Skip to content
GLP Loss
← Research
Evidence

GLP-1s and Osteoarthritis: What the Knee Trials Measured

STEP 9 cut WOMAC knee pain by 41.7 points. Its placebo arm, on diet and activity counseling alone, cut it by 27.5 — and the only other randomized trial in the class produced more weight loss and no pain benefit at all.

Owen Castellanos9 min read
Both arms of the knee trial got better68 weeks, 407 adults with obesity and knee osteoarthritisSTEP 9: change in WOMAC pain at week 68Semaglutide, 41.7 pointsPlacebo, 27.5 pointsAttributable to the drug: 14.2 points on a 0 to 100 scaleThe other randomized trial ran the diet first3.9 kg more weight lost. No pain benefit.KOOS pain difference 0.9 points, 95% CI -3.9 to 5.7Two randomized trials in the whole classBoth in the knee. None in the hip. None of tirzepatide.No GLP-1 carries an osteoarthritis indication.

Knee osteoarthritis is the one condition where the mechanical argument for weight loss writes itself. Less load across the joint, less pain. That is why a drug class built for weight keeps being pointed at an orthopedic clinic, and why the single dedicated randomized trial was read as settling the question.

It did not settle it, because the trial ran a control arm and the control arm improved a great deal. No GLP-1 receptor agonist carries an osteoarthritis indication from any regulator, and the headline obesity figures in the semaglutide weight-loss article are not what was measured here.

The dedicated trial, and the two thirds of it that was not the drug

STEP 9 was a 68-week, double-blind, placebo-controlled trial at 61 sites in 11 countries. It enrolled 407 participants with a body-mass index of 30 or above and a clinical and radiologic diagnosis of moderate knee osteoarthritis with at least moderate pain, randomized 2:1 to once-weekly subcutaneous semaglutide 2.4 mg or placebo. Both arms received counseling on physical activity and a reduced-calorie diet. The mean age was 56, the mean body-mass index 40.3, and 81.6% of participants were women.[1]

Body weight changed by −13.7% on semaglutide against −3.2% on placebo. The WOMAC pain score, on a 0 to 100 scale where higher is worse, fell by 41.7 points on semaglutide from a mean baseline of 70.9 — and by 27.5 points on placebo. Physical function on the SF-36 improved 12.0 points against 6.5.[1]

The number that belongs to the drug is the difference: 14.2 points of pain and 5.5 points of function, both significant at P < 0.001 and both a great deal smaller than the figure that travels. Adverse events leading to permanent discontinuation occurred in 6.7% of the semaglutide group against 3.0% on placebo, gastrointestinal disorders being the usual reason — the pattern described in what the trials recorded on side effects.

A clinically important improvement has a number attached

A pain score change means nothing without the threshold at which patients notice it. That threshold was estimated in a 1,362-patient multicenter cohort with knee or hip osteoarthritis, anchored on each patient’s own rating of their response. For knee osteoarthritis the minimal clinically important improvement in pain was −19.9 mm, or −40.8% in relative terms; for hip osteoarthritis, −15.3 mm and −32.0%. On the WOMAC function subscale the thresholds were −9.1 points (−26.0%) for the knee and −7.9 (−21.1%) for the hip.[2]

That threshold was anchored on a visual analog scale rather than the WOMAC pain subscale, so the percentages are comparable in kind and not in instrument. Read that way, the semaglutide arm improved 58.8% from baseline and the placebo arm 38.8%. The important word is both: a within-patient threshold of roughly 40% is one the control arm essentially reached on counseling alone. Whatever the drug added, it was added on top of something already working.

The second randomized trial removed the weight loss first

The other trial in this class inverted the design, and the result inverted with it. Patients aged 18 to 74 with knee osteoarthritis and a body-mass index of 27 or above entered an eight-week dietary intervention before randomization. Only those who had already lost more than 5% of body weight were randomized, 156 of them, to liraglutide 3 mg daily or placebo for 52 weeks.[3]

Over those 52 weeks liraglutide separated from placebo on weight: −2.8 kg against +1.2 kg, a group difference of 3.9 kg (95% CI −6.9 to −1.0, P = 0.008). On pain it did nothing. The KOOS pain subscale changed by 0.4 points on liraglutide and −0.6 on placebo, a group difference of 0.9 points (95% CI −3.9 to 5.7, P = 0.71). Gastrointestinal adverse events ran 50.2% against 39.2%.[3]

More weight off, no pain benefit, worse tolerability. The two trials differ in molecule, dose and the amount of weight involved, but the cleanest reading of the pair is that pain relief tracks the weight a patient has not yet lost. Take the first several percent away with a diet before randomizing, and the drug has little left to deliver.

