Chapter 17: BPC-157 — The "Healing Peptide": Remarkable Animal Data, Zero Human Trials, and What That Means

"Absence of evidence is not evidence of absence." — Carl Sagan, The Demon-Haunted World (1995)

Overview

That epigraph gets quoted at me about this compound more than about any other molecule in this book, almost always by someone defending it. And they are right. It is a true sentence and it is correctly applied here.

It also cuts exactly as hard the other way. Absence of evidence is not evidence of presence either. When nobody has looked, you do not get to claim the answer — in either direction. That symmetry is the whole of this chapter.

Here are the two facts you need to hold at once.

First: there is a substantial body of animal research on BPC-157, spanning tendon and ligament healing, protection of the gastrointestinal tract against damaging agents, vascular and angiogenic responses, and neuroprotection. It comes from more than one research group, it appears in peer-reviewed journals, much of it is competently conducted, and the direction of the reported effects is strikingly consistent. It is not junk science, and any treatment of this compound that implies otherwise is wrong.

Second: as of this writing in 2026, there is no completed, peer-reviewed, randomized controlled human trial of BPC-157 for any indication in the published literature. Not one. There is no published Phase I safety study either.

Most writing about this compound fails by refusing one of those facts. Sales copy drops the second. Debunking drops the first, usually with a sneer. Both leave you unable to think, because the interesting question is not which fact to believe — they are both true — but how they can both be true at once, and what a rational person does with that.

That question has a real answer, and it is not "the science is fake" or "the trials just haven't caught up yet." It is structural, and once you see it you will see it in a dozen other compounds. You already have the tools: Chapter 5's evidence ladder, Chapter 6's anatomy of how a claim becomes a market, and Chapter 4's five barriers. This is where those three chapters get their hardest workout.

The rating this chapter arrives at is ❌. Before deciding what that means, read §17.10, which writes it out in full — including, at length, what it does not say.

In this chapter, you will learn to:

• Read BPC-157's sequence and name for what they genuinely tell you, and stop where they stop
• Present the animal literature at its strongest, because that is the only version worth evaluating
• Apply Chapter 5's five reasons to a real case with real stakes
• Verify the human evidence situation yourself, in public databases, in about ten minutes
• Explain why a cut rodent tendon and a worn human tendon may not be the same problem
• Take apart an oral-peptide claim using chemistry alone
• Tell the difference between "no reported harm" and "shown to be safe"
• Write an honest ❌ — the hardest sentence in evidence communication

Learning Paths

All five paths should read §17.1, §17.5, §17.6, §17.10, and §17.11. Whatever brought you to peptides, this chapter is the book's worked example of what to do when the evidence is genuinely incomplete — and that situation is far more common than either certainty or fraud.

💊 GLP-1 — read §17.5 and §17.6 closely. You are used to a compound with an enormous human trial program. This chapter shows you what the same field looks like without one, and the contrast is the most useful thing you can carry back to Part II. 🏋️ Performance — this is your chapter. Read every word of §17.3 before you read §17.5, and do not skip §17.7; the transected-tendon mismatch is the single most important idea here for anyone thinking about soft-tissue injury. 🔬 Science — §17.2 (structure, and the absence of an identified receptor), §17.5 (the filtered literature), and §17.11 (the trial design) are the sections with the most technical content. 💄 Cosmetic — §17.8 transfers directly. The five-barrier analysis for a swallowed peptide is the same shape as the barrier analysis for a topical one in Chapter 30; learn the move here where the chemistry is unambiguous. 🏥 Clinical — §17.6, §17.9, and §17.10. You will meet patients using this. §17.10 is written partly for the conversation you are going to have with them, and the posture matters more than the content.

17.1 Why this chapter exists

Let me be direct about something, because pretending otherwise would insult you.

A large share of the people reading this book arrived because they searched for this compound. Some of you are deciding whether to try it. Some of you have been using it for months and want to know whether you have been sensible or foolish. Some of you are clinicians who have been asked about it three times this month and want something better to say than "I'd avoid that."

I do not know which you are, and it does not change what I owe you, which is an accurate account.

Here is the failure mode I want to avoid. Most authoritative-sounding writing on this compound is written by people who have already decided that anyone interested in it is a mark. You can hear it in the prose: it leads with "unproven," moves quickly to "bodybuilders," and arrives at a conclusion that was clearly available before any evidence was consulted. That writing is often factually correct and almost entirely useless, because the reader it was aimed at stops in the second paragraph and returns to sources with worse information and better manners.

