SURMOUNT-1: How Do You Read a GLP-1/GIP Weight-Loss Trial?

The short version

Reading a modern weight-loss trial well means separating four things that a headline collapses into one number: what the trial measured, how it handled people who stopped the drug or switched to rescue therapy, how long it ran, and who was enrolled. SURMOUNT-1, the phase 3 tirzepatide report published in the New England Journal of Medicine in 2022 by Jastreboff and colleagues, is a clean case study because its authors were explicit about all four. The much-quoted figures of roughly 16 to 22.5 percent body-weight reduction are the product of those design choices, not a free-standing promise. Read in that order, the percentage becomes the last mile of a chain rather than the whole story.

What the trial measured: two co-primary endpoints

Tirzepatide is a single molecule that activates two incretin receptors, GLP-1 and GIP. SURMOUNT-1 randomized 2,539 adults in a 1:1:1:1 ratio to once-weekly tirzepatide at 5, 10, or 15 mg, or to placebo, according to the NEJM report and the registration on ClinicalTrials.gov (NCT04184622).

The trial had two co-primary endpoints, and both matter. The first was the mean percent change in body weight from baseline. The second was the proportion of participants reaching at least a 5 percent reduction. Those measure different things. A mean tells you how far the average person moved, but it can be pulled by a minority of very large responders. A responder threshold tells you how many people cleared a clinically meaningful bar. A drug can look impressive on one and ordinary on the other, so a trial that pre-specifies both is harder to spin. SURMOUNT-1 met both, at every dose, against placebo.

Why the numbers come in pairs

The headline efficacy figures were reductions of 16.0 percent at 5 mg, 21.4 percent at 10 mg, and 22.5 percent at 15 mg, compared with 2.4 percent on placebo. Those come from what the trial called the efficacy estimand. A second set of numbers, from the treatment-regimen estimand, ran a few percentage points lower. Same participants, same trial, two answers. Understanding why is the heart of reading the paper.

How it handled real life: the estimand

An estimand is a precise statement of what a trial is trying to estimate, including how it treats the messy events that happen after randomization: people who stop the drug, drop out, or need rescue medication. This framework is set out in the ICH E9(R1) addendum, the international standard regulators use to keep trials from quietly choosing the flattering interpretation.

The two estimands in SURMOUNT-1 answer two honest but different questions. The treatment-regimen estimand asks, in effect, what happens to weight if you assign people to this strategy in the real world, counting the outcomes of those who came off the drug along the way. The efficacy estimand asks the more idealized question of what happens if people actually take the drug as intended for the full duration. The first is closer to a policy or population question; the second is closer to a pharmacology question. Neither is a trick. They simply frame the effect differently, which is why serious coverage reports which estimand a number came from, and why a lone percentage stripped of that label tells you less than it seems to.

How long it ran: 72 weeks

SURMOUNT-1 measured its primary outcome at 72 weeks, a little under a year and a half. That window is a deliberate compromise. It is long enough to move past the early, partly fluid-driven weight change and to show whether loss is sustained on therapy, but it is still a fixed observation period rather than a lifelong one. Obesity is chronic, and weight trajectories after a drug is stopped or after several years are separate questions that later SURMOUNT-1 follow-up analyses and other trials have gone on to address. When you read any 72-week figure, read it as a snapshot at that timepoint under continued treatment, not as a permanent endpoint.

Who was studied: the entry criteria frame everything

The single most important context for the effect size is who was allowed in. SURMOUNT-1 enrolled adults with a body-mass index of at least 30, or at least 27 with a minimum of one weight-related complication such as hypertension or dyslipidemia. Critically, people with type 2 diabetes were excluded; roughly 40 percent of participants had prediabetes at baseline. Both groups also received lifestyle counseling, a reduced-calorie diet, and increased physical activity.

Two consequences follow. First, results from a population without diabetes do not automatically transfer to people with diabetes, whose weight response to incretin drugs tends to be smaller; that is precisely why a separate trial, SURMOUNT-2, was run in adults who had type 2 diabetes along with overweight or obesity. Second, the placebo arm is not a do-nothing arm. Those participants got the same lifestyle program and still lost about 2.4 percent. The drug effect worth quoting is the gap between arms, not the raw number on the active drug, because both arms carried the same background intervention. Comparing a treated arm against zero would overstate what the molecule itself contributed.

Putting it together

Read in order, SURMOUNT-1 tells a coherent story. It measured two endpoints so a single flattering metric could not carry the claim. It reported two estimands so the difference between an idealized and a real-world question stayed visible. It fixed a 72-week window. And it defined a specific population whose BMI, comorbidity, and diabetes status bound how far the findings travel. A headline percentage is the last mile of that chain, and it means what it means only once you know the four choices behind it.

This article is educational and is not medical advice; decisions about any weight-management therapy belong to an individual and their own clinician.