GLP-1 Clinical Studies Explained: How to Read the Research

The headlines about GLP-1 medications are written in percentages. "Patients lost 15% of their body weight." "The highest dose produced 21% weight loss." These numbers are real, and they come from genuinely strong trials — but a percentage stripped of its context can mislead as easily as it informs. A 15% figure means something quite different depending on what it was measured against, how long it took, who was studied, and what happened to the people at the edges of the average. Reading the research well means learning to ask those questions before accepting the number.

This page is a guide to doing exactly that. It is not a summary of which drug wins — that comparison lives elsewhere, in the tirzepatide versus semaglutide comparison and in the broader GLP-1 science pillar. The aim here is narrower and more durable: to give you the tools to read any GLP-1 weight-loss trial — including the ones not yet published — and judge for yourself what it does and does not establish. The two trials that anchor the field, STEP 1 and SURMOUNT-1, serve as the worked examples throughout.

What Counts as Good Evidence Here

Not all evidence is equal, and the weight-loss field has historically been awash in the weaker kinds: testimonials, before-and-after photos, short uncontrolled studies, and trials funded to produce a particular answer. The reason the GLP-1 evidence base is taken seriously is that the defining trials sit at the top of the evidence hierarchy. They are large, randomised, double-blind, and placebo-controlled, with weight change specified as a primary endpoint before the trial began rather than discovered in the data afterward.

Each of those terms is load-bearing. Randomised means participants were assigned to drug or placebo by chance, so the two groups should be similar in every respect except the treatment — age, starting weight, motivation, and the hundred other variables that otherwise confound weight studies. Double-blind means neither the participants nor the researchers measuring outcomes knew who was getting the active drug, which matters enormously for a subjective-feeling-laden outcome like appetite. Placebo-controlled means there was a comparison group receiving an inert injection on the same schedule, with the same clinic visits and lifestyle support. Pre-specified primary endpoint means the researchers committed in advance to what they were measuring, so they could not fish through dozens of outcomes and report only the flattering ones. When a study has all four features, its weight-loss number deserves real weight. When it is missing any of them, treat the number with proportional caution.

What the Major Trials Actually Measured

The two trials that reset expectations for the field were STEP 1 and SURMOUNT-1, both published in the New England Journal of Medicine. Understanding precisely what each measured is the foundation for reading everything that has followed.

STEP 1, led by John Wilding and colleagues and published in 2021, tested once-weekly semaglutide at the 2.4mg dose. It enrolled 1,961 adults with overweight or obesity but without type 2 diabetes, and randomised them in a 2:1 ratio to semaglutide or placebo, with both arms receiving lifestyle counselling. The treatment ran for 68 weeks. The pre-specified co-primary endpoints were the percentage change in body weight and the proportion of participants reaching at least 5% weight loss. The mechanism behind these effects — slowed gastric emptying, central appetite suppression, reduced food reward — is the same one Daniel Drucker's 2018 synthesis in Cell Metabolism laid out, and is covered in detail in how semaglutide works for weight loss.

SURMOUNT-1, led by Ania Jastreboff and colleagues and published in 2022, did the parallel job for tirzepatide, the dual GLP-1/GIP agonist. It enrolled 2,539 adults with obesity, or overweight with a weight-related complication, again without diabetes, and randomised them to one of three tirzepatide doses — 5mg, 10mg, or 15mg — or placebo, over 72 weeks. As in STEP 1, the percentage change in body weight was a primary endpoint. The headline results sit in the Zepbound clinical trial results; what matters for reading the research is the shape of the design, which closely mirrors STEP 1 and makes the two broadly comparable — though, crucially, they were separate trials and never randomised head-to-head against each other.

Here is a summary of the headline figures from these two trials, drawn only from the published reports:

Feature	STEP 1 (semaglutide)	SURMOUNT-1 (tirzepatide)
Published	NEJM, 2021 (Wilding et al.)	NEJM, 2022 (Jastreboff et al.)
Participants	1,961 adults, no diabetes	2,539 adults, no diabetes
Duration	68 weeks	72 weeks
Dose(s) tested	2.4mg weekly	5mg / 10mg / 15mg weekly
Mean weight loss	~14.9%	15.0% / 19.5% / 20.9%
Placebo weight loss	~2.4%	3.1%

Two things are immediately worth noticing in this table, and both are the subject of the sections that follow: the active-drug numbers are averages that conceal a wide spread, and they are only interpretable in relation to the placebo column sitting next to them.

