The n e w e ng l a n d j o u r na l of m e dic i n e

Original Article

Left Atrial Appendage Closure or Medical

Therapy in Atrial Fibrillation
U. Landmesser,1-3 C. Skurk,1,2 P. Kirchhof,4-6 T. Lewalter,7 J. Hartung,1,2 A. Rroku,1
B. Pieske,8 J. Brachmann,9 I. Akin,10 C. Jacobshagen,11 B. Meder,12,13 A. Zeiher,14,15
S.D. Anker,2,16 H. Thiele,17 S. Blankenberg,4,5 S. Massberg,18,19 H. Schunkert,19,20
N. Frey,12,13 A. Joost,5,21 M. Bergmann,22 R.S. von Bardeleben,23 T. Friede,24,25
M. Placzek,24,25 A. Suling,26 K.G. Haeusler,27 M. Endres,2,28 K. Wegscheider,5,26
L.-H. Boldt,2,29 and I. Eitel,5,21 for the CLOSURE-AF Trial Investigators*​​

A BS T R AC T

BACKGROUND
The authors’ full names, academic degrees, Catheter-based closure of the left atrial appendage is an alternative to oral antico-
and affiliations are listed at the end of agulation for stroke prevention in patients with atrial fibrillation. The effectiveness
the article. Ulf Landmesser can be con-
tacted at ­ulf​.­landmesser@​­dhzc-charite​.­de of this strategy, as compared with physician-directed best medical care, in patients
or at the Department of Cardiology, Angiol- at high risk for stroke and bleeding is unknown.
ogy, and Intensive Care Medicine, Deutsch-
es Herzzentrum der Charité, Campus Ben- METHODS
jamin Franklin, Charité University Medicine In this multicenter randomized trial conducted in Germany, we assigned patients
Berlin, Hindenburgdamm 30, 12203 Berlin,
Germany. with atrial fibrillation and a high risk of stroke and bleeding to undergo left atrial
appendage closure or to receive physician-directed best medical care (including di-
*A list of the CLOSURE-AF Trial Investi-
gators is provided in the Supplemen- rect oral anticoagulants, if eligible). The primary end point, tested for noninferior-
tary Appendix, available at [Link]. ity, was a composite of stroke (ischemic or hemorrhagic), systemic embolism, major
Ulf Landmesser and Carsten Skurk con- bleeding, or cardiovascular or unexplained death, assessed in a time-to-event analy-
tributed equally to this article. sis. The noninferiority margin was a hazard ratio of 1.3.
This article was published on March 18, RESULTS
2026, at [Link].
A total of 912 adult patients underwent randomization. The primary end-point analy-
N Engl J Med 2026;394:1270-80. sis included 446 patients who were assigned to undergo left atrial appendage closure
DOI: 10.1056/NEJMoa2513310
Copyright © 2026 Massachusetts Medical Society.
(device group) and 442 who were assigned to physician-directed best medical care
(medical-therapy group). The mean (±SD) age was 77.9±7.1 years; 38.6% of the patients
CME were women, the mean CHA2DS2-VASc score was 5.2±1.5 (range, 0 to 9, with higher
scores indicating a greater risk of stroke), and the mean HAS-BLED score was 3.0±0.9
(range, 0 to 9, with higher scores indicating higher risk of bleeding). After a median
follow-up of 3 years (interquartile range, 1.7 to 4.7), a first primary end-point event
had occurred in 155 patients (incidence per 100 patient-years, 16.8) in the device group
and in 127 patients (incidence per 100 patient-years, 13.3) in the medical-therapy group
(difference in restricted mean survival time, −0.36 years; 95% confidence interval,
−0.70 to −0.01; P = 0.44 for noninferiority). Serious adverse events occurred in 368
patients (82.5%) in the device group and 342 (77.4%) in the medical-therapy group.
CONCLUSIONS
Among patients with atrial fibrillation at high risk for stroke and bleeding, left
atrial appendage closure was not noninferior to physician-directed best medical care
with regard to a composite end point of stroke, systemic embolism, major bleeding,
or cardiovascular or unexplained death. (Funded by the German Center for Cardio-
vascular Research; CLOSURE-AF [Link] number, NCT03463317.)

1270 n engl j med 394;13 [Link] April 2, 2026

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 Left Atrial Appendage Closure in Atrial Fibrillation

