The short answer
Switching from the efficacy estimand to the treatment-regimen estimand usually shrinks a drug's measured effect. In TRIUMPH-4 it shrank both active arms as expected but more than doubled the placebo arm's weight loss, narrowing the gap from both ends at once.
Key facts
- Trial
- TRIUMPH-4 (NCT05931367)
- Placebo, efficacy estimand
- -2.1% (-2.1 kg)
- Placebo, treatment-regimen estimand
- -4.6% (-5.3 kg)
- Higher dose, efficacy estimand
- -28.7%
- Higher dose, treatment-regimen estimand
- -23.7%
- Gap, efficacy estimand
- 26.6 percentage points
- Gap, treatment-regimen estimand
- 19.1 percentage points
- Trial duration
- 68 weeks
What an estimand is, briefly
An estimand is a precise statement of what a trial is estimating, including how it handles events that occur after randomisation (stopping treatment, starting something else, having surgery). Two analyses of the same data can produce different numbers not because one is wrong but because they answer different questions. The efficacy estimand asks what would happen if participants remained on randomised treatment without rescue. The treatment-regimen or treatment-policy estimand asks what happened, counting all data regardless of what participants did afterwards.
The usual direction, and the exception
Normally the treatment-regimen estimand gives a smaller effect, because it includes data from people who stopped the drug and regained weight. TRIUMPH-4's active arms behave exactly that way: -26.4% becomes -20.0%, and -28.7% becomes -23.7%. The placebo arm goes the other way. Its weight change moves from -2.1% under the efficacy estimand to -4.6% under the treatment-regimen estimand, more than doubling. A placebo arm that loses more weight when everything that happened is counted needs explaining.
Research material referenced
Retatrutide 10mg, third-party HPLC tested
The most likely explanation
The arithmetic is unambiguous; the explanation is inferential, and the sponsor's topline does not spell it out. The reading that fits is that participants who left the placebo arm did not simply resume their previous trajectory. Some of them obtained effective weight-loss treatment outside the trial. Under the efficacy estimand those post-discontinuation observations are excluded or handled as if treatment had continued; under the treatment-regimen estimand they are counted as they occurred. The placebo arm therefore becomes, in part, a treated arm, and its measured weight loss rises. Other contributors are possible: differential dropout, intercurrent illness, or lifestyle intervention effects concentrated among those who left. None of these can be separated from the published topline.
Why it narrows the gap from both ends
This is the part that matters for interpretation. The drug arms shrink because discontinuers regain; the placebo arm grows because discontinuers get treated elsewhere. The measured difference between them falls twice as fast as either movement alone. At the higher dose the gap goes from 26.6 percentage points on the efficacy estimand to 19.1 on the treatment-regimen estimand. That is a 28% reduction in the estimated treatment difference driven entirely by the analytical question asked, not by anything that happened in the joints or the gut of any participant.
Why this gets worse every year
A placebo arm in an obesity trial is only a placebo arm if it stays untreated. That was easier to arrange when there was little effective treatment to leave for. It is now much harder: participants randomised to placebo in a 68-week trial have licensed, effective options available, and there is no ethical mechanism to prevent them from seeking those options. As effective agents spread, placebo arms in this field will drift further toward being partially treated arms, and treatment-policy estimands will drift further from efficacy estimands. This is a structural problem for the field's evidence base, not a flaw in any one trial.
What to do with this when reading a result
Three things. Check which estimand a quoted number comes from, because the same trial routinely publishes both. Never compare a number from one estimand against a number from another, even within the same trial. And when a placebo arm shows unexpected movement, treat it as information about the trial's conduct rather than as background noise. It is often the most revealing number in the table.
Quick reference
| Arm | Efficacy estimand | Treatment-regimen estimand | Direction |
|---|---|---|---|
| Retatrutide 9 mg | -26.4% | -20.0% | Effect shrinks |
| Retatrutide 12 mg | -28.7% | -23.7% | Effect shrinks |
| Placebo | -2.1% | -4.6% | Loss more than doubles |
| Gap at higher dose | 26.6 points | 19.1 points | Narrows 28% |
Frequently asked questions
Is one estimand more honest than the other?
No. They answer different questions and regulators generally want both. The efficacy estimand describes the drug's pharmacological effect under ideal adherence; the treatment-regimen estimand describes what a population randomised to that strategy actually experienced. Neither is the truth on its own.
Does this mean published weight-loss figures are inflated?
It means the headline figure is usually the efficacy estimand, which is the larger of the two and describes conditions that do not obtain outside a trial. That is not inflation, but it is a systematic reason real-world results come in lower (the same reason covered in the efficacy-effectiveness gap).
Could the placebo arm result be a data error?
It is an unusual pattern but not an implausible one, and it is exactly what post-randomisation treatment in a control arm produces. The full publication would be expected to describe intercurrent events by arm, which is where the explanation would be confirmed or replaced.
Does this affect other obesity trials?
Increasingly, yes. Any long placebo-controlled obesity trial running while effective treatments are widely available faces the same pressure. Check control-arm weight change in any trial you read; a control arm losing several percent of body weight is telling you something about what happened outside the protocol.
Continue researching
Related guides and reference material.
Primary sources & clinical trials
Published papers, trial registrations, regulator pages and company announcements. Links open the original.
Written and published by UK Peptides, the trading name of AURYXRESEARCH LTD (company 17388727, registered in England and Wales).
Questions or a correction? Email hello@ukpeptides.org or call +44 7426 987848.
More GLP-1 & Incretin Science articles
- Six Trials Out of Forty-TwoA Bayesian meta-analysis compared tirzepatide, semaglutide and liraglutide indirectly. It used six of the 42 trials it found, and got a number later confirmed.
- Tirzepatide Has No Active Comparator in the NetworkIn the 2026 obesity network meta-analysis, tirzepatide links to every other drug only through placebo arms. That structure is where the assumptions live.
- Three Studies, Three Numbers, No ContradictionPublished FAERS analyses give semaglutide kidney-injury figures of 0.44, 1.25 and 3.15. All three are correct, because each uses a different comparator.
- Why a Safety Report Count ClimbsTirzepatide ketoacidosis reports rose from 1-2 per quarter in 2022 to 28-34 by 2025. Almost all of that curve is uptake and reporting behaviour, not risk.
- The Complication That Hides Behind a Normal GlucoseKetoacidosis with near-normal glucose is easy to miss: the test that usually raises the alarm looks reassuring. Both incretin drugs show reporting signals.