The Trial Landscape, Current

The GLP-1 era rests on a handful of large trials, and the honest way to read them is to separate what they measured from the percentage people quote. This page maps the landmark set — STEP, SURMOUNT, SELECT — explains what each number does and doesn't mean, and looks at the next generation of agents with the appropriate caution: early data appears promising, and follow-up is still short.

How to Judge a Weight-Loss Trial

Before the numbers, the reading order. A trial headline percentage is only as good as four things: whether the result is measured against placebo or is just a before-and-after; how long the follow-up actually ran; who was enrolled (which determines what the result applies to); and how many participants dropped out, because people who leave early often leave because the drug disagreed with them. Every number on this page should be read through that filter. The trials below were large, randomized, placebo-controlled, and ran a year or longer — that is the standard of evidence this field now holds, and it matters because the first generation of weight-loss drugs never met it.

Two honest glosses on that chart. First, these are group means at fixed time points — spread around them is wide, and a mean of −15% includes people who lost twice that and people who lost almost nothing. Second, the difference from placebo is the real effect: tirzepatide's roughly 18 percentage points over placebo is a genuinely large effect for this class of study. Neither statement makes the number a personal forecast.

The Landmark Set

The set tells a coherent story. Efficacy is established — two different molecules, several thousand participants each, consistent 10–20% losses. SELECT then moved the question from weight to outcomes: in adults with overweight or obesity and established cardiovascular disease, semaglutide reduced major cardiovascular events by roughly a fifth compared with placebo, an effect that appeared to run ahead of the weight loss itself (Lincoff et al., New England Journal of Medicine, 2023). And STEP 4 supplied the discipline the other trials glossed over — what happens when the drug stops — which the withdrawal page of this series digs into properly.

Who Was Enrolled, and What That Means

The trials enrolled people with clinical indications: obesity (a BMI of 30 or above) or overweight (27 or above) plus a weight-related condition. That matters in two directions. For people inside that envelope, the results genuinely apply. Outside it, the evidence thins quickly — the trials did not test "the last five kilograms for a beach season," and treating that as an indication is a stretch the data never made. This is also why the numbers here are far from what any one person loses on a maintenance-style dose. The parent GLP-1 era page walks the medical boundary in detail; the short version is that these are medicines with indications, decisions that belong to a prescribing clinician, and effects that are earned against a real side-effect ledger.

What the Trials Didn't Measure

• 🧊 Body composition by the book. Weight is the headline; the mix of fat and muscle is the subtext. Secondary analyses in the STEP and SURMOUNT programs found a meaningful slice of the lost weight is lean mass — the lean-mass share page owns that finding.
• ⏳ The decade view. Two to three years of trial follow-up is not a decade, and the muscle, bone, and organ questions need exactly that longer window.
• 🤢 Real-world tolerance. Trial retention is optimistic; real-world adherence is worse. Effectiveness in the wild almost always trails efficacy in the protocol.
• 🔁 Life after the drug. STEP 4 and the STEP 1 extension showed much of the loss returns on discontinuation — the durability question the withdrawal page makes its subject.
• 🧬 Longevity itself. No trial has yet shown a mortality benefit in generally healthy people. SELECT showed cardiovascular benefits in a high-risk group; that is a very different claim.

Practical Rules: Reading a Result, Not a Headline

The tools on this page compress into a checklist you can run on any weight-loss claim — a press release, a podcast quote, a friend's report — in well under a minute. Run it before the number lands in your head:

• 📐 Mean or median? A headline that quotes a mean is quoting an average of a wide, right-skewed spread; the median — the middle participant — is usually lower. "Up to" numbers are the tail, not the center.
• 🧪 Against what? A placebo-controlled difference is the real effect; a before-and-after is the effort plus regression plus the study's attention. They are not the same number dressed differently.
• ⏳ For how long? A one-year result is a one-year result. Extension data is where durability claims get made — or quietly dropped.
• 👥 In whom? Match the population to your situation before matching the number to your expectations. An obesity-trial mean says nothing about a healthy-weight supplement user.
• 🚪 Who left? High dropout is a red flag in both directions — people leave for side effects, and people leave because they stopped bothering. Either way the remaining sample is not the whole story.

That checklist is the honest bridge from trial to person. It will not tell you what will happen to you — nothing will — but it will tell you which numbers deserve your attention and which were engineered to deserve your wallet.

The Next Generation

The pipeline behind semaglutide and tirzepatide is where the honest hedging earns its keep. Newer agents — triple-hormone receptor agonists, oral peptide candidates, and fixed-dose combinations, among them retatrutide, orforglipron, and CagriSema — are in later-stage trials, and early results in some appear to reach deeper than the current pair. The phrase "appear to" is doing real work: several of these reports carry follow-up measured in months rather than years, head-to-head data is limited, and every new molecule inherits the same durability and side-effect questions the established class still has open. The practical stance for a reader is simple: a class of medicine that is this young, and this effective, is worth watching — and watching is not the same as having seen its full report card.

Questions, Answered Briefly

• 📊 Which trial result should I quote? The one with the matching population. For most people weighing a decision, STEP 1's −14.9% and SURMOUNT-1's −20.9% bracket the class; SELECT is the answer to "do outcomes improve," not "how much will I lose."
• 🔬 Are the new agents better? Early data from next-generation molecules appears promising — some show larger mean losses at short follow-up. "Promising" is the precise word: short windows, sparse head-to-head data, and the same unanswered durability questions.
• 🕐 How long was follow-up, really? About a year to 72 weeks for the headline losses, with extensions out to roughly two to three years for some populations, and a cardiovascular-outcome trial running about three years. None of that is a decade.
• 👨‍⚕️ Does this page recommend a drug? No. These are prescription medicines for stated indications; whether one fits your situation is a conversation with a clinician. This page exists to make that conversation better-informed, not to make it for you.

Related Topics