A
trial fibrillation is a serious Me thods
health care challenge owing to the in-
creasing incidence in the aging popula- Trial Design
tion and the association with an elevated risk of This trial was a pragmatic, prospective, open-
stroke and death.1,2 For decades, oral anticoagu- label, multicenter, randomized, controlled trial
A Quick Take
lation with vitamin K antagonists, such as war- with blinded outcome assessment and a parallel
is available at
farin, has been the standard of care for stroke two-group design that recruited patients at 42 [Link]
prevention in atrial fibrillation. Although this German sites with experience in left atrial ap-
treatment approach is effective, many patients pendage closure. The trial design has been pub-
were not treated owing to the therapy-associated lished previously, and the protocol is available
increased bleeding risk.3 Four direct-acting oral with the full text of this article at [Link].14 A
anticoagulants (DOACs) have shown efficacy for full list of participating sites is provided in the
stroke prevention in atrial fibrillation that is at Supplementary Appendix, available at [Link].
least similar to that of warfarin, but with a more The protocol and statistical analysis plan (avail-
favorable safety profile, especially with respect able with the protocol) were designed by the
to intracranial bleeding.4-7 Although DOACs are principal and coordinating investigators, mem-
widely available, many patients with atrial fi- bers of the steering committee, and the trial
brillation are not treated with them, often owing statisticians. Data were collected electronically
to a perceived bleeding risk, previous bleeding (secuTrial, interActive Systems) at the trial sites,
complications, or unacceptable side effects. The stored at a central location, and analyzed by the
randomized phase 3 trials comparing DOAC trial statisticians. The trial was conducted in
therapy with warfarin excluded patients with a compliance with the Good Clinical Practice
history of major hemorrhage, and 20 to 30% of guidelines of the International Council for Har-
trial patients who received DOAC therapy discon- monisation and the principles of the Declaration
tinued treatment at 2 years.4-7 Therefore, there is of Helsinki, coordinated by Charité University
a need to examine alternative approaches for Medicine Berlin, and approved by the ethics
stroke prevention in patients with atrial fibril- committee at each participating site. All the pa-
lation. tients provided written informed consent. The
Catheter-based closure of the left atrial ap- steering committee (Table S1 in the Supplemen-
pendage has been suggested to provide protec- tary Appendix) and an independent data and
tion against stroke, systemic embolism, and car- safety monitoring board (Table S2) oversaw trial
diovascular death that is similar to that provided conduct and patient safety. A clinical events com-
by anticoagulation therapy in moderate-sized mittee, the members of which were unaware of
clinical trials.8-10 Device-based strategies are as- trial-group assignments, adjudicated all clinical-
sociated with a certain degree of periprocedural outcome events (Table S3). The authors had un-
risk11,12 but have the potential to reduce bleeding restricted access to the data, prepared all drafts
events. Data comparing left atrial appendage of the manuscript before submission for publica-
closure to contemporary medical therapy are tion, and vouch for the completeness and accu-
scarce, especially with regard to patients who racy of the data and for the fidelity of the trial
are at high risk for stroke and bleeding,9,10,13 the to the protocol. The trial was sponsored by the
patient population that is considered to poten- German Center of Cardiovascular Research and
tially derive the greatest benefit. Therefore, we conducted in collaboration with the Atrial Fibril-
designed the CLOSURE-AF (Catheter-Based Left lation Network.
Atrial Appendage Closure in Patients with Atrial
Fibrillation at High Risk of Stroke and Bleeding Trial Population
as Compared with Best Medical Therapy) trial to Patients were eligible for inclusion if they had
compare left atrial appendage closure with phy- atrial fibrillation and were at a high risk for
sician-directed best medical care (including DO- stroke and bleeding or had contraindications to
ACs when considered feasible) in patients with long-term anticoagulation therapy. High risk of
atrial fibrillation and a high risk of stroke and stroke was defined as a CHA2DS2-VASc score of
bleeding. 2 or higher (range, 0 to 9, with higher scores

n engl j med 394;13 [Link] April 2, 2026 1271

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 The n e w e ng l a n d j o u r na l of m e dic i n e

indicating a greater risk of stroke). High risk of The treating physician decided what qualified
bleeding was defined as the presence of at least as best medical care, with DOAC as the default
one of the following: HAS-BLED score of 3 or therapy in patients who were deemed to be eli-
higher (range, 0 to 9, with higher scores indicat- gible for anticoagulation therapy.15 When antico-
ing higher risk); previous intracranial, intraspi- agulation therapy was ruled out, an antiplatelet
nal, or intraocular bleeding compromising vision or no-antithrombotic therapy was selected. The
(Bleeding Academic Research Consortium [BARC] patients who did not qualify for anticoagulation
type 3c); hemorrhagic or bleeding complications therapy were not prospectively identified. All the
fulfilling BARC type 3a or 3b criteria (i.e., gas- patients were followed for a minimum of 6
trointestinal, genitourinary, or respiratory tract months, with visits at 3, 6, 12, 18, and 24 months
bleeding and drop in hemoglobin level); chronic and yearly thereafter until the end of the trial.
kidney disease with an estimated glomerular fil- Clinical information was recorded at each follow-
tration rate of 15 to 29 ml per minute per 1.73 m2 up visit, including medication, adverse events,
of body-surface area; or any recurrent bleeding and quality of life.
that precluded long-term anticoagulation therapy.
Key inclusion and exclusion criteria are provided End Points
in Table S4. The representativeness of the trial The primary end point was a patient-relevant
population is provided in Table S5. composite end point of stroke (ischemic or hem-
orrhagic), systemic embolism, major bleeding
Trial Intervention and Follow-up (BARC type 3 or higher), or cardiovascular or
Randomization was performed electronically and unexplained death, assessed in a time-to-event
stratified according to trial site. Patients as- analysis. All components of the primary end
signed to undergo left atrial appendage closure point were adjudicated by an independent clini-
(the device group) could receive a Watchman cal events committee, the members of which
(Watchman or Watchman FLX, Boston Scientific) were unaware of the treatment-group assign-
or Amplatzer (Amplatzer Cardiac Plug or Amu- ments. Secondary end points included the compo-
let, St. Jude Medical)8,9 device. For patients who nents of the composite primary end point and
were not eligible to receive a Watchman or Am- death from any cause. A complete list of end
platzer device, the LAmbre device (LifeTech) was points is provided in Table S6.
available at selected sites as a bailout device. All
participating sites were required to have docu- Statistical Analysis
mented experience with catheter-based left atrial The CLOSURE-AF trial was planned as an event-
appendage closure, and operators were required driven trial with two unblinded interim analyses
to be certified for all devices used.14 Postimplan- performed according to a sequential design. The
tation antithrombotic treatment was individual- initial sample-size calculation was based on
ized according to the patient’s bleeding risk and published event rates from randomized trials or
was determined by the treating physician (Fig. registries of left atrial appendage closure,10-12,16-18
S1). After deployment of the device, therapy with involving patients who were at mainly low-to-
a dual antiplatelet agent was recommended for moderate risk, as detailed in the protocol.14 As-
at least 3 months. If transesophageal echocar- suming that 50% of the patients who were in-
diography showed complete left atrial appendage cluded would have a history of major bleeding,
closure or only a small residual leak (<5 mm) we expected that the annual control-group haz-
without device-related thrombus, discontinuation ard rate for the primary end point would be
of dual antiplatelet therapy was recommended. 11.2 events (0.5 × 6.2 + 0.5 × 16.1) per 100 patient
Discontinuation of single antiplatelet therapy years.18,19 With a hazard ratio of 1.3 as the non-
with aspirin was recommended after 6 months inferiority margin, 467 primary events would be
unless a clear indication existed for continuing required to provide the trial with 80% power to
therapy. In patients with excessive bleeding risk, show that left atrial appendage closure was non-
dual antiplatelet therapy could be shortened to inferior to best medical care. With a 36-month
6 weeks, and aspirin could be stopped after 3 recruitment period, a follow-up time of at least
months.14 24 months, and 5% annual loss to follow-up, we

