One November study found 18% more bleeding when blood-thinner patients started a GLP-1. An October study found 21% fewer clots in nearly the same population. Both numbers are real. Here is how they fit together.
Roughly one in four people with atrial fibrillation also has type 2 diabetes. That overlap is where modern prescribing gets complicated: a blood thinner to prevent stroke, and increasingly a GLP-1 drug to handle blood sugar and weight. Until this fall, nobody had solid numbers on what happens when the two prescriptions land in the same routine. Within three weeks, two studies published in the same journal went looking. One found 18% more bleeding. The other found 21% fewer blood clots. Both findings are real, and the gap between them is the interesting part.
The question had been floating around pharmacies for years. GLP-1 drugs slow down how fast the stomach empties. Direct oral anticoagulants, the apixaban and rivaroxaban family, are swallowed tablets absorbed through the gut. Change the speed of the gut and you can change how much of the drug reaches the bloodstream. Researchers flagged this as a plausible interaction long before anyone had outcomes data. This fall, the outcomes data arrived.
The first paper was published November 3 in Diabetes, Obesity and Metabolism. A team led by Fang Du at the University of Florida used Medicare data to emulate a target trial, which is a way of asking what a randomized trial would have found when running one isn't practical. They identified people aged 65 and older with type 2 diabetes and atrial fibrillation who were already taking a direct oral anticoagulant, then compared those who started a GLP-1 drug with those who started a DPP-4 inhibitor, a different diabetes pill. After matching, 930 patients were in each group, with about a year of follow-up and a mean age near 75.
The headline result: bleeding was more common in the GLP-1 group. The hazard ratio was 1.18, with a 95% confidence interval of 1.03 to 1.36. In plain language, an 18% relative increase in bleeding events. That is a modest bump on a low baseline, and the study cannot tell you which bleeds were serious and which were not.
The other two results deserve equal billing. Thromboembolic events, the clots the blood thinner is there to prevent, showed no clear difference between groups. The hazard ratio was 1.50, but the confidence interval ran from 0.22 to 10.43, which is statistical territory where almost nothing can be concluded. Clots were too rare in this cohort for the comparison to mean much either way.
Then there is the finding nobody expected. All-cause mortality was substantially lower in the GLP-1 group: hazard ratio 0.58, confidence interval 0.36 to 0.95. About 42% fewer deaths from any cause over a year. The authors treat it carefully, as they should. It was not the study's main question, the interval is wide, and observational data can make healthy-user effects look like drug effects. But it lines up with what large randomized trials of GLP-1 drugs have shown in high-risk cardiovascular patients.
Three weeks earlier, on October 12, the same journal published a research letter from Shao-Wei Lo and colleagues at the Cleveland Clinic and Montefiore. They asked a neighboring question using the TriNetX network of 167 healthcare organizations, covering 2018 through 2024. Same population on paper: adults with type 2 diabetes and atrial fibrillation or flutter. Same comparison: new users of GLP-1 drugs against new users of DPP-4 inhibitors. This time the sample was far larger, 27,548 matched pairs.
Over one year, venous thromboembolism, the clot-in-the-leg or clot-in-the-lung family of events, happened in 2.07% of GLP-1 users versus 2.56% of DPP-4 users. That works out to a hazard ratio of 0.79, a 21% relative reduction. Pulmonary embolism dropped 24%, deep vein thrombosis dropped 19%, and ischemic stroke showed up less often too, 3.03% versus 3.36%. The pattern held when GLP-1 drugs were compared against SGLT2 inhibitors, sulfonylureas, and thiazolidinediones, and it held both in patients taking anticoagulants and in those who were not.
The authors' own framing is the right one: these findings suggest GLP-1 drugs may carry vascular benefits in this high-risk group, and they do nothing to change the established role of blood thinners in preventing AF-related stroke. It is a drug-selection signal, not a substitution signal.
More bleeding and fewer clots sound contradictory until you separate what the two studies measured. Hemorrhage and thrombosis are not mirror-image outcomes driven by one dial. They have overlapping but distinct biology, and the two teams used different data, different follow-up windows, and different definitions of events.
