). The primary safety endpoint, consisting of major bleeding and
ur

procedure/device-related complications, was similar in both groups (HR 0.80, 95% CI 0.55–
Jo

1.18). However, of the 48 safety events in the LAAC group, one-half were procedural adverse

events, suggesting that with continued follow-up, one could suspect further divergence

between groups in favor of LAAC.

Stroke and systemic cardioembolic events

The annualized rates of all-stroke/TIA and ischemic stroke were 2.4% and 1.9% for

LAAC, and 2.7% and 1.5% for NOAC, respectively, in our high-risk patient cohort

(CHA2DS2-VASc = 4.7±1.5). Certainly, these data must be considered cautiously since the

study was not powered to evaluate the differences in these components of the primary

endpoint. However, the incidence of approximately 2% per year is in agreement with

16
previously published data on annualized stroke rates in other randomized trials as well as non-

randomized registries of AF patients with LAAC vs NOAC. Hildick-Smith et al. reported an

annualized ischemic stroke incidence of 2.2% in a group of 1,088 subjects (CHA2DS2-VASc

= 4.2±1.6) after LAAC using the Amulet device. (7) Boersma et al. reported annualized rates

of stroke or stroke/TIA/SE as 1.3% and 2.0% in the 2-year follow-up of 1,020 patients

(CHA2DS2-VASc = 4.5±1.6) after LAAC using the Watchman device. (8)

With regard to NOAC therapy, in the ROCKET-AF trial (mean CHADS2 = 3.5), stroke

or SE occurred in 1.7% of patients per year. In the lower risk AVERROES trial patient cohort

f
oo
(mean CHADS2 = 2.0, prior stroke or TIA in only 14%), the annualized incidence of stroke

r
(without TIA) was 1.6% in the apixaban arm and 3.7% in the aspirin arm. (9) In older
-p
warfarin vs placebo trials, the incidence of stroke/TIA/SE in patients without antithrombotic
re
treatment was significantly higher. For instance, in the AFASAK trial, the annualized
lP

incidence of stroke/TIA/SE was 2.0% on warfarin and 5.5% on placebo (10), and the annual
na

incidence of cardioembolic events in the placebo arms of other warfarin trials varied from

4.7–8.0%.(11)
ur

Further reductions in cardioembolic events may be difficult to achieve using OAC or

Jo

LAAC strategies alone. Interestingly, the recently published LAAOS III trial demonstrated the

superiority of a novel strategy that combines surgical LAA closure with oral anticoagulation

vs. oral anticoagulation alone: the composite of ischemic stroke/TIA/SE was significantly

reduced in the surgical LAAC group (HR 0.67). (12) However, this combination strategy has

never been tested using transcatheter techniques, and might present a challenge for a

standalone LAAC procedure (unlike concomitant surgical LAA ligation).

In our cohort, 26 patients (12.9%) stopped NOAC treatment during the study period.

In previous publications, NOAC discontinuation was associated with an increased risk of

subsequent cardioembolic events. (13) (14) This is particularly important since OAC

17
discontinuation occurs relatively frequently in clinical practice, and even in clinical trial

settings. In a meta-analysis of NOAC trials in AF, the drug was discontinued in 21.7% of

patients, and discontinuation of both Warfarin and NOAC was associated with an increased

risk for cardioembolism. (15) The most common reasons for discontinuation were adverse

events, especially bleeding. In the AVERROES trial, apixaban was discontinued in 21.7% of

patients during 1.8 years of follow-up. In the propensity score-matched study of Amulet vs

NOAC therapy, 20% of patients in the NOAC cohort permanently stopped NOAC treatment

after three months. (5) Discontinuation of NOAC increases the risk of cardioembolic events,

f
oo
and as such, could have contributed to similar results in the LAAC and NOAC arms relative

r
to our trial. Indeed, drug discontinuation can happen with any medication, especially those

that cause side effects.

-p
re
Bleeding
lP

In the present follow-up of ~4 years, non-procedural clinically-relevant bleeding was

na

significantly reduced with the LAAC strategy (sHR 0.55, 95% CI 0.31–0.97). The bleeding

rate in LAAC patients (3.4%/year) was similar to that reported in other LAAC studies: in
ur

PROTECT-AF, the post-implant (> 7 days) bleeding rate was 3.5%/year. (16) (17) In the
Jo

EWOLUTION and Iberian registries, both of which included patients with OAC

contraindications, the corresponding major bleeding rates were 2.7%/year and 3.9%/year,

respectively. (8) (18)

In PRAGUE-17, a temporal analysis of bleeding events reveals progressive separation

in favor of LAAC during follow-up starting at six months (Central Illustration). This again

emphasizes that the benefits of LAAC become more evident as time progresses. The early

post-implant period is burdened by both i) procedure/device-related complications which are

typically bleeding events (6 of 9 complications in PRAGUE-17 were bleeding events), and ii)

the temporarily more intense antithrombotic treatment, both of which increase early bleeding.