I am not going to do that, for a reason that has nothing to do with politeness. A dismissal is not an analysis. If I tell you BPC-157 is unproven and stop, I have given you a conclusion you cannot check, cannot update, and cannot apply to the next compound. If I show you what the evidence actually consists of, where it sits on the ladder, and precisely what would move it, you can do that work yourself — here, and on the next compound, when I am not around to help.

So this chapter has one intellectual task, and it is not "debunking."

The task is to explain how a serious, multi-model, multi-group animal literature and a completely unanswered human question coexist without contradiction.

That is not a paradox. It is the ordinary condition of biomedical research, and the field has a number for it: roughly nine in ten compounds that enter human trials never reach approval. And that number describes compounds that made it into human trials — meaning they had already cleared preclinical work, toxicology, and manufacturing, and someone had committed the money. The attrition from "promising animal result" to "approved drug" is steeper still, and most of those failures are never counted at all, because they happen before anyone has to announce anything.

Read that again, because it reframes everything. An impressive animal result is the normal starting condition of a compound that will not work in humans. It is also the normal starting condition of a compound that will. The animal result does not discriminate. That is not a criticism of animal research; it is a description of what animal research is for. It generates candidates. Humans adjudicate them.

BPC-157 has generated a candidate. Nobody has adjudicated it.

One rule for this chapter specifically

This is the highest-risk chapter in this book, so the standing rule from Chapter 1 §1.9 gets stated again at full strength. There are no doses here, no protocols, no preparation or reconstitution instructions, no administration technique, no cycles, no sources, and no vendors. Not because the information is secret, but because a chapter containing it would function as an instruction sheet regardless of what its prose said, and I am not writing one of those about a compound with no human safety data.

If you are using this compound and hoped this chapter would tell you whether you are doing it right, I cannot, and neither can anyone who has never examined you. What it can give you is more durable: an accurate picture of what is known, what is not, and how to tell the difference yourself. §17.9 has the chapter's one actionable sentence.

17.2 The molecule: fifteen residues and a gastric-juice origin story

Start with the thing itself, because a surprising amount is settled before any biology.

BPC-157 is a 15-amino-acid peptide — a pentadecapeptide — with the sequence GEPPPGKPADDAGLV and a molecular weight of about 1,419 daltons. It is described as derived from a sequence identified in human gastric juice, which is where the name comes from: BPC stands for "Body Protection Compound," and the 157 is a laboratory series number.

Place it on Chapter 1's spectrum and it sits comfortably in peptide territory — larger than a small molecule by roughly an order of magnitude, smaller than insulin by a factor of four, far too big to behave like a drug you swallow. That last point becomes §17.8.

Now read the sequence structurally, using nothing but Chapter 1 §1.2.

🧬 The Molecule — what fifteen letters tell you

text Gly-Glu-Pro-Pro-Pro-Gly-Lys-Pro-Ala-Asp-Asp-Ala-Gly-Leu-Val G E P P P G K P A D D A G L V 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Four prolines, at positions 3, 4, 5, and 8 — including a run of three. Proline's side chain loops back onto the backbone nitrogen, which locks that stretch of chain and removes it from the hydrogen-bonding pattern an alpha helix requires. This molecule cannot form a normal alpha helix through that region. It is structurally rigid there, and rigidity is chemically consequential: proline-rich stretches are also poor substrates for many endopeptidases, which is the seed of a real argument we will meet in §17.8.

Three glycines, at 1, 6, and 13. Glycine's R group is a single hydrogen — the smallest possible — which permits backbone angles no other residue allows. Glycines are hinges. A short peptide with three of them interleaved between rigid proline blocks is a molecule of alternating stiffness and flexibility, not a uniform rod and not a floppy string.

Two aspartates and one glutamate, and no arginine or histidine. At physiological pH the molecule carries net negative charge. Charged and polar molecules do not diffuse across lipid membranes; this is one reason a peptide of this size does not simply wander into cells or across an epithelium.

One lysine, at position 7. In drug development, lysine is the standard chemical handle — the place you attach a fatty acid to extend half-life, as semaglutide does (Chapter 33). Nobody has commercially developed BPC-157 that way, which is itself a small fact about who has been working on this molecule and with what resources.