1272 n engl j med 394;13 [Link] April 2, 2026

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 Left Atrial Appendage Closure in Atrial Fibrillation

calculated that a total of 1586 patients would R e sult s

need to undergo randomization.
Two prespecified blinded interim analyses of Patients
pooled data were performed for sample-size re- From March 2018 through May 2024, a total of
calculation and adaptation of the trial design, as 10,872 patients were screened and 912 under-
detailed in the Supplementary Appendix; alpha went randomization at 42 sites. One site was
allocation was not required for the two interim excluded from the analysis owing to failed data
analyses because no comparative data were used. quality control (23 patients), and one patient was
The sequential design was abandoned and re- excluded owing to having been erroneously in-
cruitment stopped after 912 patients were en- cluded in another interventional study (Fig. 1).
rolled, because published data from another Of the remaining 888 patients, 446 were as-
trial20 suggested that this trial might already be signed to undergo closure of the left atrial ap-
overpowered. pendage (device group) and 442 were assigned
The primary analysis for the comparison of to receive physician-directed best medical care
left atrial appendage closure with physician- (medical-therapy group). In the device group,
directed best medical therapy was performed 421 patients (94.4%) underwent the assigned
according to the intention-to-treat principle, and treatment (median time to implantation, 2 days;
hypothesis testing assessed a hazard ratio of interquartile range, 1 to 5), 56 (12.6%) with-
less than 1.3 as compared with a hazard ratio of drew from the trial, and 10 (2.2%) were lost to
1.3 or greater. A Cox model with randomly as- follow-up. In the medical-therapy group, 442
signed group as the only factor, with stratifica- patients (100%) underwent the assigned treat-
tion according to site, was used for hypothesis ment, 48 (10.9%) withdrew from the trial, and
testing. Because Schoenfeld residual plots indi- 23 (5.2%) were lost to follow-up. A total of 18
cated that there were some deviations from the patients in the medical-therapy group underwent
proportional-hazards assumption for the pri- left atrial appendage closure during follow-up
mary end point as well as for some of its com- without having had a primary end-point event
ponents (Fig. S2), in particular major bleeding, (Table S7).
differences in restricted mean survival times Demographic and clinical characteristics of the
were used to summarize the between-group dif- patients appeared to be well balanced between the
ferences in end-point events. To account for com- groups (Table 1 and Table S8). The mean (±SD)
peting events, the restricted mean survival time age was 77.9±7.1 years; 343 of the patients
was calculated as the area under 1 minus the (38.6%) were women. The mean CHA2DS2-VASc
cumulative-incidence curve. Pseudo-observations score was 5.2±1.5 and the mean HAS-BLED
accounting for site-related effects were comput- score was 3.0±0.9. Most of the patients in the
ed and linear models fitted to them to derive device group (79.6%) received dual antiplatelet
treatment-group means and their differences therapy as the initial antithrombotic therapy
with 95% confidence intervals. Secondary time-to- (Fig. S3 and Table S9). The majority of the pa-
event end points were reported with 95% two- tients in the medical-therapy group (85.1%) re-
sided confidence intervals on the basis of an ceived a DOAC.
analogous Cox model. Aalen–Johansen cumula-
tive-incidence curves with adjustment for compet- Primary End Point
ing events were used for visualization. Adverse The median follow-up was 3.0 years (interquartile
events were compared descriptively. Confidence range, 1.7 to 4.7), with a maximum follow-up of
intervals were not adjusted for multiplicity and 6.7 years. A first primary end-point event occurred
should not be interpreted as hypothesis tests; only in 155 patients (incidence per 100 patient-years,
P values controlled for type 1 error are reported. 16.8) in the device group and in 127 patients
No imputation was planned for missing values. (incidence per 100 patient-years, 13.3) in the
Further analyses, such as subgroup and adjusted medical-therapy group (difference in restricted
analyses, were prespecified in the statistical analy- mean survival time, −0.36 years; 95% confidence
sis plan. Additional details regarding the analyses interval [CI], −0.70 to −0.01; P = 0.44 for nonin-
are provided in the Supplementary Appendix. feriority) (Table 2 and Fig. 2). The incidence of a

n engl j med 394;13 [Link] April 2, 2026 1273

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 The n e w e ng l a n d j o u r na l of m e dic i n e

primary end-point event in the half of the trial had noncardiovascular death is shown in Figure
population that was enrolled first as compared S4. Results of prespecified subgroup analyses
to the half that was enrolled second is shown in also appeared to be consistent with those of the
Table S10. The incidence of noncardiovascular primary analysis (Figs. S5 and S6).
death did not differ meaningfully between the Stroke occurred in 27 patients in each of the
groups (Fig. 2). Results of a sensitivity analy- two groups (incidence per 100 patient-years, 2.6
sis performed in the per-protocol population and 2.7, respectively; difference in restricted
appeared to be consistent with the results of the mean survival time, 0.01 years; 95% CI, −0.17 to
primary analysis, which was conducted in the 0.20). Systemic embolism occurred in 3 patients
intention-to-treat population (difference in re- in the device group and in 1 patient in the
stricted mean survival time, −0.40 years; 95% CI, medical-therapy group (incidence per 100 pa-
−0.76 to −0.04). The occurrence of primary end- tient-years, 0.3 and 0.1; difference in restricted
point events with respect to patients who with- mean survival time, −0.02 years; 95% CI, −0.08
drew from the trial, were lost to follow-up, or to 0.03). Major bleeding occurred in 70 patients

912 Patients underwent randomization at 42 sites

458 Were assigned to left atrial appendage closure 454 Were assigned to physician-directed best medical care

11 Were excluded owing to the

exclusion of one center 12 Were excluded owing to the
1 Was excluded after undergoing exclusion of one center
randomization in error