18
A reduction in non-procedural (> 7 days) bleeding in favor of LAAC was also observed in the

5-year patient-level meta-analysis of the PROTECT-AF and PREVAIL trials, and even with a

similar HR of 0.48 (95% CI 0.32–0.71). (1)

Similar findings were also reported in non-randomized studies and registries. In the

Amulet observational study of 1,088 patients, there was a decrease in the annualized rate of

major bleeding from 10.1% for the 1st year to 4.0% for the 2nd year. (7) In the aforementioned

propensity-matched analysis, there were fewer major bleeding events after LAAC compared

to a matched NOAC control cohort (HR 0.62, 95% CI 0.49–0.79); interestingly, the time-to-

f
oo
event curves started to separate at three months post-implant, with progressive separation over

r
time. (5) In the APPLY propensity-matched analysis, the primary safety endpoint (major
-p
bleeding + procedure/device-related complications) occurred in 3.6%/year with LAAC: 52%
re
of events were procedural adverse events, and the remaining 48% were major bleeds during
lP

follow-up. (6) In the 4.5-year analysis of the CAP and CAP2 registries, together including
na

1,144 patients undergoing Watchman implantation, primary safety events composed of major

bleeding and procedure-related complications occurred in 3.1%/year. Roughly one-half of

ur

these events occurred within the first six months post-implantation. (19) Again, a significant
Jo

number of complications were bleeding events, further emphasizing the need to reduce

procedure- and device-related complications. To this point, next-generation devices like the

Watchman-FLX seem to be associated with a lower rate of procedural complications, as

recently demonstrated in the PINNACLE-FLX clinical trial. (20)

Study limitations

The composite endpoint itself contains both thromboembolism and bleeding

components, potentially with competing directions of effect. The PRAGUE-17 trial was

underpowered to evaluate the relative differences in individual components of the primary

composite endpoint, so all analyses of individual components need to be weighed carefully. In

19
the NOAC arm, no medication logs were kept. The results may not apply to all AF patients

since the study focused on high-risk patients with high CHA2DS2-VASC scores. Crossovers

from the LAAC to NOAC arm could theoretically bias toward the null hypothesis; however,

the per-protocol analysis of only patients treated as randomized yielded similar results.

Device-related thrombosis was not prospectively studied in all LAAC patients due to the

disruption caused by the COVID-19 pandemic; many of the planned TEEs had to be canceled.

CONCLUSION

Among non-valvular patients with atrial fibrillation and at high risk for stroke and

f
oo
bleeding, the non-inferiority of LAAC to NOAC relative to the composite of cardioembolic

r
events, CV death, significant procedure/device-related complications, or clinically-relevant
-p
bleeding, was maintained during long-term follow-up. The rate of non-procedural clinically-
re
relevant bleeding was significantly reduced with LAAC compared with NOAC therapy, but
lP

the study was underpowered to detect differences in stroke rate. The curves of clinically-
na

relevant bleeding appear to separate at ~ 6 months.

ur
Jo

20
CLINICAL PERSPECTIVES

Competency in Patient Care and Procedural Skills: Among patients with AF at high risk for

stroke and increased risk of bleeding, treatment with percutaneous LAAC is associated with

similar rates of AF-related events (stroke, cardiovascular death, and bleeding) compared with

NOAC treatment. The rate of post-procedural bleeding is reduced in the long-term.

Translational Outlook: Since the long-term clinical outcomes of LAAC patients are similar to

NOAC patients, further improvements in LAAC technology may further improve outcomes of

this non-pharmacological prevention strategy.

f
r oo
-p
re
lP
na
ur
Jo

21
REFERENCES

1. Reddy VY, Doshi SK, Kar S, et al. 5-Year Outcomes After Left Atrial Appendage

Closure: From the PREVAIL and PROTECT AF Trials. J Am Coll Cardiol.

2017;70(24):2964-75.

2. Ruff CT, Giugliano RP, Braunwald E, et al. Comparison of the efficacy and safety of

new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of

randomised trials. Lancet. 2014;383(9921):955-62.

3. Osmancik P, Herman D, Neuzil P, et al. Left Atrial Appendage Closure Versus Direct

f
oo
Oral Anticoagulants in High-Risk Patients With Atrial Fibrillation. J Am Coll Cardiol.

r
2020;75(25):3122-35.

4.
-p
Osmancik P, Tousek P, Herman D, et al. Interventional left atrial appendage closure vs
re
novel anticoagulation agents in patients with atrial fibrillation indicated for long-term
lP

anticoagulation (PRAGUE-17 study). Am Heart J. 2017;183:108-14.

na

5. Nielsen-Kudsk JE, Korsholm K, Damgaard D, et al. Clinical Outcomes Associated

With Left Atrial Appendage Occlusion Versus Direct Oral Anticoagulation in Atrial
ur

Fibrillation. JACC Cardiovasc Interv. 2021;14(1):69-78.

Jo

6. Gloekler S, Furholz M, de Marchi S, et al. Left atrial appendage closure versus

medical therapy in patients with atrial fibrillation: the APPLY study. EuroIntervention.

2020;16(9):e767-74.

7. Hildick-Smith D, Landmesser U, Camm AJ, et al. Left atrial appendage occlusion with

the Amplatzer Amulet device: full results of the prospective global observational study. Eur

Heart J. 2020;41(30):2894-901.

8. Boersma LV, Ince H, Kische S, et al. Evaluating Real-World Clinical Outcomes in

Atrial Fibrillation Patients Receiving the WATCHMAN Left Atrial Appendage Closure

22
Technology: Final 2-Year Outcome Data of the EWOLUTION Trial Focusing on History of

Stroke and Hemorrhage. Circ Arrhythm Electrophysiol. 2019;12(4):e006841.

9. Connolly SJ, Eikelboom J, Joyner C, et al. Apixaban in patients with atrial fibrillation.

N Engl J Med. 2011;364(9):806-17.

