GLP-1s and Atrial Fibrillation: Reading a 42% Headline

A 2024 meta-analysis reported that semaglutide cut the risk of new atrial fibrillation by 42%.^[1] That is a striking number and it has traveled a long way. Before it travels any further, it is worth knowing exactly what kind of number it is — because it was assembled from ten trials, none of which set out to study atrial fibrillation at all.

What was actually pooled

The analysis brought together ten randomized trials covering 12,651 patients, 7,285 of them on semaglutide and 5,366 on placebo, in populations at high cardiovascular risk and largely with type 2 diabetes. Pooled with a random-effects model, the relative risk of incident atrial fibrillation was 0.58, with a 95% confidence interval of 0.40 to 0.85 and no measurable heterogeneity between studies.^[1]

The consistency is the impressive part. An I² of zero means the ten trials pointed the same way rather than one outlier dragging the average. Subgroup analysis found no difference between oral and injected semaglutide, and meta-regression found no influence from the proportion of patients with diabetes or from body mass index.^[1]

42% of what?

This is where a relative risk needs its partner. A 42% reduction tells you how much smaller the risk became; it says nothing about how large the risk was to begin with. Cut a small number by 42% and you have prevented very few events.

The published summary reports the relative risk without the absolute event counts, so we cannot give you the second number, and we are not going to estimate one. What we can say is the general shape: atrial fibrillation was an uncommon event in these populations over these timeframes, so the absolute difference behind that 42% is small.

A relative risk without a baseline is half a finding. It is also the half that gets quoted.

Why 'semaglutide' and not 'GLP-1 drugs'

The authors are careful about this and it is worth preserving. The class as a whole is regarded as neutral on arrhythmia risk; what the data suggest is a possible benefit specific to semaglutide.^[1]

Drug-specific effects inside a drug class are interesting for two opposite reasons. Sometimes they are real pharmacology — molecules in a class genuinely differ. Sometimes they are an artifact of which trials happened to be run, in which populations, with which reporting practices. A finding that appears in one member of a class and not the others deserves more scrutiny than a class-wide one, not less.

There is a plausible mechanism, which is not the same as evidence: obesity, sleep apnea and hypertension all drive atrial fibrillation, and these drugs improve all three. If the effect is real, it may be largely downstream of weight rather than anything specific to the heart. Our sleep apnea article covers the one of those three with a dedicated trial behind it.

What to do with this

• It is not a reason to start one of these drugs. No GLP-1 is approved for preventing atrial fibrillation, and evidence assembled from adverse-event reporting would not support that even if it were larger.
• It says little about people who already have AF. These trials counted new cases; the direction of the data is not toward harm, but what happens to an existing arrhythmia is a separate question.
• It does not replace anything. Anticoagulation decisions, rate and rhythm control belong to cardiology and are unaffected by this.
• Palpitations are still worth reporting. Heart rate increases are a known effect of this drug class and appear in the labels, as why a GLP-1 raises your heart rate explains; a new irregular heartbeat is a reason to call someone, not to read a meta-analysis.

The honest position is that this is a promising, internally consistent signal from data collected for other purposes. That is a good reason to run the trial that would answer it properly, and not yet a reason to change what anybody does.