446 Were included in the trial 442 Were included in the trial
414 (92.8%) Received left atrial appendage closure 376 (85.1%) Received DOAC
225 (54.3%) Received Watchman device 303 (68.6%) Received DOAC alone
42 (10.1%) Received Watchman 2.5 device 67 (15.2%) Received DOAC and single antiplatelet
183 (44.2%) Received Watchman FLX device therapy
173 (41.8%) Received Amplatzer Amulet device 6 (1.4%) Received DOAC and dual antiplatelet
16 (3.9%) Received LAmbre device therapy
7 (1.6%) Underwent unsuccessful left atrial appendage 14 (3.2%) Received vitamin K antagonist
closure 11 (2.5%) Received vitamin K antagonist alone
25 (5.6%) Did not have left atrial appendage closure 2 (0.5%) Received vitamin K antagonist and single
attempted antiplatelet therapy
1 (0.2%) Received vitamin K antagonist and dual
antiplatelet therapy
7 (1.6%) Received single antiplatelet therapy alone
3 (0.7%) Received dual antiplatelet therapy alone
33 (7.5%) Did not receive antithrombotics
9 (2.0%) Had missing information regarding anti-
thrombotics

144/1065 (13.5%) Total follow-up yr 177/1134 (15.6%) Total follow-up yr

were lost were lost
128 (12.0%) Follow-up yr 138 (12.2%) Follow-up yr
were lost because 56 were lost because 48
withdrew withdrew
17 (1.6%) Follow-up yr 40 (3.5%) Follow-up yr
were lost because 10 were lost because 23
were lost to follow-up were lost to follow-up

446 Were included in the primary intention-to-treat analysis 442 Were included in the primary intention-to-treat analysis

Figure 1. Enrollment, Randomization, and Follow-up.

DOAC denotes direct-acting oral anticoagulant.

1274 n engl j med 394;13 [Link] April 2, 2026

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 Left Atrial Appendage Closure in Atrial Fibrillation

in the device group (related to the procedure in Secondary End Points

18 patients and unrelated to the procedure in 52) Death from any cause occurred in 155 patients
and in 61 patients in the medical-therapy group in the device group and 141 patients in the
(incidence per 100 patient-years, 7.4 and 6.2, re- medical-therapy group (difference in restricted
spectively; difference in restricted mean survival mean survival time, −0.13 years; 95% CI, −0.48
time, −0.12 years; 95% CI, −0.40 to 0.15). The to 0.21) (Table S15). Other secondary end-point
incidence of major bleeding events in both events are listed in Table 2.
groups before and after 3 months and 6 months Procedural outcomes are shown in Table 3. In
are shown in Table S11. Cardiovascular or unex- the device group, implantation of the device was
plained death occurred in 99 patients in the attempted in 421 patients and was successful in 414
device group and 81 patients in the medical- (98.3%). The most common reason for an unsuc-
therapy group (incidence per 100 patient-years, cessful implantation was unsuitable patient ana-
9.5 and 7.7, respectively; difference in restricted tomical features. Among the 421 patients in which
mean survival time, −0.19 years; 95% CI, −0.50 implantation was attempted, technical and proce-
to 0.11) (Tables S12 through S14). dural success was achieved in 411 (97.6%) and 406

Table 1. Characteristics of the Patients at Baseline.*

Left Atrial Appendage Physician-Directed

Closure Best Medical Care Total
Characteristic (N = 446) (N = 442) (N = 888)
Age — yr 78.5±6.8 77.3±7.3 77.9±7.1
Female sex — no. (%) 172 (38.6) 171 (38.7) 343 (38.6)
Race — no. (%)†
White 415 (93.0) 416 (94.1) 831 (93.6)
Black 2 (0.4) 1 (0.2) 3 (0.3)
Not reported 29 (6.5) 25 (5.7) 54 (6.1)
CHA2DS2-VASc score‡ 5.2±1.5 5.1±1.6 5.2±1.5
HAS-BLED score§ 3.1±0.9 3.0±0.9 3.0±0.9
Diabetes — no. (%) 175 (39.2) 186 (42.1) 361 (40.7)
Hypertension — no. (%) 417 (93.5) 417 (94.3) 834 (93.9)
Dyslipidemia — no./total no. (%) 269/433 (62.1) 241/422 (57.1) 510/855 (59.6)
Smoking — no./total no. (%) 30/415 (7.2) 50/411 (12.2) 80/826 (9.7)
Coronary heart disease — no./total no. (%) 257/441 (58.3) 232/439 (52.8) 489/880 (55.6)
Cardiomyopathy — no./total no. (%) 78/439 (17.8) 75/438 (17.1) 153/877 (17.4)
Peripheral artery disease — no./total no. (%) 59/445 (13.3) 49/439 (11.2) 108/884 (12.2)
Stroke or transient ischemic attack — no./total no. (%) 142/445 (31.9) 151/442 (34.2) 293/887 (33.0)
History of major bleeding — no. (%)¶
BARC type 3a or 3b 132 (29.6) 138 (31.2) 270 (30.4)
BARC type 3c 56 (12.6) 53 (12.0) 109 (12.3)
Recurrent nonmajor bleeding — no./total no. (%) 46/404 (11.4) 56/408 (13.7) 102/812 (12.6)
Stage IV chronic kidney disease — no. (%)‖ 114 (25.6) 100 (22.6) 214 (24.1)

* Plus-minus values are means ±SD. Percentages may not total 100 because of rounding. IQR denotes interquartile range.
† Race was reported by the patient.
‡ CHA2DS2-VASc scores (an assessment of the risk of stroke among patients with atrial fibrillation) range from 0 to 9, with higher scores indi-
cating a higher risk of stroke.
§ HAS-BLED scores (range, 0 to 9, with higher scores indicating a greater risk of bleeding) reflect the risk of major bleeding among patients
with atrial fibrillation receiving anticoagulants.
¶ Bleeding is staged according to the Bleeding Academic Research Consortium (BARC; range, 0 [no bleeding] to 5 [fatal bleeding]). BARC
types 3a and 3b denote bleeding and decrease in hemoglobin level, and type 3c denotes intracranial, intraocular, or intraspinal bleeding.
‖ Stage IV kidney disease is characterized by an estimated glomerular filtration rate of 15 to 29 ml per minute per 1.73 m2 of body-surface area.

