The GLP-1 clinical trials, in one table

Most of what is said about GLP-1 drugs, in clinics, in the press and on social media, traces back to a small set of randomised trials. Each was large, ran for a year or more, and was published in a major journal. They are not interchangeable. Some enrolled people with type 2 diabetes and some excluded them. Some compared a drug with placebo, some with another drug. Some measured weight and some measured heart attacks, kidney failure or liver scarring. The table below lists the landmark ones, and the rest of this page explains how to read them.

The landmark trials

Weight results are mean percent change in body weight from the start of the trial. "Placebo" means the comparison arm received an inactive injection or tablet plus the same diet and activity advice. All trials were sponsored by the drug's manufacturer.

The SURMOUNT-OSA figure is the trial of people not using a CPAP machine. In the parallel trial of people already on CPAP the change was -29.3 vs -5.5 events per hour. Retatrutide is investigational in all of the trials above. The two TRIUMPH reports were published in late September 2026 and are the first peer-reviewed phase 3 results for the drug.

Where to go deeper

• The STEP trials: STEP 1 to 5, STEP 8 against liraglutide, STEP TEENS and the 7.2 mg STEP UP trial.
• The SURMOUNT trials: SURMOUNT-1 to 5 and the three-year prediabetes result.
• Semaglutide vs tirzepatide: what SURMOUNT-5 and SURPASS-2 showed head to head.
• Weight regain after stopping: STEP 1 extension, STEP 4, SURMOUNT-4 and the 2026 BMJ meta-analysis.
• SELECT and the heart, FLOW and the kidneys, sleep apnoea, heart failure and fatty liver.
• Retatrutide, orforglipron and oral semaglutide.
• Safety studies, lean mass studies and real-world studies.

How to read a GLP-1 trial

Phase

Phase 1 trials are small and test safety and how the body handles the drug. Phase 2 trials, usually a few hundred people, look for the right dose and a signal of effect. Phase 3 trials are the large, long ones regulators approve drugs on. The retatrutide result of -24.2% in 2023 was phase 2, from 338 people over 48 weeks. The phase 3 TRIUMPH-1 result of -25.0% on 12 mg came from 2,339 people over 80 weeks. Phase 2 numbers are often revised when phase 3 arrives, up or down.

The placebo arm

The placebo group in these trials did not do nothing. They received diet and activity counselling, regular visits and weighing, and an injection they believed might be the drug. That package alone moved weight by roughly 2 to 4 percent in most of the trials above, and by 5.7% in STEP 3, where it included 30 counselling sessions and an initial low-calorie diet. The difference between the arms is the drug's effect. The total in the drug arm is drug plus support.

Estimands: why the same trial gives two numbers

People stop drugs during long trials. An estimand is the rule for what to do with them. The two you will meet most often:

• Treatment policy, or treatment regimen. Everyone randomised is counted in the group they were assigned to, whether they kept taking the drug or not. It answers "what happens if you are offered this drug". STEP 1, SURMOUNT-1, ATTAIN-1 and the TRIUMPH trials report this as the primary result. It is the lower, more conservative number.
• Trial product, or efficacy. Results are estimated as if everyone had stayed on the drug as prescribed. It answers "what does the drug do if you take it". The number is higher. Press releases often lead with it.

When two sources quote different numbers for the same trial, this is usually why. This site uses the trial's primary estimand unless it says otherwise.

Mean change versus your change

A mean of -20.9% on tirzepatide 15 mg in SURMOUNT-1 hides a spread. On that dose, 90.9% lost at least 5% and 56.7% lost at least 20%, according to the Zepbound label. Some people lose very little. The responder rates, the share reaching 5, 10, 15 or 20 percent, tell you more about the range than the mean does.

Population

People with type 2 diabetes lose less weight on the same drug. Tirzepatide 15 mg produced -20.9% in SURMOUNT-1 (no diabetes) and -14.7% in SURMOUNT-2 (with diabetes). Comparing results across trials with different populations, different lengths and different estimands is how most misleading comparisons get made. That is why head-to-head trials like SURMOUNT-5 exist.

Outcomes trials

SELECT, FLOW and SUSTAIN-6 did not primarily measure weight. They counted events: heart attacks, strokes, deaths, kidney failure. They run for years and need thousands of people, and the result is a hazard ratio. A hazard ratio of 0.80 means the rate of events was 20% lower in the drug group over the trial, a relative reduction. The absolute numbers matter too: in SELECT, 6.5% vs 8.0% had a primary event over a mean of 39.8 months.

Who paid

Every trial in the table was funded by Novo Nordisk or Eli Lilly. That is standard for phase 3 drug trials and does not make the results wrong, since the protocols are registered in advance and the outcomes reviewed by regulators. It is still worth knowing, especially for head-to-head trials, where one company designed a comparison with its competitor's drug.

What these trials do not tell you

They do not test compounded versions, which were not used in any of them. They mostly stop at 68 to 80 weeks, with a few running two to four years. They enrolled more women than men in most weight trials and were not sized to answer questions about every subgroup. For anything beyond the trial population, the label and your prescriber are the reference points, and real-world studies fill some of the gaps, with the caveat that they show association rather than cause.

Trials report averages. Your own result is a line on a chart: the dose and the date you changed it, your weight, the side effects that came with each step and, if you have them, your labs. Peptide AI keeps those on one timeline, and the dose log and titration schedule are free. The free GLP-1 tracker page explains what it records, or download the app.