How to read a peptide study: BPC-157 and retatrutide, broken down

You've seen the posts. "New study shows…", a screenshot of an abstract, a few million views, and a comment section split between people who think it's settled science and people who think it's all made up.

Usually it's neither. A "study" can mean a dish of cells, a dozen rats, a phone survey of seventeen people, or a trial with hundreds of participants who didn't know whether they were getting the real thing. Those are not the same kind of evidence, and the post almost never tells you which one you're looking at.

So here are five questions that do. We'll run two real examples through them: BPC-157, where the evidence is almost all animal work, and retatrutide, where there are proper human trials.

The five questions

1. Cells, animals or people? Research moves through stages. In vitro means in a dish or test tube. In vivo means in a living animal. Then come human studies. Each step is a big jump, and plenty of things that look promising in animals never make it. A review in PLoS Medicine put it bluntly, describing "a recurrent failure of interventions apparently promising in animal models to translate to the clinic". A rat study is a reason to do more research. It isn't evidence about people.

2. Was there a comparison group? Without one, you can't tell whether a change came from the compound or from time, attention or expectation. Three words to look for:

• Randomised: a coin-toss decides who gets what, so the groups start out alike.
• Placebo-controlled: one group gets an identical dummy.
• Double-blind: neither the participants nor the researchers know who got which until the end.

A study with all three is hard to fool. A study with none of them can't tell you much.

3. How many, and for how long? Seventeen people is an anecdote with a spreadsheet. Several hundred is a trial. Duration matters too: a few weeks tells you about a few weeks.

4. What was the pre-registered primary endpoint? Serious trials register in advance, on a public site like ClinicalTrials.gov, exactly what they'll measure and when. That stops anyone measuring twenty things and reporting the one that looked good. If a study has an NCT number, you can check what it promised to measure against what it reported.

5. Who ran it, and where was it published? A peer-reviewed journal is not the same as a press release, a preprint, or a clinic's own write-up. Funding matters too, and good papers declare it. A study paid for by the company developing the compound is normal. It just belongs on the list of things you notice.

The trial phases, in one table

When people say "it's in phase 3", this is what they mean. The descriptions are the FDA's own.

The FDA adds a line worth remembering about phase 2: these studies "aren't large enough to show whether the drug will be beneficial". That's the regulator saying even a good phase 2 trial is a step, not an answer.

Worked example 1: BPC-157

In 2025, a team of orthopaedic researchers published a systematic review in the HSS Journal, the kind of paper that searches the medical databases for every relevant study and sorts them. They searched PubMed, Cochrane and Embase up to June 2024 and included 36 studies.

Of those 36: 35 were preclinical (animals and cells) and 1 was clinical. The review also states: "No clinical safety data were found."

That's the whole picture in one sentence. Almost everything you've read about BPC-157 comes from animal research, which is question 1 answered.

The single human study is worth a look, because it shows what weak human evidence looks like. Published in 2021, it was a retrospective chart review: researchers looked back through the records of 17 patients at one clinic, reached 16 of them by phone, and asked them how they got on. Run it through the five questions:

Nothing about that means the researchers did anything wrong. A chart review is a legitimate first look. It just isn't the kind of study that settles anything, and the authors call it a small study themselves. So when a post says "there are human studies on BPC-157", this is the study, and now you know what it is. How BPC-157 differs from the peptide it's usually paired with is in BPC-157 vs TB-500.

Worked example 2: retatrutide

Now compare the two published phase 2 trials of retatrutide.

The first, published in the New England Journal of Medicine in 2023, was a "phase 2, double-blind, randomized, placebo-controlled trial" of 338 adults, lasting 48 weeks, registered in advance as NCT04881760 with its primary endpoint measured at 24 weeks.

The second, in The Lancet the same year, went further on design: "randomised, double-blind, double-dummy, placebo-controlled and active comparator-controlled". Two terms there are worth knowing:

• Active comparator: as well as a placebo group, one group got an already-licensed drug, so the new compound is measured against something real, not just against nothing.
• Double-dummy: the two treatments were given differently, so everyone received a dummy of the one they weren't getting. That keeps the blinding intact when the two look different.

It enrolled 281 participants at 42 centres in the USA, registered as NCT04867785, with a lab blood marker, HbA1c, as its primary endpoint at 24 weeks.

Side by side:

That's what the difference between weak and strong evidence looks like on paper. But notice where retatrutide still sits: phase 2, which by the FDA's own description isn't large enough to show whether a drug will be beneficial. Its phase 3 programmes are under way, and it's licensed nowhere. Strong design is not the same as a finished story.

Using this on the next post you see

Next time a "new study" post lands on your feed, you don't need a science degree. You need about sixty seconds:

1. Ask which study. If nobody can link the actual paper, that's your answer.
2. Find the species. Search the abstract for "rat", "mice" or "in vitro" before anything else.
3. Look for the three words: randomised, placebo-controlled, double-blind.
4. Count the people. Tens, hundreds or thousands.
5. Check for an NCT number and where it was published.

Five answers, and you'll know more about that study than most of the people sharing it.

Why a research supplier is telling you this

Because it's the honest version of our job. We can't tell you what these compounds do in a person, and we won't, for the reasons in why there is no dosing guide. What we can do is point you at the real papers and show you how to read them.

The other thing you can actually check is the vial itself: what's in it, how pure it is, and whether the report is real. How to read a certificate of analysis is the same kind of skill, applied to a lab report instead of a trial.