n engl j med 394;13 [Link] April 2, 2026 1275

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 1276
Table 2. Primary and Secondary End Points.*

Adjusted Difference in Restricted

Left Atrial Appendage Closure Physician-Directed Best Medical Care Mean Survival Time in Years
End Point (N = 446) (N = 442) (95% CI)‡
Primary end point — no. of patients with event/no. of patient-yr 155/920.8 (16.8) 127/957.0 (13.3) −0.36 (−0.70 to −0.01)
of follow-up (incidence per 100 patient-yr)‡
Per-protocol analysis — no. of patients 411 392
The

Per-protocol analysis — no. of patients with event/no. of patient-yr 144/870.0 (16.6) 108/864.5 (12.5) −0.40 (−0.76 to −0.04)
of follow-up (incidence per 100 patient-yr)
Secondary end points — no. of patients with event/no. of patient-yr
of follow-up (incidence per 100 patient-yr)
Systemic embolism 3/1042.7 (0.3) 1/1045.4 (0.1) −0.02 (−0.08 to 0.03)
Ischemic or hemorrhagic stroke 27/1019.0 (2.6) 27/1015.1 (2.7) 0.01 (−0.17 to 0.20)

n engl j med 394;13

Major bleeding 70/941.5 (7.4) 61/978.7 (6.2) −0.12 (−0.40 to 0.15)
Cardiovascular or unexplained death 99/1045.2 (9.5) 81/1045.4 (7.7) −0.19 (−0.50 to 0.11)
Death from any cause 155/1045.2 (14.8) 141/1045.4 (13.5) −0.13 (−0.48 to 0.21)

[Link]
n e w e ng l a n d j o u r na l

Myocardial infarction 14/1021.8 (1.4) 20/1016.9 (2.0) 0.07 (−0.08 to 0.22)

of

Transient ischemic attack 9/1029.1 (0.9) 10/1028.6 (1.0) 0.02 (−0.10 to 0.13)
Ischemic stroke 18/1022.8 (1.8) 15/1022.8 (1.5) −0.04 (−0.18 to 0.11)

April 2, 2026
Hemorrhagic stroke 10/1039.5 (1.0) 13/1037.4 (1.3) 0.04 (−0.08 to 0.17)
Stroke or systemic embolism 29/1016.5 (2.9) 28/1015.0 (2.8) −0.00 (−0.19 to 0.19)
m e dic i n e

Hospitalization for bleeding or cardiovascular event 284/535.9 (53.0) 250/633.9 (39.4) −0.44 (−0.80 to −0.09)
Major adverse cardiac and cerebrovascular event 92/994.1 (9.3) 81/988.6 (8.2) −0.11 (−0.41 to 0.19)

Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
* The widths of the confidence intervals have not been adjusted for multiplicity and may not be used in place of hypothesis testing.
† To account for competing events, the restricted mean survival time was calculated as the area under 1 minus the cumulative-incidence curve.
‡ The primary end point was a patient-relevant composite end point of stroke (ischemic or hemorrhagic), systemic embolism, major bleeding (BARC type 3 or higher), or cardiovascular
or unexplained death, assessed in a time-to-event analysis. P = 0.44 for noninferiority.

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
 Left Atrial Appendage Closure in Atrial Fibrillation

(96.4%), respectively. In the device group, peripro- Discussion

cedural complications in the first 7 days after im-
plantation occurred in 5.7% and consisted primar- In this multicenter randomized trial, a strategy
ily of major bleeding (in 18 patients). There were of catheter-based left atrial appendage closure
two periprocedural deaths and one device emboli- was not noninferior to physician-directed best
zation that led to surgical retrieval of the device.
medical care with regard to a composite end
Device leaks were reported in 20 patients (device point of stroke, systemic embolism, major bleed-
leak of >5 mm in 3 patients) (Table S16). ing, or cardiovascular or unexplained death in
patients with atrial fibrillation and high risk of
Safety stroke and bleeding. Left atrial appendage closure
Serious adverse events were reported in 368 pa- was associated with a higher risk of a composite
tients (82.5%) in the device group and 342 (77.4%) primary end-point event (stroke, systemic embo-
in the medical-therapy group (Table S17). Addi- lism, major bleeding, or cardiovascular or unex-
tional adverse events are provided in Table S18. plained death) over a median follow-up of 3 years.

A Primary End Point

100
90
Cumulative Incidence (%)

80
70 P=0.44 for noninferiority
60 Left atrial appendage closure
50
40
30 Physician-directed best medical care
20
10
0
0 1 2 3 4 5 6
Years since randomization
No. at Risk
Physician-directed best medical care 442 306 203 136 77 40 7
Left atrial appendage closure 446 304 202 117 71 33 9

B Noncardiovascular Death
100 20
90 Physician-directed best medical care
15
Cumulative Incidence (%)

80
70 Left atrial appendage closure
10
60
50 5
40
0
30
0 1 2 3 4 5 6
20
10
0
0 1 2 3 4 5 6
Years since randomization
No. at Risk
Physician-directed best medical care 442 306 203 136 77 40 7
Left atrial appendage closure 446 304 202 117 71 33 9

Figure 2. Incidence of Primary End-Point Event and Noncardiovascular Death.

The primary end point was a composite of stroke (ischemic or hemorrhagic), systemic embolism, major bleeding,
or cardiovascular or unexplained death, assessed in a time-to-event analysis.

n engl j med 394;13 [Link] April 2, 2026 1277

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 3. Procedural Outcomes and Periprocedural Complications with Left Atrial Appendage Closure.