Several mechanisms can hold at once. GLP-1 drugs lower weight, blood pressure, and inflammatory markers, all of which plausibly reduce clot formation, and the October study's results fit that story. In the other direction, delayed gastric emptying can raise exposure to orally absorbed drugs, and a 2026 review of motility-mediated drug interactions laid out exactly this concern for anticoagulants. More drug on board, more bleeding risk. Both forces could be operating in the same patient at the same time.
There is also a mundane explanation that doesn't involve pharmacology at all. Starting a GLP-1 means more medical contact: more visits, more lab work, more attention. People who get more attention get more diagnoses recorded. That can inflate bleeding detection in one dataset while healthier behaviors quietly lower clot risk. The authors of both studies flag residual confounding as their main limitation.
For calibration, a July 2026 study of older adults on blood thinners found that adding amiodarone raised bleeding risk by about 30%, hazard ratio 1.30. In that light, an 18% increase sits within the range of bleeding bumps that show up whenever a new drug enters an anticoagulant's routine. And a large FAERS pharmacovigilance analysis published in May found no excess thrombotic reporting for GLP-1 drugs at all, which sits comfortably with the October clot data.
This is where we have to be careful, because we are not your prescriber and this is exactly the kind of decision that belongs between you and one. But there are a few things the data supports saying plainly.
Don't stop your blood thinner. Both research teams say the same thing: nothing here changes the established role of anticoagulation in atrial fibrillation. Stroke prevention is the point of the prescription, and the studies found no clot signal that should scare anyone off.
If you are starting a GLP-1 while on an anticoagulant, the practical move is telling your prescriber about the combination and watching for bleed signals, the usual list: black or tarry stools, unusual bruising, nosebleeds that won't stop, blood in urine. Not a panic list. A know-what-to-watch-for list.
The Florida team closes their paper with the right next question: studies that factor in body weight and renal function. Both change how anticoagulants behave in the body, and both change quickly on a GLP-1. A dose that fit at 110 kilograms may not fit at 95.
For anyone weighing whether to start a GLP-1 at all: this is not a reason to avoid one. If anything, the mortality signal, the clot reduction, and the cardiovascular trial record pull in the same direction. The bleeding signal is a monitoring note, not a stop sign.
Two observational studies, one question, two different answers, and the honest synthesis is that they are measuring different things. The fear that GLP-1 drugs would amplify clotting through anticoagulant interference looks weaker after October. The suspicion that they nudge bleeding risk up a notch looks supported after November. Both effects are small in absolute terms in a population where the underlying stroke risk is large. The most interesting number in either paper is still the mortality one, and it is the least proven.
Our June breakdown of GLP-1s and atrial fibrillation covered why these drugs seem to calm the arrhythmia itself. Our guide to peptides and blood thinners walks through the interaction basics. And our November look at GLP-1s before surgery covered another moment when these drugs meet other treatments. This post is about the third such moment, and it won't be the last.
Evidence grade: B-. Two large propensity-matched observational emulations in overlapping populations, one a peer-reviewed research letter and one a full paper, with consistent thromboembolic safety findings. Held back from A by claims-based outcome ascertainment, residual confounding, a secondary mortality result with a wide confidence interval, and an age-65-plus cohort that limits how far the numbers travel.
Sources: Du F et al., Diabetes Obes Metab, Nov 3 2026 (doi:10.1111/dom.71481, PMID 42977570). Lo SW et al., Diabetes Obes Metab, Oct 12 2026 (doi:10.1111/dom.71443, PMID 42864268). Motility-mediated drug interactions with GLP-1 receptor agonists, Transl Clin Pharmacol, Sep 2026 (PMID 42846711). FAERS thrombotic adverse event analysis, Curr Drug Saf, May 2026 (PMID 42227478). Amiodarone and DOAC bleeding risk, J Cardiovasc Electrophysiol, Jul 2026 (PMID 42418755).
This article is educational and is not medical advice. Oria is a research and education platform: we do not sell peptides, and nothing here should change a prescription you have been given. Talk to a physician or pharmacist before starting, stopping, or combining any medication.