All of that is real information, obtained from fifteen letters and no biology whatsoever. It tells you what kind of object this is. It tells you nothing about whether it heals anything. Keeping those two registers separate is most of the skill this book teaches.

What the name tells you, and what it does not

BPC-157 has never received a generic drug name. It is known only by a laboratory code.

Chapter 1 §1.8 explained the naming system: international authorities assign a stem-based generic name when a compound enters serious clinical development. Semaglutide got a name. Tesamorelin got a name. A molecule still known by a laboratory code decades after its discovery has, with near certainty, never completed formal drug development.

Now hold the distinction precisely, because this is where careful people and careless people separate:

The absence of a generic name is strong evidence about regulatory history. It is no evidence at all about pharmacology.

A naming committee is not a laboratory. It does not test molecules. Its assignment reflects a sponsor's filings and a development program's existence, not a compound's biological activity. Read in the correct direction, the missing name tells you: no sponsor has carried this through the process that produces a name. Read in the wrong direction — "it has no real name, so it isn't a real drug, so it doesn't do anything" — it is a non sequitur, and one that will make you look foolish to anyone who has read the animal work.

The same discipline applies in reverse, and this one catches more people. Having a name is likewise no guarantee of efficacy for any particular claim. Plenty of named, approved drugs are ❌ for the specific things people take them for off-label.

The receptor problem

Here is the fact that most changes how you should weigh mechanistic arguments about this compound.

No receptor has been definitively established for BPC-157.

Proposed mechanisms exist and are taken seriously by the researchers who proposed them. They involve nitric-oxide-related signaling pathways, growth-factor and angiogenic signaling, effects on vasculature and on the expression of various receptors and mediators, and several others. But every one of these is inferred from downstream observations in animal models rather than derived from an identified receptor. The logic runs: we administered the peptide, we measured changes in these markers or these tissues, therefore the peptide is acting through pathways involving them.

That is legitimate science. It is also a weaker form of mechanistic knowledge than what we have for almost every other compound in this book. For semaglutide we can name the receptor, describe its structure, show the binding, and knock it out to abolish the effect. For BPC-157 the mechanism section of a paper is a description of correlated downstream changes and a hypothesis connecting them.

Two practical consequences.

For your dossier's Field 3 (mechanism), the honest entry is "not established." Not "acts on nitric oxide pathways" — that is a hypothesis with supporting downstream observations. Writing "unknown" where it is unknown is a discipline, and Chapter 5 argued that it is the discipline that most distinguishes people who can evaluate evidence from people who can only repeat it.

And it caps how much mechanistic argument is worth here. Chapter 5's rule 3 — never upgrade a rating with mechanism — applies to every compound. It applies with extra force to a compound whose mechanism is itself inferred. When someone explains at length why BPC-157 must work, notice that the explanation is built on an unidentified receptor, and that even a perfectly identified receptor would not have moved the rating.

🔍 Check Your Understanding

1. BPC-157 has never been assigned a generic drug name. State precisely what that is evidence of, and precisely what it is not evidence of.
2. Someone argues that BPC-157 must work because it acts on nitric-oxide pathways central to healing. Name two separate problems with that argument — one about the mechanism itself, one about the logic of using mechanism at all.
3. From the sequence alone, why would you predict this molecule does not cross cell membranes easily?

17.3 The animal literature, taken seriously

This section presents the case for BPC-157 at its strongest. That is not generosity; it is method. A claim you have only encountered in its weakest form is a claim you have not evaluated.

The literature spans four broad model families, and its most striking feature is that the reported direction of effect is consistent across all four.

THE FOUR MODEL FAMILIES IN THE BPC-157 ANIMAL LITERATURE

  TENDON / LIGAMENT / MUSCLE          GASTROINTESTINAL
  ─────────────────────────           ─────────────────────────
  Typical model: surgical             Typical model: chemically
  transection of Achilles or          induced ulceration or
  quadriceps tendon, or ligament      colitis; damage from
  transection, or muscle crush,       NSAIDs, alcohol, or
  in rats                             corrosive agents, in rats
                                      and mice
  Typical outcomes: histological
  healing scores, biomechanical       Typical outcomes: lesion
  load-to-failure, tendon             area and count, mucosal
  thickness, functional walking       integrity scores, survival
  or gait measures, time to           in severe models, markers
  recovery                            of inflammation