10. Petersen P, Boysen G, Godtfredsen J, et al. Placebo-controlled, randomised trial of

warfarin and aspirin for prevention of thromboembolic complications in chronic atrial

fibrillation. The Copenhagen AFASAK study. Lancet. 1989;1(8631):175-9.

11. Segal JB, McNamara RL, Miller MR, et al. Prevention of thromboembolism in atrial

f
oo
fibrillation. A meta-analysis of trials of anticoagulants and antiplatelet drugs. J Gen Intern

r
Med. 2000;15(1):56-67.

12.
-p
Whitlock RP, Belley-Cote EP, Paparella D, et al. Left Atrial Appendage Occlusion
re
during Cardiac Surgery to Prevent Stroke. N Engl J Med. 2021;384(22):2081-91.
lP

13. Park JH, Han SW, Lee KY, et al. Impact of Non-vitamin K Antagonist Oral
na

Anticoagulant Withdrawal on Stroke Outcomes. Front Neurol. 2018;9:1095.

14. Patel MR, Hellkamp AS, Lokhnygina Y, et al. Outcomes of discontinuing rivaroxaban
ur

compared with warfarin in patients with nonvalvular atrial fibrillation: analysis from the
Jo

ROCKET AF trial (Rivaroxaban Once-Daily, Oral, Direct Factor Xa Inhibition Compared

With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial

Fibrillation). J Am Coll Cardiol. 2013;61(6):651-8.

15. Chatterjee S, Sardar P, Giri JS, et al. Treatment discontinuations with new oral agents

for long-term anticoagulation: insights from a meta-analysis of 18 randomized trials including

101,801 patients. Mayo Clin Proc. 2014;89(7):896-907.

16. Holmes DR, Reddy VY, Turi ZG, et al. Percutaneous closure of the left atrial

appendage versus warfarin therapy for prevention of stroke in patients with atrial fibrillation:

a randomised non-inferiority trial. Lancet. 2009;374(9689):534-42.

23
17. Reddy VY, Sievert H, Halperin J, et al. Percutaneous left atrial appendage closure vs

warfarin for atrial fibrillation: a randomized clinical trial. JAMA. 2014;312(19):1988-98.

18. Lopez-Minguez JR, Nogales-Asensio JM, Infante De Oliveira E, et al. Long-term

Event Reduction After Left Atrial Appendage Closure. Results of the Iberian Registry II. Rev

Esp Cardiol (Engl Ed). 2019;72(6):449-55.

19. Holmes DR, Jr., Reddy VY, Gordon NT, et al. Long-Term Safety and Efficacy in

Continued Access Left Atrial Appendage Closure Registries. J Am Coll Cardiol.

2019;74(23):2878-89.

f
oo
20. Kar S, Doshi SK, Sadhu A, et al. Primary Outcome Evaluation of a Next-Generation

r
Left Atrial Appendage Closure Device: Results From the PINNACLE FLX Trial. Circulation.

2021;143(18):1754-62.
-p
re
lP
na
ur
Jo

24
FIGURE LEGENDS

Figure 1 Primary Outcome In The modified ITT Analysis

Cumulative incidence function for the primary composite outcome (cardiovascular death, all-

stroke/TIA, clinically-relevant bleeding and device/procedure-related complications) in the

presence of competing risk (non-cardiovascular death) in the mITT population. Number of

patients who are free of any event is supplemented (in brackets) by number of patients with

competing risk up to given time. CIF – cumulative incidence function, CI – confidence

interval, sHR – subdistributional hazard ratio

f
oo
Figure 2 Subgroup analysis

Summary forrest plots for the primary composite outcome in the mITT analysis. sHR –

r
-p
subdistributional hazard ratio, CI – confidence interval
re
Figure 3. Secondary Outcomes in the modified ITT analysis
lP

Cumulative Incidence Function For Cardiovascular Death (A), All-stroke/TIA (B), Clinically-
na

Relevant Bleeding (C), And Non-procedural Clinically-Relevant Bleeding (D), In The

Presence Of Competing Risk (Non-cardiovascular Death For Cardiovascular Death, And All-
ur

cause Death For Remaining Endpoints) In The mITT Population. Number of patients who are
Jo

free of any event is supplemented (in brackets) by number of patients with competing risk up

to given time. Abbreviations as in Figure 1.

Figure 4 Primary Outcome In The Per Protocol (A) And On-Treatment (B) Analyses

Cumulative incidence functions for the primary composite endpoint (cardiovascular death, all-

stroke/TIA, clinically-relevant bleeding and device/procedure-related complications) in the

Per Protocol and On-treatment analyses. Number of patients who are free of any event is

supplemented (in brackets) by number of patients with competing risk up to given time.

Abbreviations as in Figure 1.

