An approval is for a claim, not a molecule
Semaglutide was approved in 2017. It was approved again in 2021, again in 2024, and twice more in 2025, to count only five of its approvals. The molecule did not change between those dates. What changed each time was a sentence on the label: a new thing the manufacturer was allowed to say the drug does.
This is the part of drug regulation that most reporting skips, and it explains a great deal. “FDA-approved” is never a property of a molecule on its own. It is a property of a molecule for a stated use, backed by a trial that measured that use. To read any claim about what a peptide drug does, the first question is which approval the claim belongs to, and what that trial had to count.
| The claim | What the trial counted | Grade | Approved |
|---|---|---|---|
| Lowers blood sugar | HbA1c, a blood marker of average glucose | Validated surrogate | Ozempic, 2017 |
| Reduces body weight | Percent change in weight against placebo, over at least a year | Direct measurement | Wegovy, 2021 · STEP 1 |
| Prevents heart attack and stroke | Counted events: cardiovascular death, heart attack, stroke | Clinical outcome | Wegovy, March 2024 · SELECT |
| Slows kidney disease | Counted events: kidney failure, steep loss of function, death | Clinical outcome | Ozempic, January 2025 · FLOW |
| Treats fatty-liver inflammation (MASH) | Liver biopsy findings | Surrogate, accelerated approval | Wegovy, August 2025 · ESSENCE |
Three kinds of thing a trial can measure
The right-hand side of that table is doing the work. Every trial has a primary endpoint: the one measurement, chosen in advance, on which it succeeds or fails. Endpoints come in three grades.
- Clinical outcomes. how a patient feels, functions or survives. Heart attacks, strokes, kidney failure, death. These are what medicine is for, and they are slow and expensive to count: thousands of people followed for years until enough events occur
- Validated surrogates. a laboratory or physical measurement with a long record of predicting outcomes. Blood sugar and blood pressure are the classic cases. A trial can read them in months
- Reasonably likely surrogates. a measurement that should predict benefit on the biology but has not yet been shown to. These support only a provisional kind of approval, covered below
FDA publishes the list of surrogates it has accepted (FDA). The trade is always the same. A surrogate buys speed and costs certainty, because a drug can move the marker without moving the outcome. This is the same gap mechanism is not proof describes one level down: a marker moving is evidence, and it is not the thing itself.
The bar for weight is lower than the headlines suggest
Body weight is unusual, because it is both easy to measure and the thing patients are asking for. FDA treats it as an endpoint in its own right, and the bar it sets is modest. Under the agency’s January 2025 draft guidance, a weight-reduction drug is generally considered effective if mean weight loss beats the comparison group by at least five percentage points after at least a year, and the difference is statistically significant (FDA draft guidance).
Semaglutide cleared that by a wide margin in 2021: 14.9% against 2.4% on placebo (Wilding et al., NEJM, 2021). Everything since has cleared it by more. So the approval question for a new obesity drug is no longer whether it works on the scale. The live questions are the ones the five-point rule does not ask: what is being lost, what happens on stopping, and whether anything beyond weight improves. (Which of a trial’s weight figures counts toward the bar is its own subject: see two numbers, one trial.)
A cardiovascular outcomes trial is not required to approve a drug for weight reduction. An obesity drug can reach the market with no evidence either way on whether it prevents heart attacks. That is a separate claim, and it has to be earned separately.
Safe for the heart is not the same as good for the heart
Heart trials are where the most careless reading happens, because two different results are both reported as “positive.”
A cardiovascular trial counts major events (cardiovascular death, heart attack, stroke) in the drug group and the placebo group and expresses the comparison as a ratio. A ratio of 1.0 means no difference. Below 1.0 favours the drug. Because any trial is a sample, the ratio comes with a range of values the data are compatible with, and the conclusion depends on where that whole range sits.
- A safety result. the range excludes meaningful harm. The trial has shown the drug does not raise risk beyond a margin set in advance. This is called non-inferiority, and it supports the statement that the drug is not dangerous to the heart
- A benefit result. the whole range sits below 1.0. The trial has shown the drug lowers risk. This is superiority, and only this supports a claim of protection
The SELECT trial of semaglutide was a benefit result: major cardiovascular events fell by about 20% in people with obesity and existing heart disease, and that became a new line on the Wegovy label in March 2024 (Lincoff et al., NEJM, 2023).
The newer molecules are not there yet, and the two on this site illustrate both other positions. Survodutide’s cardiovascular trial reported in November 2026 that the drug did not increase events against placebo: a safety result. Retatrutide’s reported a ratio of 1.12 with a range from 0.64 to 1.96, which spans both benefit and harm and so establishes neither; the trial was too small in events to say (see the GLP-3 bull case). Neither result is bad news. Neither is a claim of protection, and assuming a new drug inherits semaglutide’s heart benefit because it produces more weight loss is an inference no trial has yet tested.
Accelerated approval: a promise with a deadline
For serious conditions with few treatments, waiting years for clinical outcomes has a cost of its own. Accelerated approval is the compromise. FDA may approve on a surrogate that is reasonably likely to predict benefit, on the condition that the sponsor runs a confirmatory trial afterwards, and the approval can be withdrawn if that trial fails (FDA).
Semaglutide’s fifth approval is of this kind. In August 2025 Wegovy was cleared for MASH, a form of fatty-liver disease with inflammation and scarring, on the basis of liver biopsies: more people on the drug saw the inflammation resolve without the scarring getting worse. Biopsy findings are a surrogate. Whether treated patients go on to have less cirrhosis and liver failure is what the continuing part of the trial has to show. The same pathway is how elamipretide reached its first approval.
An accelerated approval is a real approval. The drug can be prescribed and the label says so. But it is a different epistemic object from an approval on outcomes, and a careful reader keeps the two apart.
What this changes about how to read a claim
- Ask which claim. An approved drug is approved for something. A benefit that is not on the label has not been reviewed, however plausible it is
- Ask what was counted. Events in patients, a validated marker, or a marker that is only expected to predict benefit. The same word, approved, covers all three
- Ask who was in the trial. SELECT enrolled people with existing heart disease. Its result is evidence about them, and weaker evidence about anyone else
- For heart results, ask which test was passed. Not harmful and protective are different findings
- Do not carry a result from one molecule to the next. Sharing a receptor is a reason to run the trial. It is not a substitute for it
The same discipline applies with more force to peptides that have never been through any of this. An approved drug has cleared at least one endpoint in at least one trial. A research peptide sold on the strength of its mechanism has cleared none, which is the argument of the trial that hasn’t reported.
The short version
- Approval attaches to a claim. Each line on a label rests on its own trial with its own endpoint.
- Three grades of endpoint. Clinical outcomes, validated surrogates, and surrogates only reasonably likely to predict benefit.
- The weight bar is five points over a year. Clearing it says nothing about the heart, the kidneys or what is lost.
- Safe and protective are different results. Ruling out harm is not showing benefit.
- Accelerated approval is provisional. It rests on a surrogate and a confirmatory trial still owed.
Keep going
Educational reference on how drug approvals are structured, summarized and simplified from FDA’s public pages and the published trials. Not medical or regulatory advice. Drugs are named to explain the method, not to endorse any use. Investigational, not approved: retatrutide and survodutide. Approved in the United States: semaglutide, tirzepatide and elamipretide. Status checked November 2026.