  VASCULAR / ANGIOGENIC               NEURAL
  ─────────────────────────           ─────────────────────────
  Typical model: vessel               Typical model: peripheral
  ligation or occlusion,              nerve crush or transection;
  ischemia–reperfusion,               models of CNS insult,
  induced hypertension or             traumatic injury, or
  thrombosis in rodents               neurotoxic exposure

  Typical outcomes: collateral        Typical outcomes: functional
  vessel formation, blood flow        recovery scores, gait and
  restoration, endothelial and        reflex testing, histology of
  angiogenic markers, thrombus        nerve or brain tissue
  formation and bleeding time

  ACROSS ALL FOUR: the reported direction is consistently toward faster or more
  complete recovery in treated animals versus controls. Multiple administration
  routes have been tested in animals. More than one research group has published.

Take each in turn.

Tendon, ligament, and muscle. The best-known work uses surgical transection models: an animal's Achilles or quadriceps tendon is cut, the injury is standardized, and healing is compared between treated and control animals. Reported outcomes include improved histological organization of the healing tissue, greater biomechanical strength at defined timepoints, and faster return of function on gait or walking measures. Related work reports effects in ligament transection and in muscle crush and transection models. These are the studies that generated the "healing peptide" reputation, and they are the reason a serious athlete's training partner is not making things up when he says he read a study.

Gastrointestinal protection. This is the oldest strand and, on its own terms, the most coherent — which makes sense given the compound's stated origin in gastric juice. The models induce damage to the stomach or intestine with a known agent: nonsteroidal anti-inflammatory drugs, alcohol, corrosive substances, or agents that produce a colitis resembling inflammatory bowel disease. Treated animals are reported to develop fewer or smaller lesions, to show better preserved mucosal integrity, and in severe models to survive at higher rates. Some of this work also reports protection against damage from agents affecting other organs.

Vascular and angiogenic. Studies examining blood flow after a vessel is occluded or ligated report accelerated development of collateral circulation — alternative routes around the blockage — and faster restoration of perfusion, along with effects on thrombus formation and on markers associated with new vessel growth. The researchers involved have argued that this vascular effect is the unifying mechanism behind the whole picture: if a compound reliably improves perfusion to injured tissue, then improved healing across many tissues is not four findings but one.

Neural. The narrowest strand: peripheral nerve injury models, models of central nervous system insult, and models of neurotoxic exposure, with reported improvements in functional recovery scores and in tissue histology.

