GLP-1s and knee osteoarthritis pain: what the trial actually showed

Knee osteoarthritis and obesity travel together, and anyone who has lost a significant amount of weight with sore knees already knows the first part of this story. Less weight, less pain. What changed recently is that a drug trial put numbers on it, and a second wave of coverage started suggesting the medication itself might be doing something to the joint.

Both of those things can be partly true. They are also easy to overstate. This article separates what a randomized trial established, what observational data suggests, and what is currently only a laboratory finding.

The mechanical case is the strongest part of the story

Before any drug enters the picture, the load argument stands on its own. In a study of 142 older adults with knee osteoarthritis who completed an 18-month diet-and-exercise trial, researchers used 3D gait analysis to measure the forces moving through the knee while walking. Each kilogram of body weight lost was associated with about a 40.6-newton reduction in compressive force on the knee, roughly a fourfold reduction in load relative to the amount of weight lost.²

That multiplier is the whole reason weight loss is a first-line recommendation for knee osteoarthritis. It is not a small effect applied once. It applies to every step, several thousand times a day. If a GLP-1 medication produces 13 percent body weight loss in someone weighing 110 kilograms, that is roughly 14 kilograms, and by this relationship a large reduction in the force the joint absorbs with each stride.

Nothing about that mechanism is specific to GLP-1 medications. It would apply equally to weight lost through diet, surgery, or any other route. Keep that in mind for the rest of this article, because it is the baseline against which any additional drug effect has to be measured.

What the randomized trial measured, and in whom

The trial in question is STEP 9, published in the New England Journal of Medicine in 2024. It enrolled 407 adults with a body mass index of 30 or above and a clinical diagnosis of knee osteoarthritis with moderate radiographic changes, meaning Kellgren-Lawrence grade 2 or 3 in the target knee. Participants were randomly assigned to weekly semaglutide 2.4 mg or placebo, both alongside lifestyle intervention, for 68 weeks across 61 sites in 11 countries. Mean age was 56, mean baseline BMI was 40.3, and about 82 percent of participants were women.¹

At week 68, body weight fell 13.7 percent in the semaglutide group and 3.2 percent on placebo. WOMAC pain scores, a patient-reported osteoarthritis pain measure, fell 41.7 points with semaglutide and 27.5 points with placebo, from a mean baseline of 70.9. The SF-36 physical function score improved 12.0 points versus 6.5. All three differences were statistically significant.¹

Two details of that population matter more than they usually get credit for. First, everyone in the trial had obesity, with a mean BMI of 40.3. This is not evidence about knee pain in people at a normal weight who happen to take a GLP-1. Second, the osteoarthritis was moderate by imaging. People with severe, bone-on-bone joint destruction were not the population studied.

The placebo group is also worth sitting with. It improved by 27.5 points, roughly two-thirds as much as the semaglutide group, on a self-reported pain scale, in a trial where both arms received lifestyle intervention. Pain outcomes are subjective and respond to attention, expectation, and the passage of time. The between-group difference is what the trial establishes; the total improvement in the treatment arm is not all attributable to the drug.

How much of it is the weight, and how much is the drug

This is where the coverage gets ahead of the evidence. STEP 9 reported both a large weight loss and a pain benefit. It did not report an analysis isolating how much of the pain relief was carried by the weight change and how much, if any, came from a direct effect of the medication on the joint.¹ Given the size of the weight difference between the groups and the fourfold load relationship described above,² mechanical unloading is a sufficient explanation for what the trial found. That does not prove it is the only explanation. It means the trial cannot distinguish them.

There is real laboratory evidence for a direct pathway. A 2026 study in Cell Metabolism used a mouse model of osteoarthritis with obesity and a diet-controlled design intended to hold weight loss constant, and reported that semaglutide still reduced cartilage degeneration, osteophyte formation, and joint inflammation, acting through a GLP-1 receptor signaling pathway in cartilage cells.⁴ A 2025 systematic review of preclinical and human work describes a similar set of candidate mechanisms, including suppression of inflammatory signaling molecules such as TNF-alpha, interleukin-6, and interleukin-1 beta, while noting explicitly that pooled quantitative analysis was not possible because the number of studies was small, the study types were varied, and outcome measures were inconsistent.⁵

So the honest state of things is this. A direct anti-inflammatory or metabolic effect on joint tissue is biologically plausible and has support in animals and cells. Whether it contributes meaningfully to what a person with knee osteoarthritis feels, on top of the effect of carrying less weight, has not been demonstrated in a human trial designed to answer that question. Treating the mouse finding as if it were a human result is the single most common error in writing about this topic.

The knee replacement signal, and why it is observational

A 2026 retrospective analysis published in Regional Anesthesia & Pain Medicine drew on a large multinational health-record network to identify adults diagnosed with knee osteoarthritis between 2010 and 2024, then compared rates of knee replacement surgery in those treated with GLP-1 receptor agonists against matched patients who were not. Cumulative risk of knee arthroplasty was lower among GLP-1 users at follow-up points out to eight years, with the largest differences in patients on newer agents for longer durations.³

This is a meaningful signal at population scale. It is also observational, and the authors name the limits themselves: unmeasured differences in frailty, physical activity, functional capacity, and osteoarthritis severity were not accounted for, and prescription records do not confirm that patients actually took the medication.³ People prescribed a GLP-1 and kept on it for years differ from people who are not, in ways that also predict whether they end up in an operating room. That is not a reason to dismiss the finding. It is a reason not to read it as proof that the drug prevented surgeries.