How to Read a "% Weight Loss" Figure

The single most important habit in reading these trials is to remember that a reported figure like "14.9%" is a mean — the average across everyone in the group. Averages are useful, but they hide the distribution, and in weight-loss trials the distribution is wide. In STEP 1, about 86% of participants on semaglutide lost at least 5% of their body weight, but the rest did not, and within the responders the range stretched from modest losses to more than 20%. Some people lost a great deal; a minority lost little or nothing at all. The mean is the centre of a broad cloud, not a result each person can expect.

This is why good trial reports publish not just the mean but the responder analysis — the proportion of participants crossing thresholds of 5%, 10%, 15%, and 20% weight loss. These cumulative figures tell you far more about what to expect than the headline average does, because they describe the spread. When you read that "more than half of participants on the highest tirzepatide dose lost at least 20%," you are reading a responder figure, and it is more informative than the mean precisely because it describes the shape of the result rather than its midpoint. For thinking through where you personally might land in that distribution, setting realistic weight-loss goals on GLP-1 works through the practical implications.

A second subtlety hides in the phrase "percentage of body weight." A 15% loss means something physically different for a person starting at 140kg than for one starting at 90kg — the same percentage, very different absolute change. Percentages are used because they make trials comparable and because health benefits track proportional loss reasonably well, but when you translate a trial figure into a personal expectation, the percentage has to be applied to your own starting weight, not borrowed as an absolute number of kilograms.

Why Placebo Subtraction Matters

Here is where casual reading most often goes wrong. The drug groups in these trials did not lose weight in a vacuum; the placebo groups also lost some. In STEP 1 the placebo group lost about 2.4%, and in SURMOUNT-1 about 3.1%. That placebo weight loss is real and it is informative: it captures everything that happened to participants other than the active drug — the lifestyle counselling, the structured clinic visits, the attention, the simple fact of being enrolled in a weight study and watched.

To isolate the effect attributable to the medication itself, you subtract the placebo result from the drug result. This is the placebo-subtracted or treatment difference, and it is the number that most honestly answers "what did the drug add?" For STEP 1, roughly 14.9% minus 2.4% gives a treatment difference of about twelve percentage points. For the top SURMOUNT-1 dose, roughly 20.9% minus 3.1% gives close to eighteen points. These subtracted figures are smaller than the headlines, and they are the ones to quote when comparing across trials, because the raw drug number always includes the placebo effect bundled inside it.

A useful rule of thumb: whenever you see a single weight-loss percentage with no placebo figure beside it, you are looking at an incomplete result. The placebo arm is not a technicality — it is the baseline against which the drug's contribution is defined, and it is what makes the trials' numbers trustworthy where uncontrolled studies and testimonials are not.

Design Details That Change the Interpretation

Beyond the headline numbers, several features of how a trial is built shape what its results mean.

Duration and titration. Both anchor trials ran well over a year, and that length matters because weight loss on these medications is not linear. Dosing starts low and is increased in steps over months — a process called titration, designed to limit gastrointestinal side effects — so the early weeks understate the eventual effect. Weight typically falls fastest in the middle stretch and then flattens toward a plateau as the trial ends. A six-week study of the same drug would report a far smaller number, not because the drug is weaker but because it was measured too early.

The run-in design and what happens after stopping. A particularly instructive variation is the STEP 4 trial, led by Domenica Rubino and published in JAMA in 2021. It gave every participant semaglutide for an initial 20 weeks, then randomised them either to continue or to switch to placebo for 48 more weeks. Those who continued lost an additional 7.9%; those switched to placebo regained about 6.9% — roughly two-thirds of what they had lost. This isolates a different question from STEP 1: not "does the drug work?" but "what happens when an effective drug is withdrawn?" The answer, that much of the weight returns, was confirmed by the STEP 1 trial extension. Reading these correctly means recognising that they describe weight loss during treatment, and that regain on stopping reflects the body's defence of its prior weight rather than a failure of the drug — the biology that Priya Sumithran's 2011 work, Rudolph Leibel's 1995 energy-expenditure studies, and Erin Fothergill's Biggest Loser follow-up all document.