🔬 Read the Study

```text FIGURE 17.1 — "Gastroprotection in a rodent NSAID-damage model" [composite — real pattern, constructed specifics]

THE STUDY Rats received a nonsteroidal anti-inflammatory drug at a dose reliably producing gastric mucosal lesions. Animals were randomized to a vehicle control arm or to BPC-157 arms at several dose levels, given by injection. At a fixed interval, stomachs were examined and lesion area was measured by an assessor blinded to group assignment. Group sizes were in the range typical of this literature — several animals per arm.

THE QUESTION Does BPC-157 reduce NSAID-induced gastric mucosal damage in rats?

WHAT IT SHOWS Substantially smaller total lesion area in treated animals than in vehicle controls, with the effect present across the dose range tested rather than only at one level. Blinded assessment of the primary outcome. A clean, plausible, internally valid experiment with an unambiguous result in the model as built.

WHAT IT DOESN'T That the same thing happens in a human. Nothing about a person taking an NSAID long-term for arthritis, whose mucosal damage develops over months against a background of age, comorbidity, other medications, and Helicobacter status — none of which exist in this experiment. Nothing about symptoms, because a rat cannot report dyspepsia. Nothing about the outcome anyone actually cares about, which is bleeding events and hospitalizations, not lesion area on a scoring sheet. Nothing about safety at any exposure over any duration. And nothing about what happens when the compound is swallowed rather than injected.

THE VERDICT A well-conducted animal experiment supporting a hypothesis. It sits on the ANIMAL rung of Chapter 5's ladder, and a hundred more like it would sit on the same rung, because rungs are not reached by accumulation.

THE LESSON Notice how much this study earns and how precisely it stops. That is the correct relationship between a good experiment and its conclusion — and it is exactly the relationship that gets lost when the finding is restated by someone selling something. Watch for the sentence that drops the word "rats." ```

What "a substantial literature" means, and why the count is not the point

You will see claims about how many BPC-157 studies exist. The numbers vary wildly, and I am not going to give you one, because I would be adding a fabricated precision to a question that does not have a determinate answer.

There are several honest reasons the count is unstable. Papers on the same underlying experiments get recycled into reviews that are then counted as studies. Conference abstracts get counted alongside full papers. Work in overlapping models from the same group gets counted separately or together depending on who is doing the counting. And, critically, animal research is largely unregistered — there is no comprehensive registry of preclinical experiments the way there is for clinical trials — so nobody knows the denominator. Any count is a count of published studies, and §17.5 explains why that is a very different thing.

So: a substantial literature. More than a handful, less than a field. Enough that dismissing it as "a few rat studies" is false, and enough that citing its size as evidence is a category error. Size is not a rung.

17.4 What is genuinely impressive about it

If I were reviewing this file as a preclinical program director, several things would hold my attention. I want to name them explicitly, because a chapter that arrives at ❌ without first naming what is good about the case has not actually made an argument.

The consistency across tissue types is unusual. Most compounds that look good in one model look mediocre in the next. A reported effect in tendon, in gut mucosa, in vasculature, and in nerve is a different kind of signal — either something quite general is happening, or something quite general is wrong with how the work is being done. Both possibilities are live, and §17.5 takes the second seriously. But the first is not silly. Wound healing across tissue types genuinely does share machinery: perfusion, angiogenesis, growth-factor signaling, collagen deposition, resolution of inflammation. A compound acting upstream of those could plausibly produce exactly this pattern.

The proposed unifying mechanism is coherent. "It improves the blood supply to damaged tissue" is not a hand-wave. Perfusion really is rate-limiting for healing in many tissues, tendon notably among them because tendon is poorly vascularized to begin with, and that poor vascularity is a standard explanation for why tendons heal slowly. A compound that improved perfusion in a hypovascular tissue would be attacking a genuine bottleneck. The story hangs together, which is more than can be said for a great deal of what is sold as a peptide.

Effects have been reported across a wide range of doses. In the animal work, effects are not reported to be confined to a narrow window, and the literature does not describe a steep dose-dependent toxicity. A compound with an apparently wide margin between an active dose and a harmful one is, all else equal, a more attractive development candidate than one with a narrow one. (This is a statement about what the animal literature reports, not about human safety, which does not exist. §17.9.)

Activity has been reported by more than one route in animals. That matters for §17.8 and I am flagging it here so that section does not read as though I am hiding it.

And the work is not fringe in form. It appears in peer-reviewed journals. It uses standard, recognized models. It includes blinded outcome assessment and control arms. There is no sleight-of-hand at the level of experimental design. Whatever is wrong with the situation, it is not that the experiments are fake.

I want to say this as plainly as I can, because everything after it is going to be critical:

Nothing in the rest of this chapter is a claim that BPC-157 does not work. The animal data is real. If a pharmaceutical company had this file for a proprietary molecule, they would very plausibly fund a Phase I. That is the honest reading of the preclinical package.