Central Illustration A Summary Slide Of Primary And Secondary Endpoints

25
Shown are the patient characteristics, cumulative incidence function for primary endpoint, as

well as for stroke/TIA and non-procedural clinically-relevant bleeding in the modified

intention-to-treat population. AF – atrial fibrillation, TIA – transient ischemic attack, CV –

cardiovascular, SE – systemic embolism, CIF – cumulative incidence function, IQR –

interquartile range, sHR – subdistribution hazard ratio, CI – confidence interval

f
r oo
-p
re
lP
na
ur
Jo

26
Table 1 Baseline Characteristics

Baseline characteristics and risk factors of participants. AF – atrial fibrillation, MI – myocardial

infarction, LMWH – low molecular weight heparin

NOAC (n = LAAC (n = 201) Missing

201) values

Demographics

Age (years) 73.2 ± 7.2 73.4 ± 6.7 -

<75 years (%) 122 (60.7%) 116 (57.7%) -

f
oo
>75 years (%) 79 (39.3%) 85 (42.3%)

r
Male gender (%) 130 (64.7%) 134 (66.7%) -

Weight (kg)
-p 88.1 ± 16.2 86.9 ± 17.6 -
re
Clinical history
lP

AF type Paroxysmal (%) 67 (33.3%) 53 (26.4%) -

na

Persistent (%) 46 (22.9%) 47 (23.4%)

ur

LS persistent (%) 16 (8.0%) 18 (9.0%)

Permanent (%) 72 (35.8%) 83 (41.3%)

Jo

CHA2DS2-VASc 4.7 ± 1.5 4.7 ± 1.5 -

CHA2DS2-VASc < 3 (%) 50 (24.9%) 48 (23.9%) -

CHA2DS2-VASc = 4 (%) 40 (19.9%) 47 (23.4%)

CHA2DS2-VASc = 5 (%) 57 (28.4%) 50 (24.9%)

CHA2DS2-VASc > 6 (%) 54 (26.9%) 56 (27.9%)

HAS-BLED 3.0 ± 0.9 3.1 ± 0.9 -

Heart failure (%) 90 (44.8%) 88 (43.8%) -

Hypertension (%) 186 (92.5%) 186 (92.5%) -

27
Diabetes mellitus (%) 90 (44.8%) 73 (36.3%) -

History of cardioembolic event (%) 69 (34.3%) 73 (36.3%) -

Of which stroke (%) 63 (91.3%) 66 (90.4%) -

History of MI (%) 39 (19.4%) 30 (14.9%) -

Randomized at experienced centers 140 (69.7%) 141 (70.1%) -

Prior antithrombotic treatment

Warfarin 104 (51.7%) 85 (42.3%) -

NOACs 55 (27.4%) 66 (32.8%) -

f
oo
If no OAC, new AF appearance 30 (71.4%) 38 (76%)

r
Aspirin 32 (15.9%) 39 (19.4%) -

Clopidogrel
-p
11 (5.5%) 17 (8.5%) -
re
Dual antiplatelet treatment 6 (3.0%) 7 (3.5%) -
lP

Other (low dose LMWH, none) 19 (9.5%) 24 (11.9%) -

na
ur
Jo

28
Table 2 Number Of Events, Annualized Event Rate, And sHR For Primary And Secondary Outcomes In The mITT Analysis

Incidence of composite primary endpoint and its components in the presence of competing risk in the intention-to-treat populations. sHR –

subdistribution hazard ratio, CI – confidence interval

Total (N = 402) NOAC (N = 201) LAAC (N = 201)

f
oo
No. of No. of No. of

r
sHR P

-p
patient No. of Event patient No. of Event patient No. of Event

re
(95% CI) value
with events rate with events rate with events rate

lP
event event event

na
0.81

ur
Primary endpoint 109 139 10.27 60 81 11.92 49 58 8.60 0.27
(0.56; 1.18)
Jo
0.68
Cardiovascular death 50 50 3.69 30 30 4.42 20 20 2.96 0.19
(0.39; 1.20)

1.14
All-stroke / TIA 31 34 2.51 15 18 2.65 16 16 2.37 0.72
(0.56; 2.30)

1.38
All-stroke 25 26 1.92 11 12 1.77 14 14 2.08 0.42
(0.63; 3.03)

29
 Systemic embolism 1 1 0.07 1 1 0.15 0 0 0.00 - -

Clinically-relevant 0.75
56 69 5.10 32 40 5.89 24 29 4.30 0.28
bleeding (0.44; 1.27)

Non-procedural
0.55
clinically-relevant 50 63 4.65 32 40 5.89 18 23 3.41 0.039
(0.31; 0.97)

f
oo
bleeding

r
Procedure- or device-

-p
9 9 0.66 0 0 0.00 9 9 1.33 - -
related complication

re
lP
Non-cardiovascular 0.99
45 45 3.32 23 23 3.39 22 22 3.26 0.96

na
death (0.55; 1.77)

ur 0.81
Jo
All-cause death 95 95 7.02 53 53 7.80 42 42 6.23 0.31
(0.54; 1.22)

Subdistribution hazard ratios (sHR) for the LAAC group in comparison to the NOAC group and corresponding P values were calculated using Fine-Gray

regression models with the study group as a covariate. In case of multiple endpoints of the same type, sHR are based on the first event only.

Event rate is defined as no. of events per 100 patient-years.

30
Jo
ur
na
lP
re
-p
ro
of
Jo
ur
na
lP
re
-p
ro
of
Jo
ur
na
lP
re
-p
ro
of
Jo
ur
na
lP
re
-p
ro
of
 Online Appendix to:

Left Atrial Appendage Closure versus Non-Warfarin Oral Anticoagulation

in Atrial Fibrillation: 4-Year Outcomes of PRAGUE-17

Pavel Osmancik, MD, PhD, Dalibor Herman, MD, PhD, Petr Neuzil, MD, CSc, Pavel Hala,
MD, Milos Taborsky, MD, CSc, Petr Kala, MD, PhD, Martin Poloczek, MD, Josef Stasek,
MD, PhD, Ludek Haman, MD, PhD, Marian Branny, MD, PhD, Jan Chovancik, MD, Pavel
Cervinka, MD, PhD, Jiri Holy, MD, Tomas Kovarnik, MD, PhD, David Zemanek, MD, PhD,
Stepan Havranek, MD, PhD, Vlastimil Vancura, MD, PhD, Petr Peichl, MD, PhD, Petr

