A retrospective cohort of more than sixteen thousand adults with untreated brain aneurysms found that those on GLP-1 receptor agonists were about half as likely to rupture over five years. For the smaller group whose aneurysms ruptured anyway, the bleeds were less destructive: less blood spilling into brain tissue, less into the ventricles, less of the artery spasm that drives most post-rupture damage.
The analysis appeared online May 21 in Stroke ↗. Authors pulled records from the TriNetX US Collaborative Network across 2016 to 2024, identifying adults newly diagnosed with an unruptured intracranial aneurysm (UIA) and propensity-matching 8,088 GLP-1 RA users against 8,088 nonusers on demographics, comorbidities, and medication use. Unruptured intracranial aneurysms are weak outpouchings in brain arteries that affect up to one in twenty adults. Most never bleed. When they do, the result is a subarachnoid hemorrhage, which kills or disables a large fraction of patients before they reach a neurosurgery suite.
In the matched cohort, cumulative rupture rates split early and kept widening. At one year, 0.65 percent of GLP-1 RA users had ruptured versus 1.51 percent of nonusers. At three years, the gap was 1.18 versus 1.85. At five years, 1.49 versus 2.10. The proportional hazards model returned a hazard ratio of 0.52 (95% CI 0.38-0.69), a 48 percent reduction.
The authors then ran a second, smaller analysis. Among patients who had a rupture (n=298 after a separate match), 149 had been on a GLP-1 RA beforehand. They had less intraparenchymal hemorrhage (17.4 versus 33.6 percent, p<0.001), less intraventricular hemorrhage (10.1 versus 23.5 percent, p<0.001), and less clinically significant vasospasm (10.7 versus 24.2 percent, p=0.002). Thirty-day mortality was 9.4 percent in the prior-GLP-1 group versus 14.1 percent without, but with only 298 patients in this cohort the difference did not reach statistical significance (p=0.21).
What stands out is the layering. Most cardiovascular GLP-1 readouts work at one level. The drug reduces the rate of an outcome happening. Here the same exposure also tracks with a different outcome (severity) in patients who experienced the first event despite being on the drug. The proposed mechanisms are anti-inflammatory and vasculoprotective effects on the aneurysm wall itself. Inflammation drives wall remodeling, and GLP-1 receptor activation tones inflammation down in vessel-wall macrophages and endothelium. If the drug stabilizes the wall, both the rupture rate and the post-rupture vasospasm response could move in the same direction, which is what the data show.
This is a retrospective cohort matched on what TriNetX records, not on what unrecorded factors drive both diabetes-care intensity and cerebrovascular events. Aneurysm size, location, and morphology are not in the matching covariates. Patients prescribed GLP-1 RAs may be more closely watched, which can shift symptomatic-rupture ascertainment. The authors call the result hypothesis-generating, not practice-changing, and that framing is correct.
Liraglutide ↗ and semaglutide ↗ are the two agonists that dominate this exposure window, both indexed on the GLP-1R ↗ target page. The aneurysm result joins a string of non-glycemic GLP-1 readouts published this month, including the LEAF trial on atrial fibrillation recurrence after ablation and the propensity-matched cohort comparing bariatric surgery against GLP-1 RAs for idiopathic intracranial hypertension. The pattern across all three is that the receptor's vasculoprotective and anti-inflammatory tail is producing detectable clinical effects in indications the drug was never marketed for.
The Stroke paper does not propose a prospective trial yet. A registry-flagged cohort of UIA patients eligible for surveillance imaging would be a clean design. Randomize incidence, run severity as a hard secondary on whoever ruptures. The TriNetX numbers here are large enough to power that with a feasible sample size.