Semaglutide and knee osteoarthritis: what the trials show

A systematic review of GLP-1 receptor agonists (the drug class that includes semaglutide and liraglutide) for osteoarthritis, published as an accepted manuscript in BMC Musculoskeletal Disorders on 23 October 2026 by authors from Oxford, Imperial College London and the UK Defence Medical Services.

It included two randomised controlled trials and five observational studies, 160,089 patients in all. The studies were too different to pool, so results were described study by study. The literature search was run on 11 February 2025, so anything published since then isn't included.

Semaglutide reduced knee pain in one trial. STEP 9 randomised 407 people with knee osteoarthritis and obesity (mean age 56, mean BMI 40, 82% women) at 61 sites in 11 countries, two to semaglutide 2.4 mg weekly for every one to placebo, for 68 weeks. Both groups also had diet and exercise counselling. WOMAC pain, scored 0 to 100, fell 14.1 points more with semaglutide (95% CI 8.3 to 20.0), above the 11-point difference the review treats as clinically meaningful. Body weight fell 13.7% with semaglutide and 3.2% with placebo.

Liraglutide didn't reduce pain in the other trial. Participants first lost more than 5% of their body weight by diet, then were randomised to liraglutide or placebo. Pain didn't improve significantly.

The observational studies linked GLP-1 use with fewer joint replacements, less pain and slower cartilage loss. In one large health-records database study, knee replacement happened at 0.75 times the rate in GLP-1 users (hazard ratio 0.75, 95% CI 0.63 to 0.89) and hip replacement at 0.60 times the rate (0.45 to 0.79). In another cohort, WOMAC pain was 3.4 points lower in GLP-1 users, and cartilage thinned 0.02 mm a year more slowly, measured on MRI in 188 users and 1,267 non-users. The review rated these studies generally at serious risk of bias, and two at critical risk; those two weren't used to support its conclusions.

Certainty. The review rated the evidence low for pain, moderate for physical function and patient global assessment, low for joint structure and very low for quality of life. Adverse events couldn't be graded because they were reported inconsistently, but gastrointestinal effects were the most common in every study that reported them.

The review had no funding, and its authors declared no competing interests.

STEP 9 was funded by Novo Nordisk, which makes semaglutide, and four of its authors were from Novo Nordisk.

The randomised evidence for pain rests on one manufacturer-funded trial. The only other randomised trial, of liraglutide, found no pain benefit. One observational cohort also reported less pain, at 3.4 points on WOMAC.

STEP 9 can't separate the drug from the weight loss. People on semaglutide lost about four times as much weight as those on placebo. The review describes weight loss as "an important mediator", and neither trial was designed to test whether the drug does anything to the joint beyond reducing load. In one observational study, weight loss accounted for about a third of the link with fewer knee surgeries.

The cartilage and joint replacement findings are associations. People prescribed GLP-1 drugs differ from people who aren't, in weight, health, access to care and how closely they're followed up. The studies couldn't fully adjust for that, which is why the review rated them at serious or critical risk of bias.

The trial population was narrow. STEP 9 enrolled people with a mean BMI of 40, mostly women, so the result may not apply to people with knee osteoarthritis at a lower weight.

Semaglutide reduced knee osteoarthritis pain in one large Novo Nordisk-funded trial, in people who lost around 14% of their body weight. Evidence that GLP-1 drugs protect cartilage or prevent joint replacement is observational and at high risk of bias. So far, it seems like the pain benefit can't be separated from the weight loss.