f
Tousek, MD, PhD, Veronika Lekesova, MD, Jiri Jarkovsky, RNDr, PhD., Martina

oo
Novackova, Mgr ., Klara Benesova, Mgr., Petr Widimsky, MD, DrSc., and Vivek Y. Reddy,
MD, on behalf of the PRAGUE-17 Investigators

r
Table of Contents:
-p
re
Page 2–3: List of Sites / Investigators
lP

Page 4: Members of the DSMB and CEC

Page 5: Endpoint Definitions
Page 6: Follow-Up Imaging
na

Page 7: Statistical Plan: Adjustment For Competing Risk

Page 8–9: On-Treatment and Per Protocol Analyses
ur

Page 10: Table S1. Additional Baseline Characteristics and Risk Factors of Enrolled
Patients
Page 12: Table S2: Incidence Of Composite Primary Endpoint And Its Components In
Jo

The Presence Of Competing Risk In The Per Protocol Populations

Page 13: Table S3: Incidence Of Composite Primary Endpoint And Its Components In
The Presence Of Competing Risk In The On-Treatment Populations
Page 14: Figure S1: CONSORT Diagram
Page 15: Figure S2: Kaplan – Meier Curves Of the Incidence Of Non-Cardiovascular
Death In The Modified Intention-To-Treat Population
Page 16: Figure S3: Kaplan – Meier Curves Of the Incidence of All-cause Death In The
Modified Intention-To-Treat Populations
Page 17: Figure S4: Cumulative Incidence Functions (CIF) For Components of the
Primary Endpoint In The Presence Of Competing Risk In The Per Protocol
Population
Page 18: Figure S5: Cumulative Incidence Functions (CIF) For Components of the
Primary Endpoint In The Presence Of Competing Risk In The On-treatment
Population

1
LIST OF SITES / INVESTIGATORS

Cardiocenter, Third Faculty of Medicine, Charles University Prague and University

Hospital Kralovske Vinohrady, Prague, Czech Republic
 Pavel Osmancik, MD, PhD
 Dalibor Herman, MD, PhD
 Petr Tousek, MD, PhD
 Petr Widimsky, MD, DrSc
 Hana Linkova, MD, PhD
 Eva Paskova, MD
 Robert Petr, MD
 Karol Curila, MD, PhD
 Petr Stros, MD

f
oo
Cardiocenter, Department of Cardiology, Na Homolce Hospital, Prague, Czech Republic
 Petr Neuzil, MD, CSc

r
 Vivek Y. Reddy, MD -p
 Pavel Hala, MD
re
 Veronika Lekesova, MD
lP

Cardiocenter, Dept. of Cardiology, University Hospital Olomouc, Czech Republic

 Milos Taborsky, MD, CSc
 Irena Opavska, Mgr
na

 Miloslav Spacek MD
 Eva Souckova, Eng.
ur
Jo

Clinic of Cardiology, Masaryk University and University Hospital, Brno, Czech Republic
 Petr Kala, MD, PhD
 Martin Poloczek, MD
 Tomas Ondrus, MD
 Lumir Koc, MD

1st Department of Internal Medicine, Faculty of Medicine, University Hospital Hradec

Kralove, Charles University Prague, Czech Republic
 Josef Stasek, MD, PhD
 Ludek Haman, MD, PhD
 Josef Bis, MD, PhD
 Jiri Jäger, MD
 Martina Tomkova, MD

Department of Cardiology, Cardiocenter, Hospital Podlesí a.s., Trinec, Czech Republic

 Marian Branny, MD, PhD
 Jan Chovancik, MD, PhD
2
  Libor Sknouril, MD, PhD
 Alexandra Vodzinska, MD
 Otakar Jiravsky, MD
 Martina Hrosova, MD
 Miroslav Hudec, MD
 Libor Gajdusek, MD
 Bronislav Holek, MD
 Miloslav Dorda, MD
 Jaroslav Januska, MD, PhD
 Pavla Kufova
 Sarka Konderlova

Department of Cardiology, Krajská zdravotni a.s., Masaryk hospital and UJEP, Usti nad

f
oo
Labem, Czech Republic
 Pavel Cervinka, MD, PhD
 Jiri Holy, MD

r
-p
Cardiocenter, 2nd internal clinic – Cardiology and Angiology, Charles University, General
re
Faculty Hospital, Prague Czech Republic
 Tomas Kovarnik, MD, PhD
lP

 Stepan Havranek, MD, PhD

 David Zemanek, MD, PhD
na

Department of Cardiology, University Hospital and Faculty of Medicine Pilsen, Pilsen,

ur

Czech Republic
 Vlastimil Vancura, MD, PhD
Jo

 Jan Opatrny, MD
 Richard Rokyta, MD, PhD
 Jan Lhotsky, MD

Cardiocenter, Institute of Clinical and Experimental Medicine, Prague, Czech Republic

 Petr Peichl, MD, PhD
 Bronislav Janek, MD
 Prodrag Stojadinovic, MD
 Ivana Kuklova

Statistics and IT (eCRF and randomization software, data export, data analysis), Institute
of Biostatistics and Analysis, Faculty of Medicine, Masaryk University, Brno, Czech Republic
 Dr. Jiri Jarkovsky, PhD
 Sarka Haskova, Mgr.
 Martina Novackova, Mgr.
 Klara Benesova, Mgr.

