GLP-1s Go Under the Knife: What New Surgical Data Mean for Ozempic-Class Users

For years, the perioperative debate around GLP-1 receptor agonists has centered on a single anesthesia concern: delayed gastric emptying and the aspiration risk it creates on induction. That conversation isn't over. But the newer data look further down the timeline — at 90-day complications, readmissions, revisions — and they suggest the risk profile is not uniform across the body. It's joint-specific, and probably tissue-specific.

Start with the good news. A retrospective national-database analysis of 14,065 patients with type 2 diabetes undergoing primary total hip arthroplasty between 2016 and 2021 used propensity score matching to compare 812 GLP-1 users against 3,248 non-users. The headline: no significant differences in 90-day surgical or medical complication rates, and no difference in one-year revision or periprosthetic joint infection rates. If anything, the non-GLP-1 group was more likely to have extended hospital stays of three days or longer (odds ratio 1.25).

That's a clean signal for the hip. For a 45-year-old diabetic patient staring down a THA, the current best evidence does not flag the GLP-1 as a reason to pause or panic.

Now the part that should make you slow down. A separate 2025 analysis using the TriNetX database looked at total shoulder arthroplasty — anatomic and reverse — from 2010 to 2023. After 1:1 propensity matching, 1,259 GLP-1 users were compared to 1,259 controls. In the 90 days after surgery, GLP-1 users showed significantly higher rates of deep vein thrombosis (1.6% vs. 0.9%, OR 3.0), myocardial infarction (1.6% vs. 0.9%, OR 2.84), pneumonia (3.34% vs. 1.50%, OR 2.25), transfusion (7.1% vs. 4.3%, OR 1.7), and readmission (8.1% vs. 5.2%, OR 1.6). These held up after Bonferroni correction at a strict P<.005 threshold.

Why the shoulder and not the hip? The authors don't fully explain it, and neither will I — that's the honest answer at this stage. Possibilities range from differences in patient positioning and pulmonary mechanics during shoulder surgery, to the underlying populations selecting into each procedure, to residual confounding the propensity model couldn't catch. What's clear is that the assumption "GLP-1s look fine in knees and hips, so they're fine everywhere" doesn't hold.

There's a third piece of 2025 data that doesn't fit neatly into a surgical checklist but matters enormously for anyone making a perioperative plan. A cohort study of 69 individuals with type 2 diabetes tracked what happened when patients had to stop dulaglutide during a global supply shortage. Within three months of discontinuation, average HbA1c rose from 7.0% ± 0.9% to 8.1% ± 1.4%, and fasting glucose climbed from 129 ± 31 to 156 ± 50 mg/dL. Switching to DPP-4 inhibitors or SGLT2 inhibitors did not fully compensate.

For a surgical patient, that's a real problem. Elevated perioperative glucose is itself an independent risk factor for wound complications and infection. If your surgeon asks you to hold your GLP-1 in the days before a procedure — a reasonable request given the aspiration question — and you can't easily restart it afterward because of a supply gap, you may be trading one risk for another. This is a conversation to have on the front end, not the day before.

None of these studies are randomized trials. They are retrospective, propensity-matched cohort analyses — strong enough to flag patterns, not strong enough to dictate a protocol. The evidence here is moderate, not settled. Mechanism is largely inferred, not demonstrated. And the shoulder data, in particular, will need replication and prospective work before anyone can say with confidence whether the drug, the population, or the procedure is doing the driving.

What does change today is the quality of the conversation you can have with your own clinician. If you're on a GLP-1 and you have elective surgery on the calendar, the procedure type matters. The continuity of your therapy on the back end matters. And the default assumption that "these drugs are universally safe around surgery" is no longer adequate — not because the drugs are dangerous, but because the picture is more granular than the marketing.