Total
Outcomes or Complication (N = 446)
Patients with attempted device implantation — no. (%) 421 (94.4)
Technical and procedural success of device implantation — no./total no. (%)
Device success when deployed and implanted in the correct position 414/421 (98.3)
Technical success with peridevice leak ≤5 mm and no device-related complications 411/421 (97.6)
Procedural success 406/421 (96.4)
Leak up to day 7 or hospital discharge — no. of patients
Total         20
Leak <3 mm 16
Leak 3 to 5 mm 1
Leak >5 mm 3
Periprocedural complications by day 7 day or by hospital discharge — no. of patients
Pericardial tamponade* 5
Major bleeding leading to transfusion (BARC types 3–5) 18
Device embolization, surgical removal 1
Procedure-related transient ischemic attack 1
Peripheral embolism 1
Death within 7 days after implantation 2

* Four patients with pericardial tamponade were treated with pericardiocentesis, and one was treated surgically.

Catheter-based left atrial appendage closure conditions who are at a high risk for bleeding,
was originally developed for stroke prevention in this patient group is particularly vulnerable to
patients with atrial fibrillation and high risk of periprocedural complications and early nonproc-
bleeding who were considered to be unsuitable edural major bleeding.24,25 In our trial, no reduc-
for long-term anticoagulation,21 largely on the tion in major bleeding events was apparent in the
basis of the observation that most cardiac throm- device group as compared with the medical-
bi are detected in the left atrial appendage.22 Over therapy group, although the trial was not pow-
the past decade, procedural safety has improved ered to analyze individual components of the
and initial data from moderate-sized trials have primary end point. In the medical-therapy group,
suggested potential noninferiority when com- 13.8% of the patients had a major bleeding event,
pared with anticoagulation.10,13,16,17 Moreover, in which is a higher percentage than those report-
LAAOS III (Left Atrial Appendage Occlusion ed in large DOAC studies and reflects the high
Study III), surgical removal of the left atrial ap- bleeding risk in our trial and other recent trials of
pendage reduced the risk of ischemic stroke or left atrial appendage closure.26
systemic embolism when performed concomitant A pooled, patient-level analysis of random-
with open-heart surgery, a finding that confirmed ized clinical trials showed an increase in bleeding
the concept of stroke prevention with left atrial rates within the first 6 months after left atrial
appendage closure.23 In the present trial, the fre- appendage closure27 than after discontinuation of
quency of ischemic stroke was lower in the device adjunctive anticoagulation and dual antiplatelet
group than would be expected on the basis of the therapy. The present trial replicates this pattern,
CHA2DS2-VASc score of 5.2 without anticoagula- with the incidence of bleeding highest in the first
tion, which is consistent with a stroke-reducing 6 months after device implantation. Early bleed-
effect of left atrial appendage closure. ing that occurs after left atrial appendage closure
Although catheter-based left atrial appendage but is unrelated to the procedure has recently
closure might appear to be a good option in older been associated with dual antiplatelet therapy
patients and patients with coexisting medical and also with higher mortality.28 Ongoing trials

1278 n engl j med 394;13 [Link] April 2, 2026

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 Left Atrial Appendage Closure in Atrial Fibrillation

