Inborn genetic differences point the same way the trials did: turning up the GLP-1 receptor lowers the risk of heart failure, and about a quarter of that benefit appears to run through less of a single inflammation enzyme. That is the claim in a genetic analysis published online September 11 in Heart and Lung ↗, and the size of the effect is worth reading closely before anyone calls it settled.
GLP-1 drugs (semaglutide, sold as Ozempic and Wegovy; tirzepatide, sold as Mounjaro and Zepbound) act on the GLP-1 receptor, a docking site on cells that these molecules switch on. Large trials have shown the drugs cut heart-failure risk, but the reason has stayed murky. Is it the weight loss, the blood-sugar control, something the drug does to the heart or blood vessels directly, or some mix? The trials show the outcome without isolating the cause.
Using genes as a natural experiment
The authors did not dose anyone. They used a method called Mendelian randomization. It leans on a quirk of biology: the gene variants you inherit are dealt out roughly at random at conception, before any lifestyle or disease can confound them. Say people who carry variants that raise a gene's activity also tend to get a disease less often. That pattern is hard to explain away as reverse causation or lifestyle, because the genes came first. It is the closest thing to a randomized trial you can run on existing genetic databases.
Here the researchers picked genetic variants tied to higher expression of the GLP-1 receptor gene, drawn from the eQTLGen consortium. Those variants stand in for switching the receptor on. As a sanity check, the same variants tracked strongly with lower type 2 diabetes risk (odds ratio 0.82, meaning roughly 18 percent lower odds). That is exactly what you would expect if the instrument really reflects GLP-1 receptor activity, and it is the strongest number in the paper.
The heart-failure signal is real but thin
Genetically proxied GLP-1 receptor expression was tied to lower heart-failure risk, with an odds ratio of 0.93. In plain terms, about 7 percent lower odds. The catch is the confidence interval: 0.87 to 0.998, with a p-value of 0.044. The upper bound is a whisker away from 1.0, the line where an effect disappears. This is a statistically significant result that barely clears the bar, not a thick, durable signal. Read it as directional support, not proof of magnitude.
Then the second step. The authors screened 95 inflammation-related blood markers to see which, if any, sat on the path between the receptor and the heart. One came through: matrix metalloproteinase-1, or MMP-1, an enzyme that chews up and remodels the scaffolding between cells. Lower MMP-1 accounted for about 28 percent of the total protective effect, and an independent dataset put the figure at 21 percent. Both held up statistically.
What it adds, and what it does not
The interesting idea here is mechanistic. If the finding holds, GLP-1 receptor activation may protect the heart partly by damping an inflammatory, tissue-remodeling enzyme, not only by trimming weight and glucose. That is a testable, specific handle on why these drugs help the heart.
The honest limits are large. A genetic proxy is a small, lifelong nudge in receptor activity, not a weekly pharmacologic dose. The numbers do not translate into how much a drug would help a given patient. Mediation is a statistical inference layered on top of an already-thin causal estimate. The 28 percent through MMP-1 leaves roughly three-quarters of the effect unexplained, and external validation shrank the mediated share to about a fifth. No drug was given, no patient was followed, and the heart-failure estimate itself sits at the edge of significance.
This is the second time in recent coverage that a genetic causal analysis has been used to probe the incretin drugs. An earlier one tied the GIP arm of these drugs to markers of slower aging ↗ using the same family of methods. Genetics can nominate a mechanism and argue for causation in a way a simple correlation cannot. It cannot tell you the dose, the timeline, or whether the enzyme is a lever you can actually pull. For that, someone has to measure MMP-1 in people taking the drugs.