Diet and exercise reached almost the same weight loss

The comparator that matters is not placebo, it is an intensive lifestyle program. The IDEA trial randomized 454 overweight and obese adults aged 55 and over with painful radiographic knee osteoarthritis to intensive diet plus exercise, intensive diet alone, or exercise alone, for 18 months. Mean weight loss was 10.6 kg (11.4%) with diet plus exercise, 8.9 kg (9.5%) with diet, and 1.8 kg (2.0%) with exercise.[4]

At 18 months the diet-plus-exercise group reported less pain on a 0 to 20 scale — 3.6 (95% CI 3.2 to 4.1) against 4.8 for diet alone and 4.7 for exercise alone — with the pairwise difference against exercise alone at 1.02 points (95% CI 0.33 to 1.71, P = .004). Knee compressive force was lower in the diet arms by 200 N.[4]

So a supervised program produced 11.4% weight loss over 18 months where semaglutide produced 13.7% over 68 weeks. The gap between a modern GLP-1 and a well-run lifestyle program in this population is narrower than it is in the general obesity literature, and the lifestyle arm carries a loading and strength component the drug does not. The randomized exercise evidence on this class is in the exercise article.

The structural question gets three different answers

Pain is one endpoint. Whether the joint itself does better is another, and the observational literature contradicts itself along a line worth seeing. A five-year analysis excluded anyone with pre-existing hip or knee osteoarthritis and matched 1:1 on age, sex, race, body-mass index and hemoglobin A1c. Among patients with diabetes, incidence of new osteoarthritis was higher on a GLP-1 receptor agonist: hip hazard ratio 1.63 (95% CI 1.46 to 1.82) and knee 1.52 (1.41 to 1.64) in the obese diabetic cohort, and 1.78 and 1.58 in the non-obese diabetic cohort. In the obese, non-diabetic cohort there was no difference on any outcome, and no cohort differed on hip or knee replacement rates.[5] That last stratification is the one a cash-pay reader sits in: the unfavorable direction appeared only where diabetes was present.

A 2026 analysis asked the opposite question — not who develops osteoarthritis, but who ends up in an operating room with it. Among adults already diagnosed with knee osteoarthritis, matched cohorts of 13,351 to 42,062 patients showed lower cumulative knee replacement incidence at every interval. One year of exposure to any agent gave an absolute risk difference at eight years of −2.80 percentage points (HR 0.90, 95% CI 0.83 to 0.98); three years of semaglutide or tirzepatide gave −4.71 points (HR 0.72, 95% CI 0.67 to 0.78).[6]

A third study used a competitor as the comparator rather than nothing. In 452,445 propensity-matched pairs with type 2 diabetes, SGLT2 inhibitors carried a lower five-year osteoarthritis risk than GLP-1 receptor agonists (HR 0.921, 95% CI 0.895 to 0.946), widest for knee osteoarthritis (HR 0.831, 95% CI 0.783 to 0.882), with no difference in hip or knee replacement.[7]

The randomized record adjudicates by declining to find anything. A network meta-analysis pooled 74 randomized trials and 105,415 participants with type 2 diabetes and found no significant difference in osteoarthritis incidence between GLP-1 receptor agonists and placebo, insulin or oral agents — in pairwise analysis, in the network, and across every subgroup by formulation, duration and comparator.[8] Incident diagnosis and progression to surgery are different endpoints, measured in different populations, and nothing in this corpus licenses a structural claim in either direction.

The hip has no trial, and tirzepatide has none either

Where a topic is mostly an absence, the census is the finding. Every symptomatic figure above is a knee figure. A scoping review of five databases from January 2010 to August 2025 included 199 studies, of which 36 (18.1%) assessed GLP-1 receptor agonists directly and only 14 of those 36 (38.9%) reported original data. Of the 36, 29 (80.6%) appeared in 2024 or 2025 and 52.8% were of low level of evidence. Its summary of the distribution: heavily skewed toward knee osteoarthritis, sparse in the hip and weaker where it exists.[9]

Indexed randomized trials of this class in osteoarthritis, at title level and restricted to the randomized-controlled-trial publication type, number two: the semaglutide and liraglutide trials above, both in the knee. There is no randomized trial of any GLP-1 receptor agonist in hip osteoarthritis, and none of tirzepatide in any joint — the only tirzepatide-and-osteoarthritis record indexed at title level is an economic model rather than a study of patients. A dedicated knee trial sits unreported inside the retatrutide program, covered in the retatrutide article.