The problem is not the science. The problem is that the next step has not been taken, and the next step is the only one that answers the question.

17.5 The four structural problems with that literature

Chapter 5 §5.3 gave five reasons animal results fail to transfer to humans. All five apply here, and they group into four structural problems. None of them is a criticism of any individual study. Every one of them would apply to a perfectly conducted experiment.

That is the point, and it is the most misunderstood idea in this book. These problems are not solved by doing the animal work better. They are properties of the gap between an animal experiment and a human answer, and the only thing that crosses that gap is a human trial.

Problem 1: Different biology

A rat is not a small human.

Rodents heal faster than humans, and more completely. Their metabolic rate is higher, their proteolytic environment differs, their tendons are loaded differently and are structurally different, their gut microbiota and gastric physiology differ, their immune responses differ, and their lifespan compresses everything. Effects that look large over a two-week rodent recovery may correspond to nothing detectable over a nine-month human one — or to something larger. You cannot tell without looking.

There is a specific version of this problem for a compound with no identified receptor. When you know the target, you can at least check whether the human version of that target resembles the animal one, and whether it is expressed in the same tissues. That check is one of the standard early filters in drug development. For BPC-157 that check cannot be performed, because there is no established receptor to compare. This is not a hypothetical concern; species differences in receptor pharmacology have killed real programs.

Problem 2: Different disease, and different endpoint

These two reasons from Chapter 5 belong together here because they compound each other.

A model captures a piece of a disease, not the disease. A chemically induced rodent ulcer is not peptic ulcer disease. An induced colitis is not Crohn's. A transected tendon is not tendinopathy — that mismatch is important enough to get its own section (§17.7). Every model is a deliberate simplification, and the simplification is what makes the experiment possible. The question is always whether the discarded complexity was the part that mattered.

And animals cannot report outcomes. This is the constraint people most consistently underrate. A rat cannot tell you its shoulder hurts less. It cannot tell you it slept better, returned to work, stopped avoiding stairs, or felt able to lift again. So animal endpoints are necessarily proxies: lesion area, histological score, load-to-failure, gait parameters, marker concentrations.

Every one of those proxies is a surrogate, and Chapter 5 was blunt about surrogates: they are useful when validated against outcomes that matter and misleading when not. A tendon that fails at a higher load in a materials tester is not the same finding as a person who can run again. Sometimes they travel together. Sometimes they do not, and the history of medicine contains a long list of surrogates that improved while patients did not.

Problem 3: Different dose

The dose used in a rodent experiment is not a human dose, and it does not become one by arithmetic.

Body-size scaling between species is not linear — metabolic rate, clearance, and distribution volume all scale with body mass in a way that requires allometric conversion rather than simple multiplication, and the conversions used in early drug development are themselves rough approximations that get revised as soon as real human pharmacokinetic data arrives. Route matters too: an intraperitoneal injection in a rat, a common laboratory route, has no human clinical equivalent, and the exposure profile it produces is not the profile a subcutaneous injection produces.

I am deliberately not giving you the numbers, and I want to be transparent about why rather than leaving you to guess. Any number I printed would be read as a dose, no matter how many disclaimers surrounded it, and there is no human pharmacokinetic study to anchor it against. That absence is the actual finding here: for this compound, nobody knows what human exposure any animal dose corresponds to, because the measurement that would establish it has never been published. Anyone who tells you a human dose for BPC-157 is extrapolating from rodent data using assumptions they cannot check, and if they have not told you that, they either do not know it or have decided you would rather not hear it.

Problem 4: Different publication pressure — the filtered literature

This is the least-discussed of Chapter 5's five reasons and, for this compound, the most important.

Clinical trials in humans are, in most jurisdictions, required to be registered before they begin. That registration creates a public record of the intention to run a study, so a trial that produces an unwanted result and is quietly never published leaves a visible gap. The system is imperfect, but the accounting exists.

Animal research has no equivalent. Preclinical experiments are overwhelmingly unregistered, there is no public record of an experiment that was run and produced nothing, and there is very little incentive to publish a null result — such results are hard to place, carry no career reward, and are frequently attributed to technical failure rather than to absence of effect, sometimes correctly.

The consequence is precise, and worth stating slowly:

The published animal literature on any compound is a filtered sample from a population of unknown size. You can see the studies that were published. You cannot see how many were run. You cannot compute the ratio. Therefore you cannot interpret the consistency of the published results, because consistency is exactly what a filter produces.

This is why "there are so many studies and they all point the same way" is not the argument it appears to be. In a literature with no denominator, the number of concordant published studies is only weak evidence about the underlying effect. It is compatible with a real, robust effect. It is also compatible with a small or absent effect plus ordinary selective publication. The published record cannot distinguish these two worlds, and no amount of additional publication in the same system can, which is the deepest reason more animal studies will never resolve this.

