Of the 1,618 blood-clot reports filed against GLP-1 weight-loss and diabetes drugs in the United States over five years, the count came in lower than the rate for an average drug, not higher. If these medicines were quietly driving strokes and deep-vein clots, that is not what the national side-effect database recorded.
A pharmacovigilance analysis published online May 19 in Current Drug Safety ↗ went looking for a thrombotic signal, the technical word for a pattern of clotting reports that would flag a drug for closer scrutiny. It did not find one. The work matters because the fear is everywhere. Social posts and a handful of case reports have tied semaglutide ↗ and its relatives to clots, and patients ask their doctors about it. This is the first large attempt to check the question against the system built to catch exactly this kind of harm.
What FAERS can and cannot say
The FDA Adverse Event Reporting System, or FAERS, is a pile of voluntary reports. Doctors, patients, and companies file them when something goes wrong, so the database is good at spotting patterns and bad at proving cause. The standard tool is the reporting odds ratio, which asks a narrow question: are clots reported more often for this drug than for drugs in general? Above 1 is a warning. Below 1 is reassurance.
The researchers pulled every thrombotic report tied to GLP-1 receptor agonists between 2020 and 2025 and sorted them by type. Arterial clots, the kind behind most heart attacks and strokes, made up 55.3 percent. Venous clots, the leg and lung clots known as DVT and pulmonary embolism, were 23.2 percent, with the rest mixed. The overall reporting odds ratio was 0.67 (95 percent confidence interval 0.64 to 0.71), meaning clots were reported about a third less often than expected. Venous events were the quietest of all at 0.53.
Then the comparisons. Against orlistat, an older weight-loss pill, the GLP-1 drugs had roughly a quarter the odds of a venous-clot report (odds ratio 0.27). Against SGLT2 inhibitors, a common diabetes class, GLP-1 reports were lower across every clot type.
The within-class split, and the caveat
Inside the class, the drugs did not all look the same. Liraglutide ↗, the oldest daily injectable, carried the highest clotting reports; tirzepatide ↗, the newest dual-acting agent, carried the lowest. That gradient tracks roughly with how long each drug has been on the market and how many people take it, which is a reminder of what these numbers are not.
A reporting odds ratio below 1 does not mean a drug prevents clots. FAERS counts reports, not patients, and reporting is shaped by news cycles, lawsuits, and how long a drug has existed. A newer drug with fewer prescriptions and less litigation attention generates fewer reports almost mechanically. The authors say as much, calling the result reassuring rather than protective and warning that spontaneous reports cannot establish cause.
What the analysis does do is set the burden of proof where it belongs. The same database has surfaced real GLP-1 signals before, including a biliary signal that split five ways across the class ↗. The clotting fear, run through the same machinery, did not separate from background. That is not a clean bill of health. It is the absence of the alarm people expected to hear, which in drug safety is its own kind of finding.