PDF: CLOSURE-AF
(2026)Objective
To evaluate whether catheter-based left atrial appendage closure is noninferior to physician-directed best medical care (predominantly DOACs) for preventing stroke, systemic embolism, major bleeding, or cardiovascular death in AF patients at high risk for both stroke and bleeding.
Study Summary
• Stroke rates were similar between groups: ischemic or hemorrhagic stroke 2.6 vs 2.7 per 100 patient-years (27 events each) — device provides comparable stroke prevention to DOACs.
• Major bleeding was NOT reduced: 7.4 vs 6.2 per 100 patient-years, driven by 52 procedure-related major bleeding events offsetting any long-term bleeding reduction after the periprocedural period.
• Periprocedural complications in 5.7%: including 2 deaths, 5 tamponades, 18 major bleeds requiring transfusion, 1 device embolization.
• First large RCT comparing LAA closure vs DOAC-dominant control arm — prior trials (PROTECT-AF, PREVAIL) compared against warfarin only.
Intervention
Catheter-based LAA closure (Watchman/FLX 54.3%, Amplatzer/Amulet 41.8%, LAmbre 3.9%) with post-implant DAPT ≥3 months then aspirin to 6 months, vs. physician-directed best medical care (85.1% DOAC, 3.2% VKA, 7.5% no antithrombotics).
Inclusion Criteria
Adults with AF (paroxysmal/persistent/permanent) and CHA2DS2-VASc ≥2 plus high bleeding risk (HAS-BLED ≥3, prior intracranial/intraspinal/intraocular bleeding, prior BARC 3a/3b hemorrhage, Stage IV CKD eGFR 15-29, recurrent bleeding precluding anticoagulation, or contraindications to long-term anticoagulation).
Study Design
Arms: LAA Closure (N=446) vs. Best Medical Care (N=442)
Patients per Arm: LAA Closure: 446, Best Medical Care: 442
Outcome
• Ischemic/hemorrhagic stroke: 2.6 vs 2.7/100 pt-yr (similar).
• Major bleeding: 7.4 vs 6.2/100 pt-yr (procedure-related bleeding 52 events in device group).
• CV/unexplained death: 9.5 vs 7.7/100 pt-yr (numerically higher with device).
• All-cause death: 14.8 vs 13.5/100 pt-yr.
• Device success: 98.3% (414/421). Procedural success: 96.4%.
• Periprocedural complications: 5.7% (2 deaths, 5 tamponades, 18 major bleeds).
Bottom Line
LAA closure was NOT noninferior to best medical care (predominantly DOACs) in high-risk AF patients: composite endpoint 16.8 vs 13.3 per 100 patient-years (RMST difference -0.36 years; 95% CI -0.70 to -0.01; P=0.44 for noninferiority). Stroke rates were similar (2.6 vs 2.7/100 pt-yr), but LAA closure had more major bleeding (7.4 vs 6.2/100 pt-yr) driven by periprocedural events (52 procedure-related bleeds). First large RCT comparing LAA closure vs DOAC-dominant control.
Major Points
- FAILED noninferiority: composite 16.8 vs 13.3/100 pt-yr (RMST diff -0.36 yr; 95% CI -0.70 to -0.01; P=0.44 for NI).
- Stroke rates similar: 2.6 vs 2.7/100 pt-yr (27 events each group) — device provides comparable stroke prevention.
- Major bleeding was NOT reduced with device: 70 events (7.4/100 pt-yr) vs 61 (6.2/100 pt-yr). Of the 70 device-group bleeds, 18 were procedure-related and 52 were NON-procedure-related; the 18 periprocedural events offset any long-term bleeding reduction.
- Very high-risk cohort: mean age 78, CHA2DS2-VASc 5.2, HAS-BLED 3.0. Prior major bleeding in 30.4%. Stage IV CKD in 24.1%.
- First RCT vs DOAC-dominant control: 85.1% of medical group on DOACs (prior trials PROTECT-AF/PREVAIL compared vs warfarin).
- Periprocedural complications within 7 days post-implant occurred in 5.7% of patients. Table 3 breakdown: 5 pericardial tamponades (4 treated with pericardiocentesis, 1 surgically), 18 major bleeding events requiring transfusion, 1 device embolization requiring surgical removal, 1 procedure-related TIA, 1 peripheral embolism, and 2 deaths within 7 days of implantation.
- Procedural success 96.4%. Device success 98.3% (414/421). Watchman 54.3%, Amplatzer 41.8%, LAmbre 3.9%.
- CV/unexplained death numerically higher with device: 9.5 vs 7.7/100 pt-yr (RMST -0.19; 95% CI -0.50 to 0.11).
- Sample size was reduced twice via prespecified blinded interim analyses and emerging data from similar trials: 912 patients enrolled vs 1,586 originally planned; 282 primary events observed vs 467 needed at 80% power. Recruitment was halted because interim data suggested possible overpowering, not for efficacy/safety concerns.
- Key implication: in DOAC era, LAA closure does not offer a net clinical benefit over medical therapy, even in high-bleeding-risk patients.
- Despite NI failure on the primary composite, device-group patients had SIGNIFICANTLY MORE hospitalizations for bleeding or CV events (53.0 vs 39.4 per 100 pt-yr; RMST difference -0.44 years, 95% CI -0.80 to -0.09). This reinforces that LAA closure does not reduce cardiovascular burden in this high-risk population.
Study Design
- Study Type
- Pragmatic, prospective, open-label, multicenter, randomized, controlled trial with blinded outcome adjudication (PROBE design)
- Randomization
- Yes
- Blinding
- Open-label; blinded outcome adjudication by independent clinical events committee. 1:1, stratified by site.
- Sample Size
- 888
- Follow-up
- Median 3.0 years (IQR 1.7-4.7); maximum 6.7 years.
- Centers
- 42
- Countries
- Germany
Primary Outcome
Definition: Composite: stroke, systemic embolism, major bleeding (BARC ≥3), or CV/unexplained death
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 127 events; 13.3/100 pt-yr | 155 events; 16.8/100 pt-yr | - (RMST diff -0.36 yr (-0.70 to -0.01)) | 0.44 for noninferiority (FAILED) |
Limitations & Criticisms
- FAILED noninferiority — confidence interval did not cross zero in favor of device.
- Sample size reduced twice based on prespecified blinded interim analyses and pooled data from similar trials: original plan 1,586 patients for 467 primary events at 80% power, but recruitment was stopped after 912 were enrolled because interim data suggested the trial might already be overpowered. Observed 282 events; failure of noninferiority could reflect either a true negative result or a power shortfall given the sample size reduction.
- Germany only (93.6% White) — limited generalizability.
- Open-label — clinical events committee could not be blinded to procedural complications.
- Periprocedural bleeding (18 events in the device group) drove the lack of bleeding reduction — 52 additional device-group bleeds were non-procedural. The device's long-term bleeding benefit may improve with operator experience and refined peri-implant antithrombotic strategies.
- More lost to follow-up in medical group (5.2% vs 2.2%); more withdrawals in device group (12.6% vs 10.9%).
- 18 crossovers from medical to LAA closure.
- Only Watchman and Amplatzer devices — newer devices may perform differently.
- High event rate in medical group (13.3/100 pt-yr) reflects genuinely high-risk population.
- Early termination after LAAOS III publication reduced sample size.
- Per-protocol sensitivity analysis confirmed the primary result (RMST -0.40 years, 95% CI -0.76 to -0.04) — also failed noninferiority.
Citation
N Engl J Med 2026. DOI: 10.1056/NEJMoa2513310.