A related and more delicate point, which I want to state carefully because it is a structural observation and not an accusation. When a literature is concentrated in a small number of research groups, shared methods, shared assumptions, and shared laboratory practices are shared across the whole body of work. That is not misconduct; it is what happens when a small number of people find a question interesting. But it does mean that independent replication by unaffiliated laboratories carries unusually high information value here — higher than another study from a group that has already published many. Independent replication is the specific thing this literature would benefit from most, and it is a normal scientific request, not an insinuation about anyone.

⚠️ Hype Check — "there are hundreds of studies on BPC-157"

The claim, in its usual form:

"People say it's unproven, but there are hundreds of peer-reviewed studies. It's been researched for over thirty years. That's more evidence than most supplements have."

What's true in it. More than most supplements have — genuinely yes. The research history is long, the literature is real and peer-reviewed, and someone dismissing it as "no evidence" is wrong. If the comparison class is a bottle of something with a proprietary blend and a testimonial, BPC-157's file is in a different league.

Where it fails. Three places.

First, the count is not verifiable in the way the claim implies. There is no registry of preclinical work, reviews get counted as studies, and overlapping reports from the same experimental program get counted separately. Anyone quoting a precise number is quoting a number whose construction they cannot show you.

Second, evidence does not accumulate across rungs. This is Chapter 5's central architectural point and it is the one most often lost. Studies on the animal rung add confidence within that rung — they make the animal finding more secure — and they do not by accumulation produce a human finding. A thousand rat studies and one rat study make the same claim about humans, which is none. This is not a rhetorical trick; it is why regulators require human trials from compounds with immaculate preclinical packages.

Third, the comparison class is chosen to flatter. "More evidence than most supplements" is true and irrelevant. The relevant comparison is not to a supplement; it is to the other things you might do about a tendon injury, several of which have randomized human evidence behind them.

Verdict: the claim substitutes volume for level. The correct response is not "there aren't hundreds of studies" — you cannot verify that either — but "how many of them are in humans?" That question has a determinate answer, you can check it yourself in ten minutes, and §17.6 shows you how.

🔍 Check Your Understanding

1. Which of the four problems in this section would be fixed by running a larger, better-controlled animal study? Explain your answer.
2. Why does the absence of preclinical trial registration make the consistency of published animal results hard to interpret?
3. A rat study reports faster healing at a stated dose per kilogram. Give two distinct reasons that number cannot be converted into a human dose, and one reason specific to BPC-157.

17.6 The human literature: what exists, what doesn't, and how to check for yourself

The central fact of this chapter takes one sentence.

As of this writing in 2026, there is no completed, peer-reviewed, randomized controlled human trial of BPC-157 for any indication in the published literature.

Not for tendon injury. Not for gastrointestinal protection. Not for anything. There is also no published Phase I safety study — no dose-escalation trial establishing what happens to human beings given this compound under observation, which is normally the very first thing anyone does.

I have written that without adornment on purpose. It is not a gotcha, it does not require a triumphant tone, and it is not the end of the discussion. It is a fact about the state of the record, and it is the fact from which everything else in this chapter follows.

What people mistake for human evidence

The reason many intelligent readers arrive believing human trials exist is that several things resemble them. Learn to tell them apart and you will have acquired a skill that transfers to every compound in Part III.

A trial registration is a plan, not a result. Registries list studies that are planned, recruiting, active, suspended, terminated, withdrawn, or completed. A registration is a document filed before anything happens. It tells you someone intended to run a study. It tells you nothing about whether the study ran, finished, found anything, or was ever written up. "There's a trial on ClinicalTrials.gov" is compatible with there being no data whatsoever.

"Completed" in a registry is not "published." A registry entry can be marked completed while no peer-reviewed paper exists and no results are posted. Completion means recruitment and follow-up ended. It does not mean anyone can read what happened.

A conference abstract is not a peer-reviewed paper. Abstracts are typically a few hundred words, reviewed lightly if at all, and frequently never followed by a full publication. A substantial fraction of results presented at conferences never appear in full, and abstracts that do get published often differ from what was presented. An abstract is a promissory note.

A case report is a story with an n of 1. Someone took a compound, something improved, a clinician wrote it up. Case reports are genuinely valuable for one thing — flagging unexpected, dramatic, or harmful events worth investigating — and they are close to worthless for establishing efficacy, because they have no control, no blinding, no randomization, and no denominator. You do not know how many people took the same thing and did not improve, and neither does the author.

