When the CLOSURE-AF trial landed in the New England Journal of Medicine, Eric Topol noticed something striking: the journal published an accompanying editorial — and a cardiologist reached the opposite conclusion via a separate interview. Two careful reads of the same dataset, different maps.
The numbers: LAAC failed to prove noninferior to standard medical therapy at three years. Stroke rates were essentially identical — 2.6 versus 2.7 per 100 patient-years in the device and medical therapy groups, according to the ACC Journal Scan. Serious adverse events were higher in the device arm — 82.5 percent versus 77.4 percent. The adjusted hazard ratio was 1.28, just outside the noninferiority boundary of 1.3. The trial enrolled 912 patients at 42 centers in Germany, making it the first adequately powered, non-industry-funded randomized comparison of LAAC against physician-directed best medical therapy, predominantly modern anticoagulants.
Petr Widimský of Charles University took the straightforward read. Best medical therapy should remain the standard for most older patients with atrial fibrillation at high risk of both stroke and bleeding, he wrote in NEJM. The trial showed the device could not demonstrate it was even as good as pills in a population selected to receive it. Andrew Goldsweig of the University of Massachusetts-Baystate landed differently. The enrollment window began in 2018, capturing older devices and less experienced operators. "If you were to do the trial with current devices today, you would have very different results," Goldsweig told TCTMD. His broader argument: the field has grown ahead of the evidence regardless. More than 600,000 LAAC devices have been placed worldwide since 2010, roughly 30,000 per year in the United States alone — a volume built largely on earlier trials conducted against warfarin, not modern anticoagulants.
Gregory Katz, a cardiac electrophysiologist at NYU Langone Health, drew a narrower conclusion from the same data. His argument is not that CLOSURE-AF was a triumph. He is explicit about the noninferiority failure and the higher adverse event rate. His point is patient-specific: those who genuinely cannot take anticoagulants — documented prior gastrointestinal bleeding, prior intracranial hemorrhage, transfusion-dependent anemia. For that population, the risk calculus flips. "A medication works everywhere. A procedure works in one place." The trial's composite endpoint was driven substantially by cardiovascular death, with periprocedural bleeding accounting for roughly a quarter of the safety gap, as Jeff Healey, the ACC discussant, noted at the meeting. That does not straightforwardly condemn a device used in a narrower, higher-risk group.
The tension both perspectives dance around is what CLOSURE-AF reveals about the bar for procedural alternatives in 2026. The medical therapy arm was not warfarin — it was mostly direct oral anticoagulants. The low stroke rates across CLOSURE-AF, OPTION, and PRAGUE-17, as Katz noted, "underscore how good systemic treatments have become." That is an achievement for anticoagulation — and a problem for anyone trying to justify a device with upfront procedural risk when a pill already does the job.
The competing interpretations also rest on different theories about where strokes originate. A high CHADS-VASc score reflects multiple systemic vascular risks: carotid plaque, small vessel disease, sources outside the left atrial appendage. Katz argues LAAC can only reduce stroke risk attributable to the appendage itself. Goldsweig appears more skeptical that the device has demonstrated benefit even in that narrower context.
Widimský and Goldsweig both acknowledge what CHAMPION-AF — presented at the ACC annual meeting on March 28 — will not resolve. The WATCHMAN FLX device (Boston Scientific, funder) showed noninferiority to NOACs for the composite endpoint. But there were more ischemic strokes in the device arm: 45 versus 27. The trial's safety endpoint excluded periprocedural bleeding — precisely where LAAC's risk concentrates. Follow-up is shorter than CLOSURE-AF's three-year median. Whether device placement reduces bleeding over time requires longer observation.
CLOSURE-AF has honest limitations. Enrollment began in 2018 with superseded hardware. The noninferiority margin required a hazard ratio no worse than 1.3; the device came in at 1.28. What the trial does not answer is whether a better-device, better-operator, carefully selected-patient version would arrive at a different place. That is the bet LAAC proponents are making. The evidence base for placing that bet remains incomplete.
Topol was right to flag the discordance. The disagreement between Widimský and Goldsweig — and Katz's third read — is not a failure of the trial. It is a window into how medicine navigates genuine ambiguity. The terrain is real. The maps differ.
