Knee replacement study links GLP-1 use to a slightly lower rate of joint infection

Patients who had a prescription for a GLP-1 medicine in the year before a total knee replacement had a slightly lower rate of infection around the new joint than matched patients who did not, according to a database study published September 6, 2026 in the Journal of Orthopaedic Surgery.

The rates were 1.5 percent and 2.0 percent. That is a difference of half a percentage point.

The topic matters to a lot of people who follow obesity treatment. Knee osteoarthritis is common in people carrying extra weight, knee replacement is one of the operations they are most likely to face, and a large number of them are now taking a GLP-1 medicine such as semaglutide, tirzepatide, liraglutide, or dulaglutide for weight or blood sugar. Infection around an implanted joint is one of the outcomes surgeons work hardest to avoid, because it often means further surgery.

Infection rate was 1.5 percent with a GLP-1 prescription and 2.0 percent without

The researchers searched TriNetX, a database that pools de-identified electronic health records from many health systems. They found patients who had a total knee replacement and who had used a GLP-1 receptor agonist within the year before the operation, then matched each one to a patient who had not, pairing them on age, sex, and other medical conditions that could affect the result. They followed both groups for two years.

Periprosthetic joint infection, an infection in the tissue around the artificial joint, occurred in 1.5 percent of the GLP-1 group and 2.0 percent of the comparison group. The risk ratio was 0.74 with a 95 percent confidence interval of 0.57 to 0.96. A risk ratio below 1 means lower risk, and because the range stays below 1 the difference reached statistical significance, with a p value of 0.02.

Among patients who also had diabetes, the gap was slightly wider: 1.7 percent versus 2.5 percent, a risk ratio of 0.68 with a confidence interval of 0.52 to 0.89.

The authors also checked which knee each event happened to, so that a complication in the opposite knee would not be counted against the operated one. They describe this step as a correction to how earlier database studies of this question were done.

Loosening, fracture, and revision surgery showed no difference

The study looked at more than infection. It also counted revision surgery, mechanical loosening of the implant, fracture of the bone around the implant, and other mechanical complications. None of those differed significantly between the two groups, in the full population or in the diabetes subgroup.

That pattern is worth noting. If the medicines were simply a marker for healthier patients, a difference might be expected across most outcomes. Seeing a difference only in infection, and not in the mechanical problems, is at least consistent with the authors' reading that any effect is specific rather than general. It is not proof of that.

A prescription in the record is not proof a patient took the medicine

This is a retrospective look at records that were collected for treating patients, not for answering this question. It shows an association. It does not show that taking a GLP-1 before knee replacement prevents infection.

Patients were classified by whether a GLP-1 appeared in their record in the year before surgery. That does not establish which medicine, at what dose, for how long, or whether the patient actually took it near the time of the operation. It also does not capture how much weight anyone lost, or what their blood sugar control looked like going into surgery, either of which could be doing the work.

People who fill prescriptions and keep appointments tend to do better after surgery for reasons that have nothing to do with the medicine. Matching on recorded conditions reduces that problem but does not remove it.

The absolute difference is also small. Half a percentage point means that in this data roughly 200 people would need to be in the GLP-1 group rather than the comparison group for one additional infection to be avoided, and that estimate carries its own uncertainty.

Nothing here is a reason to start, stop, or change the timing of a GLP-1 medicine around surgery. Separate guidance exists on holding these medicines before anesthesia because of concerns about stomach emptying, and that is a question for the surgical and anesthesia team, not something this study addresses.