That economic model prices the gap anyway. Running a validated knee osteoarthritis microsimulation from a base-case cohort at WOMAC pain 71, it put tirzepatide ahead of semaglutide on both benefit and cost, at $57,400 per quality-adjusted life-year against diet and exercise, with a 64% probability of being cost-effective at a $100,000 threshold against semaglutide’s 34%. Roux-en-Y gastric bypass provided greater health benefits at lower cost than either drug.[10] Its tirzepatide inputs are extrapolated from obesity trials, because no knee osteoarthritis trial of it exists to draw on. The surgical comparison itself is in the bariatric surgery article.

The caveat that cuts the other way

The tissue that protects an arthritic knee is muscle, and rapid weight loss is not neutral toward it. In a 12,043-participant cohort with 2,008 incident cases of symptomatic knee osteoarthritis, the highest band of normalized grip strength carried a relative risk of 0.67 (95% CI 0.60 to 0.75) against the reference group, and slower chair-rising time a relative risk of 1.65 (95% CI 1.17 to 2.33). Sarcopenic obesity roughly doubled the risk, and muscle strength mediated the obesity association.[11]

That mechanism runs opposite to the one these drugs are sold on. If strength is part of the pathway by which excess weight damages a knee, losing weight in a way that also costs strength is not straightforwardly protective. What these drugs do to lean tissue is in the body-composition article, and what they do to the skeleton is in the bone density article. Neither has been measured inside an osteoarthritis trial.

One literature does favor the class, and it is the one where hip data is the stronger half. Ten matched cohort studies covering 96,356 patients — 30,350 hip and 66,606 knee replacements — pooled to a 90-day surgical complication risk ratio of 0.73 among GLP-1 users, with medical complications at 0.78 and readmissions at 0.79. For hip replacement the associations held across surgical complications (RR 0.63), medical complications (0.55) and readmissions (0.82); for knee replacement only readmission reached significance (RR 0.77). GRADE certainty was very low for all but one outcome.[12] Timing a dose around an operation is covered in the surgery and anesthesia article.

What a buyer is actually holding

STEP 9 studied branded semaglutide 2.4 mg alongside structured diet and activity counseling, in people screened for radiographic knee osteoarthritis. A vial from a cash-pay service inherits the molecule and none of the rest. Compounded semaglutide and tirzepatide have no FDA approval, and the agency does not review them for safety, effectiveness or quality before a pharmacy ships them — the detail is in what a compounded vial contains, and the sellers are collected on the compounded semaglutide board.

There is also a duration problem specific to joints. Osteoarthritis is permanent and the weight loss is not, as what happens when you stop documents, and a knee that felt better at 68 weeks has no reason to stay that way once the weight returns. Neither trial followed anyone after treatment ended.

The defensible summary is narrow. Semaglutide 2.4 mg beat placebo on knee pain by 14.2 points over 68 weeks, on top of a control arm that improved substantially on counseling alone. Adding a GLP-1 after the weight was already lost produced no pain benefit. Nothing randomized has measured the hip, tirzepatide, cartilage, or the muscle around the joint. None of that is an approval, and none of it replaces an orthopedic assessment.

Frequently asked

Does semaglutide help knee osteoarthritis pain?
In the one dedicated randomized trial, 407 adults with obesity and moderate knee osteoarthritis took semaglutide 2.4 mg or placebo for 68 weeks alongside diet and activity counseling. WOMAC pain fell 41.7 points on semaglutide and 27.5 points on placebo, from a baseline of 70.9 on a 0 to 100 scale. The difference attributable to the drug is 14.2 points, which is statistically significant and considerably smaller than the headline figure.
Why did the placebo group improve so much?
Both arms of the trial received counseling on physical activity and a reduced-calorie diet, and the placebo arm lost 3.2% of body weight. Osteoarthritis pain also fluctuates and responds strongly to attention and expectation, which is why an uncontrolled before-and-after figure in this condition is close to uninterpretable. A within-patient threshold for a clinically important improvement in knee pain is roughly 40% from baseline, and the placebo arm improved 38.8%.
Is there any evidence for tirzepatide in osteoarthritis?
No randomized trial of tirzepatide in osteoarthritis of any joint has been published. The only title-level record linking tirzepatide and osteoarthritis in the indexed literature is a cost-effectiveness microsimulation, which extrapolated its efficacy inputs from obesity trials rather than from joint outcomes. That model estimated an incremental cost-effectiveness ratio of $57,400 per quality-adjusted life-year against diet and exercise.
Do these drugs slow the arthritis itself, or just reduce the pain?
The observational evidence points in more than one direction. A five-year cohort found higher incidence of new hip and knee osteoarthritis among users with diabetes, at hazard ratios of 1.63 and 1.52, with no difference at all in the obese non-diabetic group. A separate analysis in people who already had knee osteoarthritis found lower knee replacement incidence, at a hazard ratio of 0.72 after three years of exposure. A network meta-analysis of 74 randomized trials in 105,415 participants found no difference in osteoarthritis risk at all.
Would losing the same weight through diet and exercise work as well?
The IDEA trial randomized 454 older adults with painful knee osteoarthritis and found mean weight loss of 10.6 kg, or 11.4%, in the diet-plus-exercise arm over 18 months, against 13.7% on semaglutide over 68 weeks in the drug trial. Those are different trials with different populations and cannot be compared head to head, but the weight-loss gap is narrower here than in the general obesity literature. The lifestyle arm also delivered a strength and loading component that no injection provides.
Is there a downside specific to arthritic joints?
Muscle strength is part of the pathway linking excess weight to knee osteoarthritis, not merely a bystander. In a cohort of 12,043 participants, the highest band of grip strength carried a relative risk of 0.67 for incident symptomatic knee osteoarthritis, sarcopenic obesity carried roughly double the risk, and strength statistically mediated the obesity association. No osteoarthritis trial of a GLP-1 has measured lean mass or bone density, so the size of that trade-off in this population is unknown.

