A pen sold as a research compound and a molecule in a phase 2 trial share a name and almost nothing else, and the gap between them is where a reader can get hurt. The retatrutide record is unusually good: two large randomised trials with hard numbers, a clear relationship between the amount given and the effect, a body-composition substudy, and adverse events reported by group. Reading it properly takes about ten minutes, and the label on a pen cannot be made to survive the comparison.
What the phase 2 trial actually reported
The first paper is a phase 2, double-blind, randomised, placebo-controlled trial: 338 adults with obesity, 51.8% of them men, randomised in a 2:1:1:1:1:2:2 ratio to six once-weekly subcutaneous amount groups or placebo, for 48 weeks. Eligibility was a BMI of 30 or more, or 27 to under 30 with at least one weight-related condition. The primary endpoint was percentage change in body weight at 24 weeks.
At 24 weeks the least-squares mean change was −7.2% in the lowest amount group, −12.9% in the combined 4-milligram group, −17.3% in the combined 8-milligram group and −17.5% in the 12-milligram group, against −1.6% on placebo. At 48 weeks: −8.7%, −17.1%, −22.8% and −24.2%, against −2.1% on placebo.
The responder analysis is the part that ends up on marketing material, so it should be quoted in full. At 48 weeks, a weight reduction of at least 5%, at least 10% and at least 15% had occurred in 92%, 75% and 60% of the combined 4-milligram group; in 100%, 91% and 75% of the 8-milligram group; in 100%, 93% and 83% of the 12-milligram group; and in 27%, 9% and 2% of the placebo group.
Safety reads the way a reader of a label will not find it printed: the most common adverse events were gastrointestinal, related to the amount given, mostly mild to moderate, and partially mitigated by a lower starting amount. Heart-rate increases tracked with the amount given, peaked at 24 weeks and declined thereafter. The trial was funded by Eli Lilly, and that is stated in the paper.
The substudy that scanned the fat
The second paper, from 2025, answers a question the weight number cannot: of the fat that went, how much? It is a body-composition substudy of a phase 2 trial in 42 US medical centres, in adults with type 2 diabetes, HbA1c between 7.0 and 10.5%, and a BMI from 25 to 50.
Of 281 participants in the main study, 189 entered the substudy, and 103 completed treatment with both a baseline and a week 36 DXA scan. Percent reduction in total fat mass was 26.1% in the 8-milligram group and 23.2% in the 12-milligram group, against 2.6% on dulaglutide and 4.5% on placebo. Against placebo, the least-squares mean change in total fat mass was −0.4 with 95% CI −4.0 to 3.2 for the lowest amount, and −10.7 (−17.2 to −4.2), −21.6 (−27.1 to −16.1) and −18.7 (−25.1 to −12.3) for the higher ones. Adverse events were similar between groups.
Read that first confidence interval carefully: for the lowest amount, the interval crosses zero. That is the difference between “no measurable effect on fat mass” and “a large one”, in the same trial, in neighbouring groups.
What a label on a pen cannot tell you
A label cannot tell you which of these six groups anything corresponds to, because the amount on a research pen and the amount in the trial are not the same schedule. A label cannot tell you the randomisation ratio, the responder analysis, the placebo arm, the 48-week horizon, or that the heart-rate increase peaked and then declined. It cannot tell you that 103 people were scanned twice out of 338 enrolled, which is what “a quarter of your fat” actually rests on.
It cannot tell you that this is a molecule in phase 2 for obesity with a company-sponsored programme behind it, tested in people who met a BMI threshold and whose weight was, by design, not the thing being studied long term.
And it cannot tell you what happens after the last injection, because neither paper measured that.
