An estimand is a trial's exact definition of the treatment effect it wants to measure, including how it handles people stopping treatment.
That's the practical answer to “what is an estimand” when two headlines quote different weight-loss averages from the same study. Both figures can be accurate. They may be answering different questions.
What is an estimand in official guidance?
The ICH, an international group that develops guidelines for medicine studies, describes a treatment effect through five choices. Its E9(R1) addendum is marked “Adopted on 20 November 2019.”
- Treatments: which medicine and comparison treatment are being studied.
- People: the population the question concerns.
- Measurement: the outcome, such as weight change at a particular week.
- Treatment changes: how events during treatment affect the question.
- Summary: how the groups are compared, such as a difference in average weight loss.
An intercurrent event is something after treatment begins that changes how a result should be understood. Stopping treatment or adding another weight-loss medicine are examples. The definition belongs in the trial's planning, before the results are known.
Why one trial reports two weight-loss numbers
Starting a medicine and staying on it are different things. A trial needs to describe how its result handles that difference.
STEP 1, reported by John Wilding and colleagues in the New England Journal of Medicine in 2021, studied 1,961 adults without diabetes. It compared weekly semaglutide injections with placebo, an inactive injection, alongside lifestyle support for 68 weeks.
Here is that one trial, read two ways, from the published STEP 1 paper.
| Question the calculation answers | Semaglutide average weight loss | Placebo average weight loss |
|---|---|---|
| Treatment policy: what happens after assignment, including treatment stops and additional weight-loss treatments? | 14.9% | 2.4% |
| Trial product: what would happen if everyone continued assigned treatment without additional weight-loss treatments? | 16.9% | 2.4% |
The gap is 2.0 percentage points, meaning the difference between the two percentages. It doesn't mean the trial used two semaglutide products.
Neither number forecasts an individual's result. For a semaglutide prescription discussion, the useful question is which treatment situation the quoted average describes. The STEP 1 results guide covers the participants, side effects and later follow-up.
Treatment-regimen estimand and efficacy estimand
Different companies use different labels for broadly similar questions. Novo's semaglutide papers use treatment policy and trial product.
Lilly's SURMOUNT-1 paper, Jastreboff and colleagues, NEJM, 2022, uses treatment-regimen and efficacy. Treatment-regimen includes treatment stops. Efficacy assumes the treatment was administered as intended.
| Broad question in these trials | Novo's label | Lilly's label |
|---|---|---|
| Includes what happens when people stop assigned treatment | Treatment policy | Treatment-regimen |
| Assumes people continue assigned treatment | Trial product | Efficacy |
The labels are a useful starting point. A paper's own rules still matter, especially for additional treatments and missing measurements.
At a visit, a tirzepatide average becomes meaningful once the population, treatment duration and handling of treatment stops are clear. The SURMOUNT-1 results guide keeps those details alongside the figures.
What this means for a peptide prescription
An estimand helps a provider discuss expectations from semaglutide or tirzepatide studies. It doesn't establish that a pharmacy's preparation has been tested.
Semaglutide and tirzepatide prescribed through Promise are dispensed by a licensed U.S. compounding pharmacy as compounded medications, which are different from an FDA-approved product: the formulations offered here are not FDA-approved. Compounding means a licensed pharmacy prepares medication for a prescription.
As of September 30, 2026, the day this article was written, the FDA's compounding explanation carries “Content current as of: 06/22/2026.” It states that the agency doesn't review compounded drugs for safety, effectiveness or quality before marketing.
A licensed provider may still prescribe a compounded formulation; that decision is between the patient and the doctor. The prescriber sets the actual dose and follow-up plan.
Mistakes that make a headline misleading
The bigger figure isn't automatically more applicable. A trial-product estimate describes a situation with continued treatment. That can be useful, provided the condition stays attached to the number. Leaving it out changes the meaning.
Press releases don't all lead with the larger figure. Novo's June 4, 2020 announcement put 14.9% first, with 16.9% qualified in the same headline. Highlighting a larger conditional estimate can be accurate; it still leaves out part of the picture for someone starting treatment.
“Only people who finished” misses the point. A trial-product estimate uses a statistical model, a mathematical method for estimating outcomes under stated conditions. It isn't simply an average of people who finished. Garvey and colleagues' STEP 5 methods, Nature Medicine, 2022, show how measurements during treatment inform this estimate for all assigned participants.
Including treatment stops doesn't make a trial ordinary care. Participants still had study visits, eligibility rules and lifestyle support. A treatment-policy result can't predict the outcome for every clinic or patient.
Mixing approaches can distort a comparison. As of September 30, 2026, the day this article was written, a September 28 analysis in the International Journal of Obesity had compared semaglutide and tirzepatide using both approaches. Its broad conclusions were similar, but some findings met the cutoff for separating a treatment difference from chance under only one approach. Lilly funded the analysis, which compared separate trials rather than assigning people directly between the medicines.
For semaglutide, keeping the calculation beside the percentage helps a provider explain what the research actually supports.
Related terms that keep a result in context
An endpoint is the outcome a trial measures, such as percentage weight change. An estimate is the calculated number; the estimand defines the effect behind it.
Intention to treat means analyzing people in the groups they were assigned to. That phrase alone doesn't explain every rule for treatment changes or missing measurements.
Clinical trial phases describe the stage of drug testing. The phases of clinical trials guide explains what each stage can establish.
At Promise, a licensed provider reviews every request, and not everyone qualifies.