Sources

  1. [1] Bliddal H, Bays H, Czernichow S, et al. (2024). Once-Weekly Semaglutide in Persons with Obesity and Knee Osteoarthritis. N Engl J Med. PMID 39476339
  2. [2] Tubach F, Ravaud P, Baron G, et al. (2005). Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: the minimal clinically important improvement. Ann Rheum Dis. PMID 15208174
  3. [3] Gudbergsen H, Overgaard A, Henriksen M, et al. (2021). Liraglutide after diet-induced weight loss for pain and weight control in knee osteoarthritis: a randomized controlled trial. Am J Clin Nutr. PMID 33471039
  4. [4] Messier SP, Mihalko SL, Legault C, et al. (2013). Effects of intensive diet and exercise on knee joint loads, inflammation, and clinical outcomes among overweight and obese adults with knee osteoarthritis: the IDEA randomized clinical trial. JAMA. PMID 24065013
  5. [5] Lavu MS, Porto JR, Hecht CJ 2nd, et al. (2024). The Five-Year Incidence of Progression to Osteoarthritis and Total Joint Arthroplasty in Patients Prescribed Glucagon-Like Peptide 1 Receptor Agonists. J Arthroplasty. PMID 38857711
  6. [6] Carter V, Desverreaux E, Amin I, et al. (2026). Glucagon-like peptide 1 receptor agonist use and risk of arthroplasty for knee osteoarthritis: retrospective database analysis. Reg Anesth Pain Med. PMID 42229941
  7. [7] Jao WY, Lai CC, Chien CS, Lin WT (2026). Comparative effects of SGLT2 inhibitors and GLP-1 receptor agonists on osteoarthritis risk in patients with type 2 diabetes mellitus: A multi-institutional cohort study. PLoS One. PMID 42475344
  8. [8] Zhou L, Wu P, Xu Y, Xie B (2026). The effect of GLP-1 receptor agonists on osteoarthritis risk in individuals with type 2 diabetes mellitus: a systematic review and network meta-analysis of randomized controlled trials. Acta Diabetol. PMID 42101662
  9. [9] McCann CJ, Clement ND, Murray A, et al. (2026). GLP-1 receptor agonists and weight-loss strategies for individuals with obesity and hip or knee osteoarthritis: a scoping review. Br J Sports Med. PMID 42321007
  10. [10] Betensky DJ, Smith KC, Katz JN, et al. (2025). The Cost-Effectiveness of Semaglutide and Tirzepatide for Patients With Knee Osteoarthritis and Obesity. Ann Intern Med. PMID 40953447
  11. [11] Yan L, Ge H, Wang Z, et al. (2024). Roles of low muscle strength and sarcopenic obesity on incident symptomatic knee osteoarthritis: A longitudinal cohort study. PLoS One. PMID 39361683
  12. [12] Rudran B, Little CDJ, Palmer AJR, et al. (2026). Glucagon-like peptide-1 receptor agonists are associated with improved 90-day outcomes after total hip and knee arthroplasty : a systematic review and meta-analysis. Bone Joint J. PMID 42538001

Where to get it

Best GLP-1 injections

Every injectable seller we can verify, with the price each one publishes and an honest read of what the trials measured.

Compare providers →

More in Evidence