Separately, a cost-effectiveness analysis published in Annals of Internal Medicine in 2025 modeled semaglutide and tirzepatide for patients with knee osteoarthritis and obesity using a validated simulation model. Tirzepatide produced greater modeled health benefit at lower cost than semaglutide, with an incremental cost-effectiveness ratio of about $57,400 per quality-adjusted life-year against diet and exercise. At a $100,000 per QALY threshold, tirzepatide had a 64 percent probability of being cost-effective and semaglutide 34 percent. For eligible patients, bariatric surgery came out ahead of both.⁶ A model is a set of assumptions, not an observation, and the authors note the result was most sensitive to medication price, treatment efficacy, and baseline BMI.⁶

What this does not mean

A few things get claimed that the evidence does not carry.

These are not arthritis drugs. Semaglutide and tirzepatide are approved for weight management and for type 2 diabetes. Neither carries an approved indication for osteoarthritis, and the 2025 systematic review frames GLP-1 use in knee osteoarthritis as an emerging therapeutic area rather than established treatment.⁵

No human study shows cartilage regrowth. The cartilage restoration finding is from a mouse model.⁴ STEP 9 measured pain and physical function, not joint structure.¹ Anyone telling you a GLP-1 rebuilds a worn knee is going beyond what has been published in people.

The trial does not generalize past its population. Adults with obesity and moderate radiographic knee osteoarthritis is a specific group. Extending the result to mild osteoarthritis, to severe end-stage disease, or to people without obesity is an assumption, not a finding.¹

The muscle problem sitting underneath all of this

Here is the tension that rarely makes it into coverage of the knee findings. The muscle that stabilizes and protects the knee is the quadriceps, and quadriceps weakness is not a neutral variable in osteoarthritis. A systematic review and meta-analysis of longitudinal studies found that low knee extensor strength increased the risk of worsening patellofemoral osteoarthritis, with evidence that the relationship to tibiofemoral disease may differ by sex.⁷ Exercise therapy, largely strengthening, is itself an evidence-based treatment for knee osteoarthritis: a Cochrane review of land-based exercise found reductions in knee pain and improvements in physical function.⁸

Meanwhile, lean mass loss is a well-documented feature of GLP-1-induced weight loss. A 2024 review of body composition changes across GLP-1-based therapies describes lean mass making up a substantial share of total weight lost and discusses mitigation strategies precisely because of that.⁹ A 2026 review raises the further point that strength changes in older adults may not track muscle mass changes one-for-one, so measuring mass alone can understate the functional cost.¹¹

Put those together. A GLP-1 reduces the load on your knee, which helps. It may simultaneously reduce the strength of the muscle that controls that knee, which does not help. The net effect on any individual depends heavily on whether that person trains. This is the same argument we make in our articles on the best exercises to preserve muscle and a beginner's resistance-training plan, but the knee is where the stakes are easiest to see.

The intervention that addresses both sides is not exotic. A meta-analysis of resistance training during caloric restriction in older adults with obesity found that resistance training prevented much of the lean mass loss that dieting otherwise causes.¹⁰ That is the same tool that shows up in the osteoarthritis literature as exercise therapy.⁸ One habit, two problems.

What to actually do

Nothing here is a substitute for a conversation with the clinician managing your knee. With that said, the practical implications are consistent across every source above.

Train the legs, and progress the load gradually. Quadriceps strength is the variable with the clearest link to knee osteoarthritis structure and symptoms,⁷ and resistance training is what preserves lean mass through weight loss.¹⁰ If a squat pattern hurts, that is worth raising with a clinician or physical therapist rather than abandoning lower-body work entirely.

Eat enough protein. Protein intake is the nutritional lever with the most direct bearing on whether lean mass survives an aggressive calorie deficit, which is what a GLP-1 creates. Our guide to protein targets on a GLP-1 covers the numbers and how to hit them with a suppressed appetite.

Watch the transition off. Weight regain and the body composition that comes back with it are their own problem, especially for older adults with joint disease. We cover that in coming off a GLP-1 after 65.

Track the composition, not just the scale. Our free muscle loss calculator gives a rough estimate of how much of your loss is likely lean mass, which is more informative for a knee than total pounds lost. If joint or tendon pain is your main concern, our review of GLP-1s and tendon or joint injury risk covers the connective-tissue side of the same question.

What the evidence does not say

To state the limits plainly. One randomized trial has tested a GLP-1 against placebo for knee osteoarthritis pain, in 407 people with obesity and moderate radiographic disease, over 68 weeks.¹ It reported a real between-group difference in a self-reported pain score, in a trial where the placebo group also improved substantially. It did not separate weight-mediated relief from any direct drug effect. The knee replacement finding is observational and its authors flag confounders they could not measure.³ The cost-effectiveness result comes from a simulation model whose output moves with drug pricing assumptions.⁶ The mechanism said to work independently of weight loss has been shown in mice and in cells, not in a human trial designed to isolate it.⁴⁵ And no published human evidence shows that these medications rebuild cartilage.

The bottom line

If you have obesity and moderate knee osteoarthritis, the weight loss a GLP-1 produces is very likely to reduce your knee pain, and there is now randomized evidence consistent with that.¹ The mechanism most likely doing the work is the one that has been measured most carefully: less body mass means dramatically less force through the joint with every step.² The more exciting claim, that the drug is treating the arthritis itself, is a live hypothesis with animal support and no human confirmation.⁴ Meanwhile, the muscle that protects the knee is exactly the tissue at risk during rapid weight loss.⁹ Resistance training and adequate protein are how you keep the benefit and avoid paying for it somewhere else. See every study behind these articles if you want to read the sources yourself.

Frequently asked