The earlier article on this site covered the retatrutide phase 2 programme and ended on a caveat: no phase 3 outcome data. That caveat has partly expired. TRIUMPH-1, the first of the pivotal obesity trials, was reported as topline results in May 2026 and published in full in the New England Journal of Medicine in September 202612.
This article works through what the trial was built to measure, what it found, and why the same trial produces two different headline numbers depending on who is quoting it.
The design
TRIUMPH-1 enrolled 2,339 adults with obesity, or overweight with at least one weight-related condition, and without type 2 diabetes. Participants were randomised to placebo or to once-weekly subcutaneous retatrutide at a maintenance dose of 4 mg, 9 mg or 12 mg, and followed for 80 weeks1.
Three things distinguish it from the phase 2 trial that preceded it4. It is roughly seven times larger. It runs 32 weeks longer. And it was designed to support a regulatory filing rather than to find a dose, which means the endpoints, the statistical plan and the handling of missing data were fixed in advance and are now part of the public record.
The primary result
Mean percentage change from baseline, n=2,339
Data from Jastreboff et al., NEJM 2026 [1]. Chart by TrueLab Care.
Under the treatment-regimen estimand, mean body-weight change from baseline at 80 weeks was −17.6% at 4 mg, −23.7% at 9 mg and −25.0% at 12 mg, compared with −3.9% on placebo1.
Those are the figures in the peer-reviewed paper. They are also lower than the figures that circulated after the May topline release, which reported −19.0%, −25.9% and −28.3% for the same three doses, with −2.2% on placebo2. Both sets are correct. The difference is the estimand.
Two estimands, two answers
Treatment-regimen vs efficacy estimand
NEJM paper [1] and Lilly topline release [2]. Chart by TrueLab Care.
An estimand is a precise statement of what a trial is estimating, including how participants who stop treatment or drop out are counted. Modern obesity trials pre-specify at least two.
The treatment-regimen estimand asks: what happened to everyone who was randomised, whether or not they stayed on the drug? A participant who stopped injecting at week 30 and regained weight still counts in the group they were assigned to. This is closest to real-world use and is the more conservative number.
The efficacy estimand asks: what did the drug do in people who actually took it as directed? Data after discontinuation are handled as if treatment had continued. This number is larger, and it is the one companies tend to lead with in press releases.
Neither is wrong, but they are not interchangeable, and secondary coverage that quotes −28.3% as "the phase 3 result" without stating the estimand is quoting the more flattering of two pre-specified answers. For a reader trying to understand what the compound does to a population over 80 weeks, the treatment-regimen figure is the more honest one.
Secondary endpoints that explain the trial's shape
TRIUMPH-1 reported reductions in knee-osteoarthritis pain and in the severity of obstructive sleep apnoea alongside the weight endpoint1. This looks odd for an obesity trial until one notices that the registration programme was designed to support several indications at once: obesity itself, sleep apnoea and osteoarthritis-related knee pain are the three named in Lilly's filings3. Collecting those endpoints inside the first pivotal trial means they do not require separate programmes.
What TRIUMPH-1 does not settle
Durability after stopping. Eighty weeks of exposure says nothing about what happens at week 120 without the compound. Weight regain after discontinuation is a documented feature of the incretin class, and this trial was not designed to measure it.
Rare adverse events. A trial of 2,339 people can characterise events that occur in roughly one in a few hundred participants. It cannot characterise events that occur in one in ten thousand. That is what post-marketing surveillance is for, and it has not started.
Generalisability. Participants were selected, monitored, and dose-escalated under protocol. The treatment-regimen estimand brings the result closer to real-world conditions than the efficacy estimand does, but it is still a trial population.
Status
Retatrutide is not approved in any jurisdiction. Lilly has stated that the clinical data package is complete and that a biologics licence application to the U.S. FDA is planned for the first quarter of 20273. Review timelines after submission are typically ten to twelve months, which places any approval decision in 2028 at the earliest.
Until then, material labelled retatrutide is a research compound. The trial literature describes what the molecule does under controlled conditions; it does not describe, and cannot describe, what an unverified vial of unknown provenance does.