aim to address bleeding by comparing different Among patients with atrial fibrillation at
periprocedural antithrombotic strategies,29,30 and high risk for stroke and high risk for bleeding,
device coatings may also mitigate the use of in- left atrial appendage closure was not noninferior
tense antithrombotic treatment after left atrial to physician-directed best medical care with re-
appendage closure. gard to a composite end point of stroke, sys-
The recent OPTION (Comparison of Antico- temic embolism, major bleeding, or cardiovascu-
agulation with Left Atrial Appendage Closure lar or unexplained death at a median follow-up
after Atrial Fibrillation Ablation) trial13 investi- of 3 years.
gated whether left atrial appendage closure can Supported by a public grant from the German Center for Car-
reduce bleeding risk associated with oral antico- diovascular Research (DZHK).
Disclosure forms provided by the authors are available with
agulation while maintaining a low risk of stroke the full text of this article at [Link].
in patients with atrial fibrillation undergoing A data sharing statement provided by the authors is available
catheter ablation. Left atrial appendage closure with the full text of this article at [Link].
was associated with fewer non–procedure-related
major or clinically relevant nonmajor bleeding Author Information
Ulf Landmesser, M.D.,1-3 Carsten Skurk, M.D.,1,2 Paulus Kirchhof,
events and was noninferior to oral anticoagula- M.D.,4-6 Thorsten Lewalter, M.D.,7 Johannes Hartung, M.D.,1,2
tion with respect to a composite of death from Andi Rroku, M.D.,1 Burkert Pieske, M.D.,8 Johannes Brachmann,
any cause, stroke, or systemic embolism at 36 M.D.,9 Ibrahim Akin, M.D.,10 Claudius Jacobshagen, M.D.,11 Benja-
min Meder, M.D.,12,13 Andreas Zeiher, M.D.,14,15 Stefan D. Anker,
months. However, the patient population had a M.D.,2,16 Holger Thiele, M.D.,17 Stefan Blankenberg, M.D.,4,5 Stef-
lower risk of bleeding than the patients in our fen Massberg, M.D.,18,19 Heribert Schunkert, M.D.,19,20 Norbert
trial. Frey, M.D.,12,13 Alexander Joost, M.D.,5,21 Martin Bergmann, M.D.,22
Ralph Stephan von Bardeleben, M.D.,23 Tim Friede, Ph.D.,24,25 Mar-
Ongoing clinical trials are comparing catheter- ius Placzek, Ph.D.,24,25 Anna Suling, Ph.D.,26 Karl Georg Haeusler,
based left atrial appendage closure to oral anti- M.D.,27 Matthias Endres, M.D.,2,28 Karl Wegscheider, Ph.D.,5,26
coagulation in patients with atrial fibrillation at Leif‑Hendrik Boldt, M.D.,2,29 and Ingo Eitel, M.D.5,21
intermediate risk for stroke22 as well as strate- 1
Department of Cardiology, Angiology, and Intensive Care
gies combining left atrial appendage closure Medicine, Deutsches Herzzentrum der Charité, Campus Benja-
min Franklin, Charité University Medicine Berlin, Berlin; 2 Ger-
with anticoagulation in patients at high risk for man Center for Cardiovascular Research (DZHK) Partner Site
stroke.31 These trials will help to refine appropri- Berlin, Berlin; 3 Friede Springer Cardiovascular Prevention Cen-
ate patient selection and treatment strategies in ter, Berlin; 4 Department of Cardiology, University Heart and Vas-
cular Center Hamburg, Hamburg, Germany; 5 DZHK Partner Site
the future. Hamburg-Kiel-Lübeck, Hamburg, Germany; 6 Institute of Cardio-
Our trial has several limitations. First, the trial vascular Sciences, University of Birmingham, Birmingham, United
was not powered to detect differences in the indi- Kingdom; 7 Department of Cardiology and Intensive Care Unit,
Hospital Munich South, Peter Osypka Heart Center, Munich,
vidual components of the primary end point. Germany; 8 Division of Cardiology, Department of Internal
Second, the trial was conducted in Germany, a Medicine, University Medicine Rostock, Rostock, Germany;
factor that resulted in a predominantly White
9
Department of Cardiology, Klinikum Coburg, Coburg, Germany;
10
Department of Cardiology, Hemostaseology, and Medical In-
population, and findings may not be generaliz- tensive Care, University Medical Center Mannheim, Heidelberg
able to other ethnic groups or health care sys- University, Mannheim, Germany; 11 Department of Cardiology,
tems. Third, the end points were based on the use Intensive Care Medicine, and Angiology, Vincentius-Diakonissen
Hospital, Karlsruhe, Germany; 12 Precision Digital Health, De-
of two approved left atrial appendage closure de- partment of Internal Medicine III, Cardiology, University Hei-
vices or therapy with DOACs in eligible patients. delberg, Heidelberg, Germany; 13 DZHK Partner Site Heidel-
Results may change as best medical therapy and berg, Heidelberg, Germany; 14 Department of Medicine III,
Goethe University, Frankfurt am Main, Germany; 15 DZHK Part-
devices evolve. Fourth, complete blinding of data ner Site Rhein-Main, Frankfurt, Germany; 16 Institute of Health
reviewed by the clinical events committee with Center for Regenerative Therapies, Charité University Medi-
respect to procedural and device complications cine, Berlin; 17 Department of Internal Medicine–Cardiology,
Heart Center at Leipzig University, Leipzig, Germany; 18 Depart-
was not possible. Fifth, more patients were lost to ment of Medicine I, LMU University Hospital, Ludwig Maximil-
follow-up in the medical-therapy group, whereas lian University of Munich, Munich, Germany; 19 DZHK Partner
the device group had more withdrawals, and pri- Site Munich, Munich Heart Alliance, Munich, Germany; 20 De-
partment of Cardiovascular Diseases, German Heart Center,
mary end-point events may have been masked School of Medicine and Health, TUM University Hospital,
by noncardiovascular death. Sixth, as a result of Technical University of Munich, Munich, Germany; 21 University
blind interim analyses and findings in similar Heart Center Lübeck, Medical Clinic II (Cardiology–Angiology–
Intensive Care Medicine), Lübeck, Germany; 22 Department of Car-
trials, the sample size was reduced twice on the diology, Asklepios Klinik Altona, Hamburg, Germany; 23 Depart-
basis of new pooled data. ment of Cardiology, University Medical Center, Mainz, Germany;

n engl j med 394;13 [Link] April 2, 2026 1279

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.
 Left Atrial Appendage Closure in Atrial Fibrillation

24
Department of Medical Statistics, University Medical Center University Hospital Ulm, Ulm, Germany; 28 Department of
Göttingen, Göttingen, Germany; 25 DZHK Partner Site Lower Neurology, Charité University Medicine Berlin, Berlin; 29 De-
Saxony, Göttingen, Germany; 26 Department of Medical Biome- partment of Cardiology, Angiology, and Intensive Care Medi-
try and Epidemiology, University Medical Center Hamburg- cine, Deutsches Herzzentrum der Charité; Campus Virchow,
Eppendorf, Hamburg, Germany; 27 Department of Neurology, Berlin.

