If you only want the short version. "No human evidence" covers four different situations, and a product page shows the same blank space for all of them.
Nobody ever tried. Somebody tried and stopped. Somebody tried and it did not beat what already existed. Or somebody tried and never published the result.
BPC-157 is the fourth kind. It went through two human trials inside a pharmaceutical company in the early 2000s, and neither was ever published.
A full development programme runs roughly a billion dollars over 10 to 15 years, and around 86% of drugs entering human trials never reach approval. For most peptides nobody has run that arithmetic and decided against it. Nobody has run it at all.
Someone already ran the trial. They finished it, and then never told anyone what they found.
BPC-157 spent thirteen years inside a pharmaceutical company. PLIVA, in Croatia, took it into human trials twice. Then the company was broken up and sold, the new owner dropped the project, and the results went with them.
So when a peptide is described as having no human evidence, that can mean several very different things. It can mean nobody ever looked. It can also mean somebody looked, and never said.
The four situations
| Situation | What happened | How common | How it looks on a product page |
|---|---|---|---|
| Nobody ever tried | Interesting in animals, no company took it further | By far the most common | No human evidence |
| Somebody tried and stopped | Development began, money was spent, the project was dropped | Sometimes for safety, often for business reasons | No human evidence |
| It did not beat what existed | The compound worked and matched an existing treatment | A commercial failure, not a scientific one | No human evidence |
| Tried, never published | The trial ran, results exist, no paper followed | Unpublished results lean negative across medicine | No human evidence |
Nobody ever tried. By far the most common. Something looked interesting in animals and no company took it further. Nothing was found because nothing was looked for.
Somebody tried and stopped. A company began development, spent real money, and then dropped it. Sometimes for safety, often for business reasons that have nothing to do with whether it worked.
Somebody tried and it did not beat what already existed. A drug does not just have to work. It has to work better than whatever doctors are already using. A compound that matches an existing treatment is a commercial failure even if it is a scientific success.
Somebody tried and never published. The trial happened, the results exist somewhere, and they never became a paper. It has been tested, and nobody can tell you what happened.
Studies that find nothing are published far less often than studies that find something, across all of medicine. So an unpublished trial leans negative. Not proof of anything, and a long way from nothing.
From outside, the last three are indistinguishable from the first. A product page shows you the same blank space either way.
What a trial costs
Discovery to approval takes 10 to 15 years and somewhere between one and three billion dollars.1 Around 86% of compounds that reach human testing never get approved.2
So nobody is asking whether a compound might work. They are asking whether the odds of success, multiplied by the earnings if it succeeds, clear a bill in the hundreds of millions that will most likely buy nothing.
For most peptides the answer is no, and it has nothing to do with anyone thinking the compound is bad.
The patent problem is not the one people describe
You will read that peptides cannot be patented, so nobody can profit from them, so nobody tests them.
A sequence that occurs in nature generally cannot be patented in its own right, which is true for compounds like thymosin alpha-1.3 But BPC-157 is patented, with an application going back to 1989.4 So that explanation collapses on the compound people use it about most.
The actual weakness is that a peptide patent protects a very small thing. Change one amino acid, or stick a chemical cap on one end, and you have a molecule the patent does not cover.5 Spend a decade and hundreds of millions proving yours works, and the month you succeed somebody sells a near-identical version you cannot stop. The protection exists. It is one amino acid deep.
What actually happened with BPC-157
People reach for BPC-157 to argue that nobody ever tested these compounds. Its history says the opposite.
It was isolated in Croatia in 1989. In 1993 the Croatian pharmaceutical company PLIVA licensed it and gave it a drug code, PL 14736. They ran two early human trials, both giving it as an enema for ulcerative colitis. The second was randomised, double-blind and placebo-controlled, which is the real thing.
A short write-up in 2005 said that trial found a positive effect, and that the effect did not reach statistical significance. In plain terms, the numbers moved in the right direction and not far enough to rule out chance. Neither trial was ever published in full, and neither shows up in PubMed.
Then the company came apart. In 2006 PLIVA's research institute was sold to GlaxoSmithKline and the rest of the business went to Barr Pharmaceuticals, then to Teva. GSK inherited BPC-157 and dropped it.
Sikiric, the researcher who found it, says his colleagues were ready to start Phase 3 and that a promising drug was abandoned in the restructuring. Nobody outside those companies can confirm or contradict that.
A trial of an oral version was registered in 2015. Data was submitted, then withdrawn before outside review, and the trial is listed as cancelled.4
So the most important evidence about the most popular peptide on the market is in a filing cabinet, and has been for twenty years.
What "it worked in rats" is actually worth
A 2024 analysis tracked 367 treatments across 54 diseases from their first animal study onward. Half made it into human studies at all. Forty percent reached a large randomised trial. Five percent ended up approved.6
But the same analysis found something that cuts against the simple version of this story. When a treatment did reach human trials, the animal and human results agreed 86% of the time.6
Both figures are real. Animal work points in the right direction reasonably often, and tells you almost nothing about how big the effect will be in a person or whether something will go wrong at human scale.
