Every large trial has a sentence buried in its protocol that decides what the whole study is allowed to prove. It is called the primary endpoint, and almost nobody reads it. People read the percentage that comes out the other end, quote it in an argument, and move on.
SURMOUNT-1 is a good place to see why that habit causes trouble. It is one of the most quoted obesity trials of the decade, and the gap between what it measured and what people believe it measured has become wide enough to drive a market through.
What a co-primary endpoint actually is
A primary endpoint is the outcome a trial is designed and powered to detect. Everything else in the statistical plan arranges itself around it. Sample size, treatment length and analysis method are all chosen to answer that one question with a defined level of confidence.
A co-primary endpoint means there are two such questions, and both have to be answered before the trial counts as positive. Secondary endpoints are different animals. They are measured, reported and often interesting, and they are not what the study was built to settle. Exploratory endpoints sit a rung lower still.
Once that ladder is clear, a trial stops being a single number and becomes a structured claim with edges.
What SURMOUNT-1 measured
SURMOUNT-1, registered as NCT04184622 and sponsored by Eli Lilly and Company, was a 72-week randomised, double-blind, placebo-controlled trial in adults with obesity or overweight who did not have type 2 diabetes. Participants were assigned in a one-to-one-to-one-to-one ratio across three tirzepatide dose levels and placebo, given once weekly by subcutaneous injection alongside a lifestyle intervention. The design included a two-week screening period and an escalation period of up to 20 weeks before participants reached their assigned level.
Its co-primary endpoints were the percentage change in body weight from baseline to week 72, and the proportion of participants achieving a weight reduction of at least 5 percent at week 72.
That is the entire mandate. Two weight measurements, both at a single time point, both compared against placebo. The full results published in the New England Journal of Medicine report arms of 630, 636 and 630 participants on tirzepatide against 643 on placebo, with the proportion reaching the 5 percent threshold at week 72 running from 85.1 percent in the lowest dose arm to 90.9 percent in the highest, against 34.5 percent on placebo. A separate reported threshold, a reduction of at least 25 percent, was reached by 36.2 percent of the highest dose arm and 1.5 percent of placebo.
Those are real published numbers from a peer-reviewed paper with named arms, defined time points and confidence intervals. They are also narrower than the conversation they get used in.
What the trial did not measure
It did not measure cardiovascular outcomes. It did not measure mortality. It did not measure what happens in year three, year five or year ten, because the treatment period ended at week 72 and a trial cannot report on time it did not run.
It did not study people with type 2 diabetes, who were excluded by design and studied in a separate programme with separate endpoints built around glycated haemoglobin rather than weight.
It did not establish anything about compounded, repackaged or research-grade material carrying the same molecule name. A trial supply chain is a documented, audited thing: characterised material, released batches, chain of custody from manufacture to participant. The registered study describes that supply and no other.
And it did not measure the single thing most people assume a weight trial measures, which is what happens after treatment stops. Separate withdrawal studies exist for that question precisely because this one could not answer it.
Approval attaches to a product, not to a word
Tirzepatide is an approved prescription medicine in the United States, sold under trademarked names by its manufacturer. That approval is not a property of the molecule in the abstract. It attaches to a specific finished product, made to a specific specification, by a specific manufacturer, for specific indications, dispensed under prescription.
A label elsewhere reading the same chemical name inherits none of that. This is the distinction that collapses most often in online discussion, and it collapses in both directions: people assume research material carries the authority of the approved product, and people assume the approved product’s safety record applies to anything sharing its name. Neither follows.
The honest framing is narrow and dull. A trial tested a defined product under defined conditions and reported defined outcomes. Everything outside those conditions is outside the evidence.
Why this reads differently in Spanish
The SURMOUNT-1 paper is in English. The registry entry is in English. The regulatory documents are in English. For a reader in Costa Rica, the primary literature on this compound sits behind a language barrier while the marketing arrives in Spanish, which is a reliable recipe for the research-grade and the approved-medicine categories bleeding into each other.
Local Spanish-language pages are where that line either holds or breaks. On tirzepatida.cr, the FAQ answers the question directly: the material is for scientific research only, and it is not for human consumption. The site describes the compound as one studied for its interaction with the GLP-1 and GIP receptors, which matches how the trial literature frames the mechanism rather than overstating it.
The same framing continues on the about page at Tirzepatida Costa Rica, which describes lyophilised material supplied with a certificate of analysis and a specification sheet, stored under controlled temperature, intended for laboratory work. Whatever a reader concludes about the market itself, a Spanish page that repeats the English literature’s boundaries is doing something that a page translating marketing copy never does.
Three questions before you quote a trial number
Ask what the primary endpoint was, and at which week it was assessed. A number pulled from week 72 means nothing attached to a claim about year five.
Ask what the comparator was. Against placebo is a different statement from against another active drug, and trials of each type exist for this compound.
Ask who was excluded. Trial populations are built by exclusion criteria, and the people left out are often exactly the people reading about the result.
None of this requires statistical training. It requires treating a trial as a question with a defined scope rather than a verdict on a molecule. The SURMOUNT-1 paper has been publicly readable since 2022, and the two sentences that define what it set out to prove sit in the abstract, above everything anyone has argued about since.