Patients who were already on a GLP-1 drug when they had a stroke severe enough to need a clot pulled out of a brain artery developed aspiration pneumonia at more than twice the rate of similar patients who were not on the drugs. In a single-center cohort, the rate was 26.8 percent in the GLP-1 group versus 10.3 percent in everyone else, an absolute gap of 16.5 percentage points. After the two groups were matched to look alike on the variables that usually muddy these comparisons, being on a GLP-1 was tied to roughly triple the odds of the lung infection (adjusted odds ratio 3.25, 95 percent confidence interval 1.43 to 7.40, p equals 0.008).
Aspiration pneumonia is what happens when stomach contents go up the esophagus and down into the lungs instead, usually while a patient is sedated or has a depressed gag reflex, and then the lung gets infected. GLP-1 receptor agonists slow how fast the stomach empties. That is part of how they curb appetite, and it is the exact mechanism behind a three-year argument over whether patients should stop the drugs before planned procedures.
The new numbers come from a retrospective cohort published online June 5 in the Journal of NeuroInterventional Surgery ↗. The authors pulled every patient who had an endovascular thrombectomy for acute ischemic stroke at one tertiary care center between January 2020 and May 2025, 740 patients in all, of whom 41 (5.5 percent) were on a GLP-1 drug when they arrived. Endovascular thrombectomy is the emergency procedure where a catheter is threaded up into the brain and the clot is mechanically removed. The primary outcome was aspiration pneumonia within 30 days. The researchers used propensity score matching, the standard technique for making two real-world groups comparable, to check that the signal survived once age, weight, and the other obvious differences were balanced out. It did: the unmatched odds ratio was 3.19 and the matched one barely moved, to 3.25.
Why this setting is different
The notable part is where the signal showed up. In elective settings, the same fear has mostly failed to materialize. A meta-analysis we covered in May found that holding GLP-1s before an upper endoscopy ↗ nearly quadrupled the odds of arriving with food still in the stomach, yet aspiration itself did not budge. A separate cohort of 366,476 surgical patients in JAMA Network Open ↗ found no postoperative aspiration pneumonia difference either. The reading at the time was that the prevention was doing its job: when a procedure is planned, the patient is told to fast, the drug can be paused, and a stomach that turns out to be full can simply send the case home to be rebooked.
A stroke does not offer any of that. The patient arrives with whatever is in their stomach, the clock is measured in minutes of dying brain, and nobody is canceling a thrombectomy to wait out a pharmacokinetic tail. This paper is the first to look in the one setting where every workaround that kept the elective numbers flat is unavailable, and the risk that elective data could afford to hold did not.
The risks that did not move
Two other outcomes stayed put. Symptomatic bleeding into the brain, the complication clinicians most fear during a thrombectomy, was no more common in the GLP-1 group (4.9 versus 9.1 percent, a difference that did not reach significance). Ninety-day disability, scored on the standard modified Rankin Scale, was also statistically indistinguishable between the groups. So the finding is narrow and specific. The drugs did not make the procedure more dangerous or the recovery worse. They raised the odds of one downstream infection that has a known set of countermeasures: airway protection, suction, head-of-bed positioning, swallowing assessment before the first meal.
That distinction is the authors' actual conclusion. The fix is not to withhold thrombectomy from stroke patients on GLP-1s, which would trade a treatable pneumonia for irreversible brain loss. The fix is to assume a full stomach and protect the airway accordingly.
The same population is where the drugs have looked their best. A separate matched cohort found that GLP-1 use before a thrombectomy stroke cut three-year mortality to about a third ↗ of the unexposed rate. Put the two together and the picture for a patient on semaglutide ↗ or tirzepatide ↗ who has a severe stroke is not a tradeoff to agonize over. They are likely to survive the next three years at a far better rate, and the immediate job is to manage a peri-procedural aspiration risk that is real but routine to prevent.
The caveats are the ones every single-center retrospective carries. Forty-one exposed patients is a small denominator, the wide confidence interval reflects it, the exposure flag is class-level rather than drug-specific, and a second center could land a different number. What it should not do is reopen the question of whether to treat the stroke. It should change what the airway team assumes before they start.