Prepared by: Research Desk. Sources are linked for verification.

Status: Investigational. Retatrutide is not approved by the FDA, which addresses it by name on its page about unapproved GLP-1 drugs [6]. See where retatrutide stands with the FDA.

Most headlines about retatrutide trial results reduce a long paper to a single figure. That figure is usually real, but it describes one group, at one time point, compared with one control, in one population. This page explains how the papers are built so readers can tell what a number does and does not show. It covers method, not findings. For the findings, see the published retatrutide results.

Why headline numbers mislead

A single percentage cannot carry the context needed to interpret it. Before relying on any retatrutide figure, five questions are worth asking:

1. Which group? Trials randomize participants to several retatrutide groups, and headlines usually quote the one with the largest effect.

2. Compared with what? Placebo groups also change weight. The meaningful number is the difference between groups.

3. Over how long? Phase 2 obesity results were measured at 48 weeks [3]. TRIUMPH-1 and TRIUMPH-2 report 80-week results [1][2].

4. In whom? TRIUMPH-1 enrolled adults with obesity or overweight, while TRIUMPH-2 enrolled adults with obesity and type 2 diabetes [1][2].

5. Counting whom? Results can be calculated for everyone randomized or only for those who stayed on the study drug, and the two can differ.

The rest of this guide takes each question in turn.

Who was enrolled in each trial

Every trial starts with eligibility criteria: who could join and who was excluded. These rules shape the results as much as the drug does.

Because TRIUMPH-1 and TRIUMPH-2 enrolled different populations, their weight-change figures should not be read as if they came from one experiment [1][2]. A gap between the two trials may reflect who was enrolled as much as anything else, and a cross-trial comparison cannot separate those explanations.

The full inclusion and exclusion lists are on each trial's ClinicalTrials.gov record [4][5]. They matter for generalisability. A result applies most directly to people who resemble those enrolled. It says much less about groups that were excluded or barely represented.

Primary and secondary endpoints

An endpoint is a prespecified measurement that a trial is designed to assess. Endpoints come in a hierarchy.

• Primary endpoint. This is the main question the trial was sized and powered to answer. The registry records for TRIUMPH-1 and TRIUMPH-2 list percent change in body weight from baseline among their primary outcome measures [4][5].

• Secondary endpoints. These are additional prespecified questions. In phase 2, one was the share of participants reaching at least 15% weight reduction: up to 83% in a retatrutide group versus 2% with placebo [3]. In TRIUMPH-1, outcomes in the prespecified subgroups included knee-osteoarthritis pain and the apnea-hypopnea index for sleep apnea [1].

• Exploratory endpoints. These are hypothesis-generating. They can point to questions for future trials but are not designed to confirm anything.

Large trials often test secondary endpoints in a fixed order. If an earlier test fails, later ones are not formally claimed. When a paper calls a secondary endpoint "significant," it is worth checking whether it sat inside that controlled testing hierarchy or was reported outside it.

The time point is part of the endpoint. "Percent change at 80 weeks" and "percent change at 48 weeks" are different endpoints, even when they come from the same drug.

Placebo comparisons and percent weight change

Results are reported as percent change from baseline weight, not kilograms. A 20% change means very different absolute amounts for people of different starting weights. Percent change makes participants comparable within a trial.

Placebo groups change too. In the phase 2 obesity trial, mean weight change with placebo was −2.1% at 48 weeks, against −8.7% to −24.2% across the retatrutide groups [3]. Everyone in a trial receives study visits, measurements and usually the same background guidance on diet and activity. The placebo figure captures those effects. The question a randomized trial answers is how much additional change occurred in the retatrutide groups.

Simple subtraction gives a rough gap of about 6.6 to 22.1 percentage points between the retatrutide groups and placebo in that trial [3]. That is our arithmetic, not the paper's statistics. Papers report an estimated between-group difference with a 95% confidence interval, which shows the range of values consistent with the data. A narrow interval means a precise estimate. A wide one means more uncertainty, which is common in small subgroups.

A second subtlety is the estimand: the precise question an analysis answers. Recent obesity trials often report two versions of the same endpoint. One counts everyone randomized, regardless of whether they kept taking the study drug (often called the treatment-regimen or treatment-policy estimand). The other estimates what would have happened had everyone continued as assigned (often called the efficacy estimand). The second is usually the larger number. The methods section states which analysis is primary, and a headline may quote either one without saying so.

