TRIUMPH-4: The First Retatrutide Phase 3 to Read Out

Why this trial read out first

Retatrutide's programme is named for obesity, but the first pivotal trial to report was the one in knee osteoarthritis. TRIUMPH-4 started in August 2023 and completed in November 2025; topline came on 11 December 2025, ahead of the pure obesity readouts that followed in 2026. That ordering matters for how the numbers were received. The first Phase 3 evidence anyone saw for this compound came from a population selected for joint disease, not for metabolic risk, and enrolled under criteria that excluded diabetes entirely.

Who was actually enrolled

The entry criteria are more specific than most coverage conveys, and they shape every number the trial produced. Participants needed a BMI of at least 27 kg/m2, index knee pain for more than 12 weeks before screening and on more than 15 days of the previous month, radiographic change of Kellgren-Lawrence grade 2 or 3 on a centrally read knee X-ray, and a diagnosis meeting American College of Rheumatology clinical and radiological criteria for osteoarthritis. Diabetes was an exclusion. So were steroid joint injections within 90 days, other joint injections or procedures within six months, joint disease other than osteoarthritis, weight-loss drugs within 90 days, a self-reported or documented weight change greater than 5 kg in the previous 90 days, and prior or planned bariatric surgery.

• BMI threshold 27, not 30 - a lower bar than most obesity trials
• Kellgren-Lawrence grade 2 or 3: moderate structural change, centrally read
• No pain-severity threshold at entry, only pain duration and frequency
• Diabetes excluded, which removes the population where incretins were first studied
• Recent weight change over 5 kg excluded, to keep the baseline stable

The weight numbers, and why there are two sets of them

Lilly reported body weight under two estimands, and they are not interchangeable. The efficacy estimand - what happens if participants stay on randomised treatment without rescue - gave -26.4% (-29.1 kg) on the arm escalated to 9 mg, -28.7% (-32.3 kg) on the arm escalated to 12 mg, and -2.1% (-2.1 kg) on placebo. The treatment-regimen estimand, which counts what happened to everyone regardless of what they did afterwards, gave -20.0% (-22.9 kg), -23.7% (-27.2 kg) and -4.6% (-5.3 kg). The 28.7% figure that travelled in headlines is the efficacy estimand at the higher dose. Both estimands are legitimate; they answer different questions, and a comparison that mixes them is meaningless. These are dose arms in a registered trial conducted under clinical supervision, described here as facts about the trial and nothing else.

The pain result, stated properly

The WOMAC pain subscale was scored on the 0-20 Likert version, and mean baseline was 6.0 points - toward the mild end of that range. Under the efficacy estimand, the 9 mg arm fell 4.5 points (-75.8%), the 12 mg arm fell 4.4 points (-74.3%), and placebo fell 2.4 points (-40.3%). Complete resolution of index knee pain was reported in 14.1% of the 9 mg arm, 12.0% of the 12 mg arm and 4.2% of placebo. Two things in that paragraph deserve more attention than they have had. The placebo arm improved by 40% of its baseline pain, which is enormous and entirely typical of osteoarthritis pain trials. And the higher dose, which produced more weight loss, did not produce more pain relief.

Tolerability, and what it costs to go from 9 mg to 12 mg

Discontinuation due to adverse events was 12.2% on the 9 mg arm, 18.2% on the 12 mg arm and 4.0% on placebo. The most common adverse events were gastrointestinal: nausea in 38.1-43.2% across the active arms, diarrhoea in 33.1-34.7%, constipation in 21.8-25.0% and vomiting in 20.4-20.9%. Put the efficacy and tolerability together and the higher dose bought 2.3 percentage points more weight loss and six percentage points more discontinuation, while measuring slightly worse on both pain endpoints. That is the kind of arithmetic that decides which dose is filed, and it is invisible if you only read the headline number.

What TRIUMPH-4 does not establish

It does not show that retatrutide treats osteoarthritis, and nothing here should be read that way. It shows that in a trial of 445 people meeting specific radiographic and symptomatic criteria, a randomised comparison moved two pre-specified endpoints over 68 weeks. The structural question - whether cartilage loss slows - was not a primary endpoint and the trial was not designed to answer it. WOMAC is a patient-reported symptom instrument; it records what people say they feel, which is the outcome that matters to them, but it is not a measure of joint structure.

Quick reference