3
DATA SAFETY AND MONTIORING BOARD:
Jaromir Hradec, MD, CSc
Cardiocenter, 3rd internal clinic, First Faculty of Medicine, Charles University, General Faculty
Hospital, Prague Czech Republic

Frantisek Tousek, MD
Department of Cardiology, Nemocnice Ceské Budejovice, a.s., Ceske Budejovice, Czech
Republic

f
Rostislav Polasek, MD

oo
Department of Cardiology, Regional Hospital Liberec, Liberec Czech Republic

r
-p
re
CLINICAL ENDPOINT COMMITTEE:
lP

Jaroslav Ulman, MD
Cardiocenter, Third Faculty of Medicine, Charles University Prague and University Hospital
na

Kralovske Vinohrady, Prague, Czech Republic

ur

Frantisek Bednar, MD, PhD

Jo

Cardiocenter, Third Faculty of Medicine, Charles University Prague and University Hospital
Kralovske Vinohrady, Prague, Czech Republic

Jiri Knot, MD, PhD

Cardiocenter, Third Faculty of Medicine, Charles University Prague and University Hospital
Kralovske Vinohrady, Prague, Czech Republic

Tomas Peisker, MD, PhD

Clinic of Neurology, Third Faculty of Medicine, Charles University Prague and University
Hospital Kralovske Vinohrady, Prague, Czech Republic

4
ENDPOINT DEFINITIONS
Primary endpoint
The primary endpoint of the study will be the occurrence of any of the following events
following randomization:
(1) stroke (any type) or transient ischemic attack (TIA)
(2) systemic cardioembolic event
(3) clinically significant bleeding
(4) cardiovascular death
(5) a significant peri-procedural or device-related complication.
A stroke is defined as the sudden onset of a focal neurologic deficit, from a nontraumatic cause,

f
in a location consistent with the territory of a major cerebral artery and categorized as either

oo
ischemic, hemorrhagic, or unspecified. An imaging examination (CR, MRI) will be required.

r
A TIA is defined according to current standard criteria, i.e., as a stroke with relief of symptoms
within 24 hours.
-p
re
Systemic cardioembolic events will be defined as an acute vascular occlusion of an extremity
or organ, documented with imaging (CT angiography, percutaneous interventional
lP

angiography), surgery, or autopsy.

Clinically significant bleeding will be a composite of major and non-major clinically
na

significant bleeding. Clinically significant major bleeding will be defined according to the
ur

criteria of the International Society on Thrombosis and Hemostasis (ISTH), as clinically overt
bleeding accompanied by one or more of the following: a decrease in hemoglobin level ≥ 20
Jo

g/l over a 24-hour period, transfusion of 2 or more units of packed red cells, bleeding at a critical
site (intracranial, intraspinal, intraocular, pericardial, intramuscular with compartment
syndrome, or retroperitoneal), or fatal bleeding. Clinically significant non-major bleeding
will be defined as any bleeding requiring hospitalization or invasive procedure, which doesn´t
meet the ISTH major criteria. This definition is similar as that used recently in all trials
comparing NOAC with warfarin.
Peri-procedural and device-related complications will be evaluated as significant if there is
(1) a pericardial effusion requiring drainage (pericardiocentesis), or surgery, (2) a
cardioembolic event during the procedure, (3) significant peri-procedural bleeding (i.e., major
vascular bleeding requiring surgical revision, or blood concentrate transfusion), (4) device
embolization, or (5) a thrombus on the device with a consequent cardioembolic event, or others
serious adverse events associated with the procedure/device.

5
FOLLOW-UP IMAGING

According to the protocol, systematic follow-up TEE imaging should have been

performed (6 – 18 months after randomization) to assess for the presence of late device-related

thrombosis. However, due to the COVID-19 pandemic, the vast majority of these follow-up

TEEs were canceled.

f
r oo
-p
re
lP
na
ur
Jo

6
STATISTICAL PLAN: ADJUSTEMENT FOR COMPETING RISK

When the trial was initially designed, there was no pre-specified plan to adjust for the

competing risk of mortality. But at the trial conclusion (i.e. primary results in 2019),

consistent with the prevailing changes in convention in statistical methodology, it was decided

that all primary analyses should be conducted after adjusting for the competing risk for

mortality. Accordingly, for the primary composite endpoint and the cardiovascular mortality

endpoint, the data was calculated after adjusting for non-cardiovascular mortality. Similarly,

f
other (non-mortality) endpoints were calculated after adjusting for all-cause mortality.

oo
All of the data in the Main Manuscript is expressed after adjusting for competing risk.

r
-p
Also, in this Online Appendix, as indicated for each Table and Figure, the data include
re
adjustments for competing risk.
lP
na
ur
Jo

7
ON-TREATMENT ANALYSIS

Out of 201 patients allocated to the LAAC group, the procedure was not performed, or

was not successful in 20 (not performed in 14, not successful in 6). In one of these patients, a

procedure-related endpoint (device embolization) occurred during the unsuccessful LAAC

procedure. Surgical extraction was required, during which the left atrial appendage was excised;

therefore, this patient entered the on-treatment analysis along with the procedure-related

endpoint, but was censored from further analysis for other endpoints. In two other patients (a

patient with suspicion for infective endocarditis on TEE a day before the procedure, and a

f
oo
patient with pericardial effusion), the procedure was not performed, and these patients were not

r
crossed-over to NOAC. The remaining 17 patients subsequently received NOACs, and were
-p
analyzed in the NOAC group for the on-treatment analysis.
re
On the other hand, for 2 patients allocated to the NOAC group, early (within one month)
lP

cross-over to LAAC occurred because of patient preference. Accordingly, these patients were
na

analyzed as being treated with LAAC in the on-treatment analysis. Ultimately, 184 patients in

the LAAC group, and 216 patients in the NOAC group entered the on-treatment analysis. In 17
ur

