Patients who were already taking a GLP-1 drug when they had a knee replaced went on to have fewer infections in the new joint, 1.5 percent versus 2.0 percent over the following two years. Everything mechanical about the implant, the loosenings, the fractures, the revisions, looked the same in both groups. That split is the finding, from a study published September 6 in the Journal of Orthopaedic Surgery ↗.
The complication in question is a periprosthetic joint infection, or PJI. It is one of the worst things that can happen after a joint replacement. Bacteria settle on the metal and plastic hardware, where the immune system and antibiotics struggle to reach them, and clearing it often means more surgery to wash out or replace the implant. Diabetes and obesity both raise the risk, which is why anything that improves blood sugar or trims weight before surgery is of interest to orthopedic surgeons.
GLP-1 drugs do both. Semaglutide ↗ and tirzepatide ↗, the drugs behind Ozempic, Wegovy, Mounjaro, and Zepbound, are prescribed for diabetes and weight loss, and a patient who is better controlled and lighter going into an operation is, in theory, a patient less likely to get infected. The question this study asked is whether that theory shows up in the records.
What the database showed
The researchers used TriNetX, a network of electronic health records, to find adults who had a total knee replacement and had filled a GLP-1 prescription in the year before surgery. They matched each GLP-1 user one to one against a non-user with a similar profile of age, sex, and the comorbidities that drive complications, then followed both groups for two years.
One methodological wrinkle sets this apart from earlier database work. The team verified laterality, meaning they checked that the complication they counted actually happened on the same knee that got the implant. That sounds trivial, but a records database is full of people who have had both knees done, or who have arthritis coded on the other side, and a naive query can pin a problem on the wrong joint. Confirming the infection and the surgery were on the same side gives cleaner attribution than most prior studies managed.
The result was narrow and consistent. GLP-1 users had a periprosthetic joint infection rate of 1.5 percent against 2.0 percent in non-users, about a quarter lower in relative terms (risk ratio 0.74). Among the diabetic patients specifically, the gap was wider: 1.7 percent versus 2.5 percent, roughly a third lower (risk ratio 0.68). Revision, mechanical loosening, periprosthetic fracture, and other mechanical failures showed no significant difference in either the full cohort or the diabetics.
Why the flat outcomes matter
The selective pattern is what makes this more than another database association. When a records study shows a drug beating its comparator on every single outcome at once, the usual explanation is not the drug but who was selected to receive it. Healthier, better-resourced, more engaged patients tend to be the ones on an expensive weekly injectable, and that pattern manufactures a benefit across the board. Here the board is not uniformly better. Infection moved. The purely mechanical outcomes, which have little to do with blood sugar or wound healing, did not. A blanket healthy-user effect would be expected to drag those down too.
That is a point in the finding's favor, not a clean bill of health. Residual confounding can be infection-specific. The same things that get a patient prescribed a GLP-1 drug, tighter diabetes control, fewer obesity-related wound problems, closer medical follow-up, are exactly the things that also lower infection risk on their own, and no amount of matching on baseline codes fully erases them. The study is retrospective, the groups were matched rather than randomized, and the absolute difference is half a percentage point overall. The broader literature is unsettled: some systematic reviews find a GLP-1 infection benefit after knee replacement, and at least one recent review found no mid-term benefit at all.
Both semaglutide and tirzepatide act on the GLP-1 receptor ↗, with tirzepatide adding a second action on the GIP receptor ↗. The same selective-outcome logic showed up when GLP-1 users had fewer 90-day complications before joint replacement ↗ in a meta-analysis earlier this year. This new piece narrows the claim to one complication, infection, and adds a cleaner check on which knee it happened to. It is a reason to run the randomized trial of GLP-1 drugs as pre-surgical optimization, not evidence the trial has already passed.