Dialysis patients with type 2 diabetes who were started on a GLP-1 drug were about 12 percent less likely to have a major heart event than similar patients put on an older diabetes pill. That is the headline from a new analysis of 1,688 closely matched patient pairs, and it lands in a group of people who are almost never studied.
People on dialysis with diabetes carry some of the highest heart and death risk in all of medicine. The large trials that made GLP-1 drugs famous mostly left them out. So doctors have been prescribing into an evidence gap. The new study, published June 8 in the Clinical Kidney Journal ↗, tries to close part of it by mining real-world records rather than running a trial.
What the study did
The authors pulled de-identified records from the TriNetX US Collaborative Network covering 2013 to 2022. They found patients on dialysis with type 2 diabetes who were newly started on a GLP-1 receptor agonist, the injectable class that includes semaglutide ↗, dulaglutide ↗, and liraglutide ↗. Each one was matched to a similar patient started instead on a dipeptidyl peptidase-4 inhibitor, an older oral drug that nudges the body's own incretin hormones. Matching on age, other illnesses, and medications left 1,688 pairs. The idea is to make the two groups look alike on everything except which drug they got, so the difference in outcomes can be pinned on the drug rather than on who was sicker to begin with.
The GLP-1 receptor agonists are peptides. They mimic a gut hormone that tells the pancreas to release insulin after a meal. Unlike some diabetes drugs, they are broken down by enzymes rather than cleared by the kidneys, which is part of why they can be used at all in people whose kidneys have failed.
The numbers
Across the board the GLP-1 group did better. Major adverse cardiovascular events fell 12 percent (hazard ratio 0.88, 95 percent confidence interval 0.78 to 0.99). All-cause death dropped 16 percent (HR 0.84, 0.72 to 0.99). Heart attacks fell 16 percent (HR 0.84, 0.70 to 0.99) and heart failure 13 percent (HR 0.87, 0.78 to 0.98). Hospital stays and emergency visits were lower too.
The single biggest drop was not a heart number at all. It was sepsis, the body's runaway response to infection, down 19 percent (HR 0.81, 0.71 to 0.92). Infection is a leading cause of death on dialysis, and a diabetes drug is not the tool anyone would reach for to prevent it. That the largest signal in a cardiovascular study points at infection is the part worth arguing about, and the part the authors cannot fully explain.
What to hold back
This is not a randomized trial. It is a matched comparison of people who happened to get one drug or another, and the usual caveat applies: doctors may have steered GLP-1 drugs toward patients they judged healthier, and no amount of matching fully erases that. Every confidence interval here also brushes right up against the no-difference line, so the effects are real but not enormous. The study looked at the class, not head-to-head between individual drugs, so it cannot say whether semaglutide beats liraglutide in this group.
What it does do is put numbers where there were none. Dialysis is one of the few remaining corners of diabetes care where the GLP-1 evidence was close to blank, and a 1,688-pair signal pointing consistently in one direction is a reason to run the actual trial.
The drugs in question all act on the same target, the GLP-1 receptor ↗, which anchors a large share of the metabolic cards on this platform. The dialysis result widens the population where that target looks worth pressing, and it does so in patients the original outcome trials skipped.