LAAC patients, long-term NOAC treatment was initiated at some point during the study period
Jo

for different reasons (deep vein thrombosis, peri-device leak, ischemic event). In 3 (18%) of

these patients, NOACs were initiated due to device-related thrombi by TEE, in 3 (18%) due to

peri-device leak, and in 5 (29%) due to the presence of an ischemic event (stroke or TIA). All

17 patients were censored in the on-treatment analysis at the point of NOAC initiation. Finally,

NOAC therapy was permanently stopped during follow-up in 26 NOAC group patients – in 15

of them (58%) due to clinically-relevant bleeding, with or without subsequent cross-over to

LAAC. Cross-over to the LAAC group occurred in 13 patients: 12 procedures were successful,

but one was unsuccessful for anatomical reasons (an overly large LAA). No complications

occurred in any of these 13 late crossover patients. For the purpose of the on-treatment analysis,

8
those patients remained in the NOAC group until the point that NOAC treatment ceased, and

were then censored.

PER-PROTOCOL ANALYSIS

Out of 201 patients allocated to the LAAC arm, the procedure was either not performed

or was not successful in 20. Therefore, 181 patients entered the per protocol analysis in the

LAAC group. In 17 LAAC patients, long-term NOAC treatment was initiated at some point

during the study period for various reasons. All of these 17 patients were censored in the per

f
oo
protocol analysis at the point of NOAC initiation.

r
Out of 201 patients allocated to the NOAC group, two patients were crossed-over
-p
early (within one month, without endpoint occurrence) to LAAC. Thus, 199 patients entered
re
the per protocol analysis in the NOAC group. In 26 patients in the NOAC group, NOACs
lP

were permanently stopped during follow-up – primarily for bleeding, with or without
na

subsequent cross-over to LAAC. For the per-protocol analysis, those patients who crossed-

over to LAAC because of bleeding were censored at the time of cross-over. Conversely,
ur

patients in whom NOAC treatment was stopped and did not cross-over to LAAC were
Jo

maintained in the NOAC group for the per protocol analysis.

9
Table S1: Additional Baseline Characteristics And Risk Factors

NOAC (n = 201) LAAC (n = 201) Missing

Values
Clinical History
AF duration < 1 year (%) 59 (29.4%) 49 (24.4%) -
1 – 2 years (%) 38 (18.9%) 47 (23.4%) -
> 3 years (%) 104 (51.7%) 105 (52.2%) -
Uncontrolled hypertension (%) 62 (30.8%) 56 (27.9%) -
History of CABG (%) 22 (10.9%) 23 (11.4%) -
History of clinically-relevant bleeding -
95 (47.3%) 109 (54.2%)
or bleeding predisposition (%)

f
oo
Hx of Clinically-relevant bleeding 91 (45.3%) 101 (50.2%)
Abnormal renal/liver function (%) 47 (23.4%) 44 (21.9%) -

r
Labile INR (%) 88 (43.8%) 109 (54.2%) -
Alcohol abuse (%)
-p
8 (4.0%) 7 (3.5%) -
re
Drugs increasing bleeding risk (%) 29 (14.4%) 28 (13.9%) -
Pacemaker 47 (23.4%) 48 (23.9%) -
lP

ICD 23 (11.4%) 18 (9.0%) -

Echocardiography
na

Left ventricular ejection fraction (%) 53.3 ± 12.6 52.9 ± 12.1 7/4
Left atrial size (mm) 51.8 ± 6.4 51.7 ± 7.5 27 / 23
ur

Prior Medication Treatment

ACEI -
Jo

114 (56.7%) 127 (63.2%)

Beta-blocker 157 (78.1%) 155 (77.1%) -
Diuretics 129 (64.2%) 133 (66.2%) -
Spironolactone 58 (28.9%) 57 (28.4%) -
Ca channel blockers 68 (33.8%) 58 (28.9%) -
Other antihypertensives 47 (23.4%) 52 (25.9%) -
Insulin 19 (9.5%) 19 (9.5%) -
NSAID 17 (8.5%) 14 (7.0%) -
Oral corticosteroids 8 (4.0%) 6 (3.0%) -
Blood Count and Biochemistry
Hemoglobin (g/L) 139 (129; 148) 140 (127; 151) 13 / 12
(median and IQR)
Platelet count (10^9/L) 210 (174; 253) 196 (163; 236) 14 / 14
Serum creatinine (µmol/L) 96 (81; 116) 97 (80; 119) 12 / 8
Creatinine clearance (mL/s/1.73m2) 62.4 (49.4; 74.7) 60.8 (47.6; 74.7) 12 / 8
NT-pro BNP (pmol/L) 454 (131; 1 469) 528 (162; 1 351) 54 / 53

