Patients who were taking a GLP-1 drug before a hip or knee replacement had fewer complications in the three months after surgery, according to a new pooled analysis of 96,356 patients. Every complication category the authors measured moved in the same direction, which is both the reason the paper is interesting and the reason to read it carefully.
The analysis ↗ was published August 1 in The Bone & Joint Journal. A team led from the University of Oxford's Nuffield Department of Orthopaedics, with a co-author at the Hospital for Special Surgery in New York, pooled ten matched cohort studies. Those studies covered primary total hip arthroplasty (30,350 operations) and total knee arthroplasty (66,606). In all of them, patients had been prescribed a GLP-1 receptor agonist for type 2 diabetes or for obesity before going under the knife.
The headline number is the 90-day surgical complication rate. Combining hips and knees, GLP-1 users were about a quarter less likely to hit one of the tracked surgical problems (pooled risk ratio 0.73). That group of problems includes periprosthetic joint infection, wound breakdown, a fracture around the new implant, a bleed, or a surgical-site infection. Medical complications and hospital readmissions within 90 days followed the same slope, each roughly a fifth lower (risk ratios 0.78 and 0.79).
Hips moved, knees mostly did not
The cleaner signal is in the split. When the authors separated the two operations, the hip cohort carried almost all of the benefit. Hip-replacement patients on a GLP-1 drug had significantly fewer 90-day surgical complications (risk ratio 0.63, so about 37 percent lower), fewer medical complications (0.55), and fewer readmissions (0.82). Knee-replacement patients, by contrast, showed a significant drop only in readmissions (0.77); their surgical and medical complication rates did not separate from the controls in a statistically reliable way.
Why the two joints would respond differently is not something the data can answer. It could be anatomy, or that hip and knee patients differ in ways the matching did not capture, or noise from a smaller pooled effect. The paper does not claim to know.
Why the all-favorable pattern is the catch
The mechanism is plausible on its face. Obesity and diabetes both raise the risk of osteoarthritis and worsen surgical outcomes, and GLP-1 drugs push on exactly those two levers by lowering blood sugar and body weight. A patient who loses weight and tightens glucose control before an elective operation is, in theory, a lower-risk patient on the table.
But the study is built from matched cohorts, not randomized trials, and the authors are blunt about what that costs. Under the GRADE system, which rates how much confidence to place in each result, the certainty of evidence was very low for every outcome except healthcare cost. Two of the ten studies carried a serious risk of bias and the other eight a moderate risk.
The deeper problem is the very pattern that makes the result eye-catching. When every outcome favors one group, that is also the fingerprint of confounding by indication and healthy-user bias. Patients who are prescribed a GLP-1 drug, tolerate it, and stay on it long enough to reach an elective joint replacement are, on average, healthier and better resourced than patients who are not. Their better outcomes may trail their better baseline rather than the drug. The same records-based pattern showed up in a study of GLP-1 use before carotid stenting ↗. There, an apparent cut in cardiovascular events turned out to be driven entirely by lower all-cause mortality, the classic tell of a healthier group.
The one result the authors graded with moderate rather than very low confidence was cost. GLP-1 use before surgery was associated with substantial healthcare savings, which is the finding a hospital system can act on even while the clinical numbers stay soft.
None of the pooled studies named a specific agent as the driver; the class is dominated by semaglutide ↗ and tirzepatide ↗, both of which hit the GLP-1 receptor ↗. The useful reading is narrow. This is a reason to run the randomized test of GLP-1 as pre-surgical preparation, especially for hip replacement, not evidence that the test has already been passed.