Retatrutide: why 28.3% and 25.0% can both be correct

A retatrutide headline quoting 28.3% weight loss and a journal abstract quoting 25.0% need not contradict each other. They answer different statistical questions. For a practice discussing an investigational medicine, losing that distinction turns an evidence summary into a promise the trial did not make.

The September 29 NEJM abstract reports TRIUMPH-1, an 80-week, randomized, double-blind, placebo-controlled phase 3 trial in adults with obesity without diabetes. A total of 2,339 participants were randomized. Retatrutide activates GIP, GLP-1 and glucagon receptors. Trial doses below identify randomized arms, not prescribing instructions.

Start with the treatment-regimen result

The paper's abstract reports mean weight changes of −17.6%, −23.7% and −25.0% at 4, 9 and 12 mg, respectively, versus −3.9% with placebo. Its weight endpoints use a treatment-regimen estimand, described as intention-to-treat. That framework addresses intercurrent events rather than pretending that everyone remained on assigned therapy.

The May 21 sponsor topline highlighted a 28.3% reduction at 12 mg using the efficacy estimand. Its footnote defines that estimand as the effect had all randomized participants remained on study intervention, with possible interruptions and modifications, without starting prohibited weight-management treatments. Its treatment-regimen definition instead includes effects regardless of adherence or starting such treatments. The 28.3% figure is not the unqualified average outcome every future patient should expect.

Keep the safety denominator beside the result

In Lilly's original distributed release, 12-mg nausea was reported in 42.4% versus 14.8% with placebo; diarrhea in 32.0% versus 13.5%; and vomiting in 25.3% versus 4.8%. Discontinuations because of adverse events were 4.1%, 6.9% and 11.3% at 4, 9 and 12 mg, respectively, versus 4.9% with placebo. These are sponsor-reported trial-arm rates, not independently estimated real-world risks. The journal abstract identifies gastrointestinal events as the most common adverse events; this review has not checked its complete safety tables or supplements against the earlier release.

The higher-dose efficacy finding should therefore not be separated from tolerability, follow-up or the population enrolled. Nor does a shorter adverse-event list establish long-term safety or safety in groups absent from the trial.

The 104-week extension is a selected population

The May release also describes a 104-week extension. It enrolled 532 participants with baseline BMI at least 35 who completed the main trial on study drug and met further continuation criteria, including tolerating treatment. It is not the same denominator as all 2,339 randomized participants. An extension headline cannot simply replace the main randomized result in a patient-facing comparison.

Separate the endpoints and the commercial status

The paper includes knee osteoarthritis and sleep-apnea subgroups. Its abstract describes a hybrid-treatment estimand for those endpoints, with a hypothetical strategy for events suggesting treatment failure; intention-to-treat results are also reported. A single weight percentage does not summarize those separate outcomes. No cross-trial ranking against tirzepatide or semaglutide follows from this placebo-controlled study.

Retatrutide is described as investigational in the reviewed sponsor record. Publication of a positive phase 3 trial is not approval of a marketed product, and these trial data do not validate products sold outside the trial supply chain. This article reviews the complete abstract and targeted sponsor passages, not the full journal paper, protocol or supplements. Eli Lilly funded the trial. The next useful evidence step is a full-paper benefit–risk extraction, not another larger-number headline.

Evidence record

Event date: 2026-09-29. Source review: 2026-10-29. Editorial publication: 2026-10-29. Original source links below retain their publication dates. These dates are not interchangeable. This is targeted source synthesis, not an exhaustive archive or a qualified clinical/legal review.

Revision history

29 October 2026: original analysis added from the specific source passages identified in the evidence audit. Journal abstracts and sponsor reporting are labeled separately; full-paper and supplementary safety assessment remain outstanding. Report a proposed correction through Contact us; do not include patient data.