10
CRP (mg/dL) 3.0 (1.6; 6.4) 3.0 (1.9; 5.9) 20 / 18

f
r oo
-p
re
lP
na
ur
Jo

11
Table S2: Incidence Of Composite Primary Endpoint And Its Components In The Presence Of Competing Risk In The Per Protocol
Populations
Total (N = 380) NOAC (N = 199) LAAC (N = 181)
No. of No. of No. of
sHR P
patient No. of Event patient No. of Event patient No. of Event
(95% CI) value
with events rate with events rate with events rate
event event event
0.80
Primary endpoint 103 124 10.15 60 75 11.62 43 49 8.50 0.25
(0.54; 1.18)

f
oo
0.70
Cardiovascular death 42 42 3.44 26 26 4.03 16 16 2.78 0.27
(0.38; 1.31)

r
1.01

-p
All-stroke / TIA 28 30 2.46 15 17 2.63 13 13 2.26 0.99
(0.48; 2.11)

re
1.30
All-stroke 23 24 1.96 11 12 1.86 12 12 2.08 0.53
(0.57; 2.93)

lP
Systemic embolism 1 1 0.08 1 1 0.15 0 0 0.00 - -

na
Clinically-relevant 0.76
53 64 5.24 32 39 6.04 21 25 4.34 0.32
bleeding (0.44; 1.31)
Non-procedural
clinically-relevant 47 58 4.75 ur 32 39 6.04 15 19 3.30
0.52
0.039
Jo
(0.29; 0.97)
bleeding
Procedure- or device-
related complication 8 8 0.65 0 0 0.00 8 8 1.39 - -

Non-cardiovascular 0.99
42 42 3.44 23 23 3.56 19 19 3.30 0.98
death (0.54; 1.82)
0.83
All-cause death 84 84 6.87 49 49 7.59 35 35 6.07 0.40
(0.54; 1.28)
Subdistribution hazard ratios (sHR) for the LAAC group in comparison to the NOAC group and corresponding P values were calculated using Fine-Gray
regression models with the study group as a covariate. In case of multiple endpoints of the same type, sHR are based on the first event only.

Event rate is defined as no. of events per 100 patient-years.

12
Table S3: Incidence Of Composite Primary Endpoint And Its Components In The Presence Of Competing Risk In The On-Treatment
Populations
Total (N = 400) NOAC (N = 216) LAAC (N = 184)
No. of No. of No. of
sHR P
patient No. of Event patient No. of Event patient No. of Event
(95% CI) value
with events rate with events rate with events rate
event event event
0.82
Primary endpoint 107 129 10.18 63 79 11.55 44 50 8.57 0.30
(0.56; 1.20)

f
oo
0.70
Cardiovascular death 43 43 3.39 27 27 3.95 16 16 2.74 0.26
(0.38; 1.30)

r
0.86

-p
All-stroke / TIA 31 33 2.60 18 20 2.92 13 13 2.23 0.68
(0.42; 1.75)

re
1.12
All-stroke 25 26 2.05 13 14 2.05 12 12 2.06 0.77
(0.52; 2.45)

lP
Systemic embolism 1 1 0.08 1 1 0.15 0 0 0.00 - -

na
Clinically-relevant 0.78
54 65 5.13 33 40 5.85 21 25 4.29 0.38
bleeding (0.45; 1.35)
Non-procedural
clinically-relevant 48 59 4.66 ur 33 40 5.85 15 19 3.26
0.54
0.049
Jo
(0.30; 1.00)
bleeding
Procedure- or device-
related complication 9 9 0.71 0 0 0.00 9 9 1.54 - -

Non-cardiovascular 1.03
43 43 3.39 24 24 3.51 19 19 3.26 0.92
death (0.56; 1.89)
0.84
All-cause death 86 86 6.79 51 51 7.46 35 35 6.00 0.44
(0.55; 1.30)
Subdistribution hazard ratios (sHR) for the LAAC group in comparison to the NOAC group and corresponding P values were calculated using Fine-Gray
regression models with the study group as a covariate. In case of multiple endpoints of the same type, sHR are based on the first event only.

Event rate is defined as no. of events per 100 patient-years.

13
Figure S1 CONSORT Diagram

f
oo r
-p
re
lP
na
ur
Jo

14
Figure S2. Cumulative Incidence Function For Non-Cardiovascular Death In The

Presence Of Competing Risk (Cardiovascular Death) In The Modified Intention-To-

Treat Population

f
r oo
-p
re
lP
na
ur
Jo

15
Figure S3. Kaplan–Meier Estimates Of The Incidence Of All-Cause Death In The

Modified Intention-To-Treat Populations

f
r oo
-p
re
lP
na
ur
Jo

16
Figure S4. Cumulative Incidence Functions (CIF) For Components of the Primary

Endpoint In The Presence Of Competing Risk In The Per Protocol Populations

Shown are the cumulative incidence function curves for cardiovascular death (A), all-

stroke/TIA (B), clinically-relevant bleeding (C), and non-procedural clinically-relevant

bleeding (D), in the presence of competing risk (non-cardiovascular death for cardiovascular

death, and all-cause death for remaining endpoints) in the Per Protocol Population

f
r oo
-p
re
lP
na
ur
Jo

17
Figure S5. Cumulative Incidence Functions (CIF) For Components of the Primary

Endpoint In The Presence Of Competing Risk In The On-Treatment Population.

Shown are the cumulative incidence function curves for cardiovascular death (A), all-

stroke/TIA (B), clinically-relevant bleeding (C), and non-procedural clinically-relevant

bleeding (D), in the presence of competing risk (non-cardiovascular death for cardiovascular

death, and all-cause death for remaining endpoints) in the On-treatment Population

f
r oo
-p
re
lP
na
ur
Jo

18