Who was studied. Both STEP 1 and SURMOUNT-1 deliberately excluded people with type 2 diabetes, because diabetes tends to blunt the weight-loss response and the researchers wanted a clean read on weight. That exclusion is a strength for interpretability but a limit on generalisability: the results apply most directly to people without diabetes, and the figures in a diabetic population are typically somewhat lower. Always check the entry criteria before assuming a trial's number applies to you.

What "Clinically Meaningful" Actually Means

The phrase "clinically meaningful weight loss" appears constantly in this literature, and it has a specific, evidence-based meaning rather than a marketing one. The conventional threshold is 5% of body weight: that is the point at which trials begin to show measurable improvements in blood pressure, blood glucose, lipids, and other cardiometabolic markers. A 5% loss is modest on the scale of these drugs — most participants far exceed it — but it is the floor at which the health benefits, as opposed to the cosmetic ones, become detectable.

Above that floor, benefits generally scale with the amount lost. A 10% loss produces larger improvements than 5%; 15% larger still. This is why the responder thresholds matter clinically and not just statistically: the share of people reaching 10% or 15% predicts how much population-level health benefit a drug will deliver. It is also why the magnitude of the GLP-1 results was treated as a genuine shift rather than an incremental one. Older weight-loss drugs struggled to push the average past the 5% clinically meaningful line; the mean losses of roughly 15% in STEP 1 and up to 21% in SURMOUNT-1 sit far above it, in territory that overlaps at the high end with bariatric surgery. For context on what these losses mean for the people who achieve them, weight loss on semaglutide covers the practical trajectory.

The Limitations Every Reader Should Hold Onto

Even strong trials have boundaries, and reading the research honestly means keeping them in view.

• Averages are not promises. The mean describes a population. Individual response varies widely, and the trials do not identify in advance who will fall at which end of the distribution. Your result may land well above or well below the headline.
• The numbers describe treatment, not the aftermath. The weight loss was achieved while taking the drug. STEP 4 and the STEP 1 extension both show that stopping is typically followed by substantial regain, because the underlying biology was countered rather than cured.
• Cross-trial comparisons are indirect. STEP 1 and SURMOUNT-1 were separate trials with different participants. Their numbers can be compared loosely, but because the two drugs were never randomised against each other within a single trial, any "X beats Y" conclusion drawn from comparing them carries more uncertainty than a direct head-to-head would.
• Trial populations are selected. Participants met specific BMI criteria, often lacked diabetes, were willing to inject weekly for a year or more, and received structured support. Real-world populations differ, and real-world adherence is generally lower, which tends to make real-world results more modest than trial results.
• Trial length is finite. The longest of these trials runs a couple of years. The consequences of using these medications for a decade or more — as a chronic-disease framing implies — are still accumulating, simply because the drugs have not been in wide use long enough to know.

None of these limitations undermines the core finding, which is robust: at the population level, these medications produce weight loss of a magnitude no previous pharmacotherapy approached, established through the strongest study designs the field routinely produces. The limitations are not reasons to dismiss the evidence. They are the context that turns a headline percentage into an honest understanding of what the percentage means.

Putting It Together

Reading a GLP-1 trial well comes down to a short checklist. Find the design — is it randomised, blinded, and placebo-controlled? Find the placebo arm and subtract it to see what the drug actually added. Look past the mean to the responder thresholds to understand the spread. Check who was studied and for how long. Note whether the result describes weight loss during treatment or after stopping. And measure the figure against the 5% clinically meaningful line to gauge whether it represents real health benefit.

Apply that checklist to STEP 1 and SURMOUNT-1 and the conclusion holds up: large, well-designed trials showing placebo-subtracted weight loss of roughly twelve to eighteen percentage points, well above the threshold for clinical benefit, achieved during treatment and dependent on its continuation. That is a genuinely strong result, stated honestly — and the same checklist applied to a glossy testimonial shows just as reliably why such sources deserve far less confidence. The broader landscape of this research — the trials, the mechanisms, and the open questions — is mapped across the wider GLP-1 science cluster, which collects the individual studies into a single connected picture.