Cardiologists at the University of Adelaide ran a small randomized trial of liraglutide ↗, the daily GLP-1 receptor agonist sold as Saxenda for obesity, in 59 overweight or obese adults preparing for catheter ablation of atrial fibrillation. Three months of the drug, added to standard risk-factor modification, almost halved the rate of AF coming back at one year. The fat depot the trial was built to shrink, the layer of epicardial adipose tissue wrapped around the left atrium, did not separate between the two arms.

The numbers, published online in JACC: Clinical Electrophysiology ↗: 81 percent of patients on liraglutide were free of AF or atrial flutter at twelve months, against 54 percent of patients on risk-factor modification alone (log-rank p = 0.007). In a logistic-regression model that controlled for change in left-atrial epicardial adipose tissue (LAEAT), liraglutide cut the odds of a recurrence by a factor of five (odds ratio 0.19, 95 percent CI 0.05 to 0.73). The trial was registered as LEAF, NCT03856632.

That is a strong clinical effect for a three-month run-in. It is also a clinical effect that the trial's prespecified mechanism does not explain.

The mechanism that didn't move

LEAF was designed around a specific bet. AF ablation outcomes are well-correlated with obesity and with the volume of epicardial fat immediately adjacent to atrial myocardium. Risk-factor modification programs can shrink that fat. GLP-1 receptor agonists drive weight loss and have been shown in smaller series to reduce epicardial fat specifically. The hypothesis: liraglutide added to risk-factor modification should reduce LAEAT more than risk-factor modification alone, and that bigger LAEAT reduction should drive a better ablation outcome.

The first half of the chain broke. Pooled across both arms, LAEAT volume came down (median -1.0 mL, p = 0.02) and weight came down (-2.8 kg, p < 0.001). Neither difference between groups was significant. Three months of liraglutide on top of risk-factor modification did not buy a measurable LAEAT or weight advantage at the time of the procedure.

The second half held anyway. Liraglutide-arm patients were less likely to be in AF or flutter at one year, the effect was large, and the confidence intervals do not include the null. The trial's authors call this out plainly in the conclusion: the proposed mechanism did not separate the arms, but the clinical endpoint did. They reach for pleiotropic GLP-1 receptor effects to bridge the gap.

That is a polite way of saying nobody knows.

What could close the gap

LAEAT volume is one number that captures the fat depot. The trial also measured EAT density, a tissue-property measure that has been moving up the cardiology literature as a marker of the metabolic activity of adipose tissue rather than its bulk. Change in EAT density was associated with lower one-year recurrence (odds ratio 0.55 per unit, 95 percent CI 0.36 to 0.84). That hints at the actual handle. Liraglutide may be reshaping the activity profile of the fat without shrinking it enough, in three months, to register on CT volumetry. Whether GLP-1 receptor agonism reaches atrial myocardium directly, through ion-channel modulation or anti-inflammatory autonomic effects, is open. Preclinical work has shown GLP-1R expression in atrial tissue and effects on conduction. LEAF gives that biology a clinical hook without confirming it.

There is also the question of why three months of liraglutide produced an effect this large in a 59-patient study while large observational analyses of GLP-1 use in unselected patients have shown smaller or absent effects on incident AF. The plausible reconciliation: incident AF in obese diabetics is a different problem from recurrence after ablation in patients with predominantly persistent AF. The substrate is already remodeled, the trigger landscape is different, and the question liraglutide is being asked in LEAF is closer to "does this slow remodeling enough to make the procedure stick" than to "does this prevent fibrillation from ever happening." The two literatures should not be expected to converge.

The size of the trial limits what it proves

Fifty-nine patients, one center, primary endpoint negative, secondary endpoint positive. The cleanest reading is hypothesis-generating, not practice-changing. AF recurrence is a real endpoint that the operators looking at the LEAF arms care about, but a secondary endpoint at this n-size will be wrong some fraction of the time. The investigators do not claim a definitive result and the field should not either.

What LEAF buys is a clear signal worth powering. A multicenter RCT with AF recurrence as the prespecified primary endpoint, sized to detect a hazard ratio in the range LEAF observed, is the next move. Whether semaglutide ↗ or tirzepatide ↗ reproduce the signal at a longer pre-procedure run-in is another open question. Both are now off-label common in this population because of weight loss. Whether any of them is doing real work on atrial substrate is what the next trial needs to settle.

Platform note

The GLP-1R target page ↗ on peptidemodel hosts the canonical receptor and 200+ candidate ligands. None of the cards model cardiac-tissue selectivity, but the LEAF finding is exactly the kind of result that argues for that property to become a first-class card attribute. If the next confirmatory trial holds, the receptor's relevance to atrial electrophysiology stops being a footnote and becomes a design target.