How peptide research works, and where most peptides stop

Every drug claim rests on a study, and every study sits somewhere on a ladder that runs from a dish of cells to tens of thousands of patients followed for years. The rung matters more than the result. A peptide that heals tendons in rats has shown something real about rats. Whether it does the same in people is a separate question that only a human trial answers.

The ladder

Preclinical. Cell studies (in vitro) test whether a compound does anything at a receptor or in a tissue. Animal studies, mostly in mice and rats, test whether it does something in a living body and whether it is obviously toxic. Most compounds that look good here fail later.

Clinical. Human testing follows a sequence the FDA describes on its own drug development pages:

The GLP-1 drugs went the whole way. Semaglutide's STEP 1 weight trial randomised 1,961 adults for 68 weeks. Its cardiovascular trial, SELECT, enrolled 17,604. Those are the numbers behind the claims on the GLP-1 clinical trials page.

Where most popular peptides stop

BPC-157 is the most discussed example. A 2025 systematic review in orthopaedic sports medicine screened 544 articles and included 36 studies: 35 preclinical and one clinical, a retrospective series in which 7 of 12 people given knee injections reported relief lasting more than six months. It found no clinical safety data. A separate 2025 narrative review counted three human pilot studies in total and called for well-designed trials before the compound is considered more than investigational. The animal work is extensive. The human evidence is a handful of small, uncontrolled reports.

TB-500 is in a similar place. It is sold as a fragment of thymosin beta-4, which is how FDA's compounding pages name it. The BPC-157 and TB-500 pages have the full study lists.

Why do so many stop at the animal rung? The reasons usually cited are cost and incentive. A phase 3 programme is expensive, and a short, natural or long-published sequence can be hard to protect commercially, so the usual sponsor of a large trial has less reason to fund one. Popularity does not generate trials on its own.

How to read a PubMed abstract

Most claims online trace back to an abstract. Six questions sort most of them:

1. What species? Look for "rats", "mice", "in vitro" or "cell line" in the methods. If it is there, the result is preclinical.
2. What design? "Randomised, double-blind, placebo-controlled" is the strongest common design. "Retrospective", "case series" and "observational" describe associations, not cause.
3. How many people, for how long? Small trials can miss real effects and overstate chance ones. The 2008 MK-677 trial, with 65 participants, states plainly that it was too small to evaluate function.
4. What was the primary endpoint? A trial is built to answer one question. Secondary and exploratory results are hypotheses.
5. Absolute or relative? SELECT's headline is a 20% lower risk (hazard ratio 0.80, 95% CI 0.72 to 0.90). In absolute terms, events fell from 8.0% to 6.5% over a mean of about 40 months. Both are true; the second tells you more.
6. Who paid? Industry funding does not make a result wrong, and most large drug trials are industry funded. It is context, not a verdict.

Surrogate endpoints

A surrogate endpoint is, in the FDA's words, "a marker, such as a laboratory measurement, radiographic image, physical sign, or other measure" that is not itself a direct measure of benefit. HbA1c and LDL cholesterol are on FDA's table of surrogates accepted for traditional approval, because decades of outcome data link them to what matters. A rise in IGF-1 after CJC-1295, or a change in a tissue stain in a rat, is a measurement. Whether it predicts a benefit a person would notice is unproven until someone measures that benefit directly.

Where regulators place these compounds

FDA keeps a list of bulk substances that may present significant safety risks in compounding, often called Category 2. As of the version current on 22 April 2026, it includes ibutamoren (MK-677), kisspeptin-10, GHRP-2 and GHRP-6, and ipamorelin for outsourcing facilities. BPC-157, TB-500, CJC-1295 and several others appear in a separate table of substances "previously in category 2" that "were withdrawn by the nominators". Leaving that list does not make a substance compoundable: under section 503A a pharmacy may only compound from a bulk substance with a USP monograph, one that is a component of an approved drug, or one on the 503A bulks list. FDA's Pharmacy Compounding Advisory Committee reviewed seven of these peptides on 23 and 24 July 2026, and its recommendations are advisory. Our write-up of that meeting has the votes.

Products sold online as "research use only" sit outside all of this. FDA's page on unapproved GLP-1 drugs says it has warned companies selling semaglutide, tirzepatide, retatrutide and others "falsely labeled" as being for research purposes or not for human consumption, while selling them to consumers. The research chemical entry explains the label.

How we use this on the site

Every compound page says which rung its evidence reaches, and the editorial policy sets out how we grade it. For the case for why researchers keep pursuing peptides at all, see why researchers study peptides.

If you are using any compound, a dated log of dose, symptoms and, on Premium, bloodwork turns a single experience into a record a clinician can read. Logging is free in the app; the download page has the links.