Why results are reported as ranges across study groups

The retatrutide trials did not compare one drug group with placebo. They compared several retatrutide groups, each assigned a different fixed dose level, with placebo. Each group gets its own result, which is why figures appear as ranges, such as −8.7% to −24.2% in phase 2 [3].

Two points follow:

• The top of a range describes one group only. An "up to" figure is a statement about the most responsive group, not the average across everyone who received retatrutide.

• The bottom of a range is equally real. A group with a smaller effect may also have had fewer side effects. Effect and tolerability need to be read side by side, group by group.

Averages also hide spread within each group. A mean of −20% includes participants who changed far more and others who changed far less. Threshold endpoints (the share reaching at least 5%, 10% or 15% reduction) and, where published, individual-level plots give a better sense of that variation.

Discontinuations and adverse-event tables

Adverse-event tables usually report the proportion of participants with at least one occurrence of an event. In the phase 2 obesity trial, the most common adverse events were gastrointestinal: nausea, diarrhoea, vomiting and constipation [3]. Reading tables like these well means:

1. Comparing with the placebo column. Some events occur at background rates in any population. The excess over placebo is the signal.

2. Remembering that frequency is not severity. A participant with one mild episode counts the same as one with weeks of symptoms. Severity, timing and duration are usually in the full text or supplementary appendix.

3. Separating discontinuation types. Discontinuation because of adverse events and overall discontinuation, for any reason, are different figures. The second is usually larger.

4. Looking at time course. In phase 2, heart rate rose in a dose-dependent way, peaked at 24 weeks and declined afterwards [3]. A single end-of-trial number would have missed that pattern.

Discontinuations also feed back into efficacy. Participants who stop early leave gaps in the data, and how those gaps are handled is one reason the two estimands above can differ. Rare adverse events are a further limit. A trial of a few thousand people can miss events that would appear only in much larger or longer use.

Phase 2 versus phase 3: what changes

Phase 2 and phase 3 trials answer different questions:

• Purpose. The phase 2 obesity trial (338 participants, 48 weeks) was designed mainly to choose dose levels and gauge the size of the effect [3]. Phase 3 trials such as TRIUMPH-1 (80 weeks) are designed to confirm prespecified endpoints [1].

• Size and duration. Larger, longer trials produce narrower confidence intervals and a bigger safety database.

• Population breadth. Phase 3 programmes run separate trials in distinct populations, such as type 2 diabetes in TRIUMPH-2 [2][5].

Comparing a phase 2 figure with a phase 3 figure is an indirect comparison: different trials, participants and time points. It can suggest a pattern, such as weight change levelling off over the second year, but it cannot prove one. Only data from within a single trial, measured over time, show the shape of the curve directly.

Registry records versus published papers

Each TRIUMPH trial is registered on ClinicalTrials.gov with its planned outcomes, target enrolment, eligibility criteria and status: TRIUMPH-1 as NCT05929066 [4] and TRIUMPH-2 as NCT05929079 [5]. The published paper is the peer-reviewed analysis.

Reading the two together is a basic integrity check:

• Do the published primary endpoints match the registered ones? Changes are not always improper, but they should be explained.

• Does the reported enrolment match the registry?

• Is everything prespecified actually reported? Some outcomes may appear in a later paper or a supplement rather than the main report.

The abstract is the shortest and most quoted layer. The full text adds methods and limitations. The supplementary appendix often holds the detailed adverse-event and discontinuation data. Headline coverage usually draws only on the abstract.

Further reading on evaluating research

The skills on this page apply to any randomized trial: identifying the population, finding the primary endpoint, reading a confidence interval and checking a side-effect table against its control column. A good general book on statistics and research evidence makes those habits faster and more reliable.

When choosing one, useful criteria are: whether it is written for general readers or assumes prior statistics; whether it works through real published studies rather than invented examples; whether it covers confidence intervals, absolute versus relative differences, bias and multiple testing; the author's background in statistics or epidemiology; how recent the edition is; and format and price. Print and ebook editions often differ noticeably in cost.

Any titles we add here will be linked through Bookshop.org, which sells books on behalf of independent bookshops. They will be chosen on the criteria above, and paid links will be labelled. Our sourcing and disclosure rules are explained on how this site works.

Last updated: 18 October 2026.

Not medical advice. This page reports published research and regulatory records for information only.

Research-only notice: Content on this site is provided for research and educational purposes only and is not intended for human use. It is not medical advice.

Translation notice: Some content on this site may have been machine-translated using Google Translate and may contain errors.