References
1. Hindricks G, Potpara T, Dagres N, et appendage closure with the WATCHMAN 21. Sievert H, Lesh MD, Trepels T, et al.
al. 2020 ESC guidelines for the diagnosis device: peri-procedural outcomes from the Percutaneous left atrial appendage trans-
and management of atrial fibrillation de- EWOLUTION registry. Eur Heart J 2016;​37:​ catheter occlusion to prevent stroke in
veloped in collaboration with the European 2465-74. high-risk patients with atrial fibrillation:
Association for Cardio-Thoracic Surgery 12. Landmesser U, Tondo C, Camm J, et early clinical experience. Circulation 2002;​
(EACTS): the task force for the diagnosis al. Left atrial appendage occlusion with 105:​1887-9.
and management of atrial fibrillation of the AMPLATZER Amulet device: one-year 22. Landmesser U, Skurk C, Tzikas A,
the European Society of Cardiology (ESC) follow-up from the prospective global Falk V, Reddy VY, Windecker S. Left atrial
Developed with the special contribution of Amulet observational registry. EuroInter- appendage closure for stroke prevention
the European Heart Rhythm Association vention 2018;​14(5):​e590-e597. in atrial fibrillation: current status and
(EHRA) of the ESC. Eur Heart J 2021;​42:​ 13. Wazni OM, Saliba WI, Nair DG, et al. perspectives. Eur Heart J 2024;​45:​2914-32.
373-498. Left atrial appendage closure after abla- 23. Whitlock RP, Belley-Cote EP, Paparella
2. Joglar JA, Chung MK, Armbruster AL, tion for atrial fibrillation. N Engl J Med D, et al. Left atrial appendage occlusion
et al. 2023 ACC/AHA/ACCP/HRS guideline 2025;​392:​1277-87. during cardiac surgery to prevent stroke.
for the diagnosis and management of atrial 14. Landmesser U, Skurk C, Kirchhof P, N Engl J Med 2021;​384:​2081-91.
fibrillation: a report of the American Col- et al. Catheter-based left atrial appendage 24. Sanjoy S, Choi Y-H, Holmes D, et al.
lege of Cardiology/American Heart Associ- CLOSURE in patients with atrial fibrilla- Comorbidity burden in patients undergo-
ation Joint Committee on Clinical Practice tion at high risk of stroke and bleeding as ing left atrial appendage closure. Heart
Guidelines. Circulation 2024;​149(1):​e1- compared to best medical therapy: ratio- 2021;​107:​1246-53.
e156. nale and design of the prospective ran- 25. El Nasasra A, Brachmann J, Lewalter
3. Verheugt FWA, Granger CB. Oral anti- domized CLOSURE-AF trial. Am Heart T, et al. Comparison in patients < 75 years
coagulants for stroke prevention in atrial J 2026;​292:​107273. of age - versus - those > 75 years on one-
fibrillation: current status, special situa- 15. Steffel J, Verhamme P, Potpara TS, et year-events with atrial fibrillation and left
tions, and unmet needs. Lancet 2015;​386:​ al. The 2018 European Heart Rhythm As- atrial appendage occluder (from the Pro-
303-10. sociation Practical Guide on the use of spective Multicenter German LAARGE
4. Connolly SJ, Ezekowitz MD, Yusuf S, non-vitamin K antagonist oral anticoagu- Registry). Am J Cardiol 2020;​136:​81-6.
et al. Dabigatran versus warfarin in pa- lants in patients with atrial fibrillation. 26. Lakkireddy D, Thaler D, Ellis CR, et al.
tients with atrial fibrillation. N Engl J Med Eur Heart J 2018;​39:​1330-93. 3-Year outcomes from the Amplatzer Am-
2009;​361:​1139-51. 16. Holmes DR, Reddy VY, Turi ZG, et al. ulet Left Atrial Appendage Occluder ran-
5. Patel MR, Mahaffey KW, Garg J, et al. Percutaneous closure of the left atrial ap- domized controlled trial (Amulet IDE).
Rivaroxaban versus warfarin in nonvalvu- pendage versus warfarin therapy for pre- JACC Cardiovasc Interv 2023;​16:​1902-13.
lar atrial fibrillation. N Engl J Med 2011;​ vention of stroke in patients with atrial 27. Price MJ, Reddy VY, Valderrábano M,
365:​883-91. fibrillation: a randomised non-inferiority et al. Bleeding outcomes after left atrial
6. Granger CB, Alexander JH, McMurray trial. Lancet 2009;​374:​534-42. appendage closure compared with long-
JJV, et al. Apixaban versus warfarin in pa- 17. Holmes DR Jr, Kar S, Price MJ, et al. term warfarin: a pooled, patient-level
tients with atrial fibrillation. N Engl J Med Prospective randomized evaluation of the analysis of the WATCHMAN randomized
2011;​365:​981-92. Watchman Left Atrial Appendage Closure trial experience. JACC Cardiovasc Interv
7. Giugliano RP, Ruff CT, Braunwald E, device in patients with atrial fibrillation 2015;​8:​1925-32.
et al. Edoxaban versus warfarin in patients versus long-term warfarin therapy: the 28. Mesnier J, Cruz-González I, Guedeney
with atrial fibrillation. N Engl J Med 2013;​ PREVAIL trial. J Am Coll Cardiol 2014;​64:​ P, et al. Early nonprocedural bleeding af-
369:​2093-104. 1-12. ter left atrial appendage occlusion. JACC
8. Landmesser U, Holmes DR Jr Left 18. Holmes DR Jr, Doshi SK, Kar S, et al. Cardiovasc Interv 2024;​17:​1765-76.
atrial appendage closure: a percutaneous Left atrial appendage closure as an alter- 29. Freixa X, Cruz-González I, Cepas-Guillén
transcatheter approach for stroke preven- native to warfarin for stroke prevention in P, et al. Low-dose direct oral anticoagula-
tion in atrial fibrillation. Eur Heart J 2012;​ atrial fibrillation: a patient-level meta- tion vs dual antiplatelet therapy after left
33:​698-704. analysis. J Am Coll Cardiol 2015;​65:​2614- atrial appendage occlusion: the ADALA
9. Reddy VY, Doshi SK, Kar S, et al. 23. randomized clinical trial. JAMA Cardiol
5-year outcomes after left atrial append- 19. Kuramatsu JB, Gerner ST, Schellinger 2024;​9:​922-6.
age closure: from the PREVAIL and PRO- PD, et al. Anticoagulant reversal, blood pres- 30. Reinhardt SW, Gibson DN, Hsu JC, et
TECT AF trials. J Am Coll Cardiol 2017;​70:​ sure levels, and anticoagulant resumption al. Anticoagulation alone vs anticoagu-
2964-75. in patients with anticoagulation-related in- lation plus aspirin or DAPT following
10. Osmancik P, Herman D, Neuzil P, et tracerebral hemorrhage. JAMA 2015;​ 313:​ left atrial appendage occlusion. J Am Coll
al. 4-Year outcomes after left atrial ap- 824-36. Cardiol 2024;​84:​889-900.
pendage closure versus nonwarfarin oral 20. Osmancik P, Herman D, Neuzil P, et 31. McIntyre WF, Jolly SS, Whitlock RP,
anticoagulation for atrial fibrillation. J Am al. Left atrial appendage closure versus Healey JS. Introducing the fourth left
Coll Cardiol 2022;​79:​1-14. direct oral anticoagulants in high-risk pa- atrial appendage occlusion study. JACC
11. Boersma LVA, Schmidt B, Betts TR, et tients with atrial fibrillation. J Am Coll Clin Electrophysiol 2025;​11:​1041.
al. Implant success and safety of left atrial Cardiol 2020;​75:​3122-35. Copyright © 2026 Massachusetts Medical Society.

1280 n engl j med 394;13 [Link] April 2, 2026

The New England Journal of Medicine is produced by NEJM Group, a division of the Massachusetts Medical Society.
Downloaded from [Link] at Univ Hospital Birmingham NHS Foundation Trust on April 27, 2026.