Around 500 compounds have been reported to reduce damage in animal models of stroke. Two have worked in people.7
More than 200 treatments have been reported effective in the standard mouse model of Alzheimer's disease. None has worked in a human trial.7
Neither field is careless. Both are among the most heavily funded in medicine. What defeats them is that a mouse bred to develop a mouse version of a disease has never been a person with that disease.
Most peptides have not entered that pipeline at all, so the five percent does not even apply to them yet.
The part people get backwards
Untested and failed are the same thing from where you are standing. Both mean nobody has shown it works in a person, and "nobody checked" is evidence of nothing except that nobody checked.
Safety is the sharper version. Early trials exist specifically to find safety problems, and those are the trials that never happened.
What the economics change is not your confidence in the compound. They change how the gap reads. A peptide sitting in animal-study limbo for thirty years is the ordinary outcome, not a cover-up and not a verdict.
What to look for on a compound page
Ask which of the four situations applies. Our compound pages say when a programme was started and dropped, which is different information from never having been tried.
And separate "no evidence" from "evidence against." Opposite claims, both routinely written up the same way.
Common questions
Does no human evidence mean it does not work?
Usually it means nobody ran the study. Untested and failed look the same from where you are standing, and neither is a reason to assume it works.
Was BPC-157 ever tested in people?
Yes, twice. A pharmaceutical company ran a Phase 1 safety study and a randomised placebo-controlled Phase 2 study in ulcerative colitis in the early 2000s. A 2005 write-up said the second found a positive effect that did not reach statistical significance. Neither was published in full.
Why do not drug companies just test these?
A full programme runs one to three billion dollars over 10 to 15 years, and around 86% of compounds entering human trials never get approved. The expected return has to clear a bill that will most likely buy nothing.
If it worked in animals, does that count for anything?
Some. Across 367 treatments in 54 diseases, five percent ended up approved, but animal and human results agreed 86% of the time when something did reach trials. Animal work points in the right direction and is poor at predicting size. Roughly 500 compounds have reduced damage in animal stroke models. Two have worked in people.
If thousands of people use it, is not that evidence?
No. Reports cannot separate the compound from everything else happening in those people's lives, which is the entire job of a control group. Popularity generates confidence and no data.
References
- Wouters OJ, McKee M, Luyten J. [Estimated Research and Development Investment Needed to Bring a New Medicine to Market, 2009-2018](https://pubmed.ncbi.nlm.nih.gov/32125404/). *JAMA* 2020;323(9):844-853. *Peer-reviewed analysis, abstract read at source. After accounting for the costs of failed trials, median capitalised research and development investment to bring a new drug to market was $985.3 million, 95% CI $683.6 million to $1,228.9 million, with a mean of $1.3 billion. The paper notes that published estimates range from $314 million to $2.8 billion, which is why this page gives a range rather than a figure. The higher end, commonly quoted as $2.6 billion, comes from DiMasi JA, Grabowski HG, Hansen RW, Innovation in the pharmaceutical industry: new estimates of R&D costs, Journal of Health Economics 2016;47:20-33, and is disputed on methodology.* ↩
- Clinical trial success rates, as applied in [Wouters et al. 2020](https://pubmed.ncbi.nlm.nih.gov/32125404/). *Derived from the same paper, read at source. The analysis assumes success rates of 14% for compounds entering Phase 1, 35% entering Phase 2 and 59% entering Phase 3. A 14% success rate from first human dosing means roughly 86% of compounds entering clinical trials never reach approval.* ↩
- [Why promising peptides will never be proven](https://www.sacfirm.com/blog/why-promising-peptides-will-never-be-proven-the-hidden-economics-behind-therapies-your-doctor-cant-recommend/), 2026. *Legal and economic commentary. Naturally occurring peptides generally cannot be patented in their native form, limiting the commercial case for large trials.* ↩
- [The hidden history of BPC-157](https://undark.org/2026/02/03/bpc-157-peptide-fda/), Undark and STAT News, supported by the Pulitzer Center, 2026. *Investigative reporting. Isolated 1989; licensed by PLIVA in 1993 as PL 14736; two unpublished ulcerative colitis enema trials, the second reporting a positive effect that did not reach statistical significance; research institute sold to GSK in 2006 and the project dropped; a 2015 oral trial withdrawn before outside review.* ↩
- [Why BPC-157 is not FDA approved](https://abud.substack.com/p/the-truth-about-bpc-157-why-it-isnt), 2025. *Commentary. Describes how small peptide patents are circumvented by single amino acid changes or added chemical modifications, undermining the commercial case for expensive trials.* ↩
- [Analysis of animal-to-human translation shows that only 5% of animal-tested therapeutic interventions obtain regulatory approval for human applications](https://journals.plos.org/plosbiology/article?id=10.1371%2Fjournal.pbio.3002667). *PLOS Biology*, 2024. *Systematic review and meta-analysis of 367 interventions across 54 diseases. 50% progressed to human studies, 40% to randomised trials, 5% to approval, with 86% concordance between positive animal and clinical results.* ↩
- [Why do over 90% of behavioral neuroscience results fail to translate to humans](https://academic.oup.com/ilarjournal/article/55/3/438/645837). *ILAR Journal*, 2014. *Review. Cites over 200 interventions effective in the APP mouse model of Alzheimer's with none effective in humans, and approximately 500 compounds effective in animal models of acute ischaemic stroke with two effective in humans.* ↩