A case series is several of those stacked, which multiplies the story and does nothing about any of the four missing elements.

And a narrative review can read exactly like a body of human evidence. This is the most effective of the lot, and I want to name the mechanism precisely, because once you see it you cannot unsee it. A review paper summarizes dozens of animal studies. Its abstract, compressing for space, says something like "BPC-157 has demonstrated efficacy in models of tendon, gut, and vascular injury." A second source cites the review and writes "BPC-157 has demonstrated efficacy in tendon, gut, and vascular injury." A third cites the second: "BPC-157 is effective for tendon injury." The word "models" fell out between steps one and two, and by step three the sentence describes humans. Nobody lied. Each step was a small compression. The species disappeared through paraphrase, and the final sentence is now circulating with three citations behind it.

Chapter 6 called this citation drift, and it is the most common way a preclinical literature turns into a marketing claim without anyone committing fraud.

On earlier development history: you will encounter references to BPC-157 having been explored clinically for gastrointestinal indications under a development code. Rather than characterize that history from memory, I want you to do something better, which is check the primary sources yourself and see what returns. The standard is not "did someone once plan a trial" — it is "is there a completed, published, randomized result you can read." Apply that standard and the answer is determinate.

How to check this yourself, in about ten minutes

This is the most valuable thing in this chapter. Not the rating — the rating is mine, it is date-stamped, and it may be out of date by the time you read this. The procedure is yours, it works on any compound, and it will still work when this book is old.

THE TEN-MINUTE CHECK — any compound, any claim

STEP 1 — PubMed (pubmed.ncbi.nlm.nih.gov)
  a. Search the compound name. Note the total.
  b. Now apply the ARTICLE TYPE filter: "Randomized Controlled Trial".
  c. Now apply the SPECIES filter: "Humans".
  d. Look at what survives both filters. This is the number that matters.
  e. Repeat with alternate names and laboratory codes — compounds hide under
     several identifiers, and a search on one name misses the others.

STEP 2 — READ THE METHODS OF WHATEVER SURVIVED
  Filters are imperfect and indexing has errors. Open each surviving paper and
  find the sentence naming the subjects. If it says rats, mice, rabbits, or
  "in vitro," it is not a human trial regardless of how it was indexed.

STEP 3 — ClinicalTrials.gov
  a. Search the compound. Read the STATUS field on every hit:
     Not yet recruiting / Recruiting / Active, not recruiting / Suspended /
     Terminated / Withdrawn / Completed / Unknown status
  b. "Unknown status" means the sponsor stopped updating. It is common and it
     usually means what you think it means.
  c. For anything marked Completed, look for a "Study Results" tab. Registered
     ≠ run ≠ completed ≠ results posted ≠ peer-reviewed publication. Those are
     five different achievements and only the last one settles anything.

STEP 4 — WIDEN THE NET
  ClinicalTrials.gov is US-centered. Check the WHO International Clinical Trials
  Registry Platform (ICTRP), which aggregates national registries, plus the
  EU Clinical Trials Register. A trial run elsewhere would appear there.

STEP 5 — DATE AND RECORD WHAT YOU FOUND
  Write down the date, the searches you ran, and the counts. This is now a
  finding with a timestamp, and it is checkable — by you, later, when someone
  tells you the situation has changed.

Do this. Genuinely, put the book down and do it. It takes less time than reading the rest of this section, and it converts everything here from something you are being told into something you have verified. If your result differs from what this chapter says, your result is the current one and mine is the historical one — that is what date-stamping a rating is for, and §17.11 explains exactly what a changed answer would mean.

📊 Evidence Rating

Claim: BPC-157 protects the human gastrointestinal tract against damage from NSAIDs, alcohol, or inflammatory disease.

Rating: ❌ Hype outpaces evidence. (as of 2026)

Why: The gastrointestinal work is the oldest and arguably most internally coherent strand of the animal literature, and the origin story gives it a natural plausibility. It remains animal work. There is no completed, peer-reviewed, randomized controlled human trial for this endpoint, and no published human safety data to support use for a chronic condition.

What would change it: a randomized, placebo-controlled human trial in a defined population — for example, patients requiring long-term NSAID therapy — with a pre-specified clinical endpoint such as endoscopically confirmed ulceration or clinically significant bleeding events, rather than a symptom score alone. Existing proton pump inhibitors are an effective, cheap, well-characterized comparator with decades of human evidence, so a serious trial here would need an active-control arm to be worth running.

Note — this is the same compound rated in §17.10, on a different claim. Chapter 5's rule 6: one molecule, many ratings. The ratings happen to match here, and the reasoning that produced them is not identical — this one carries the additional problem that a well-established alternative already exists, which changes what a trial would have to show.

17.7 The transected tendon and the worn tendon: arguably not the same disease

This section is the most practically important in the chapter for anyone thinking about a soft-tissue injury, and its point is easy to state and easy to miss.

The animal studies that built this compound's reputation cut a tendon. The human problem most people are trying to solve did not involve anything being cut.

TWO INJURIES THAT SHARE A WORD AND LITTLE ELSE

  THE RODENT MODEL                      THE HUMAN PROBLEM
  ────────────────────────────          ────────────────────────────
  Mechanism:  surgical transection      Mechanism:  cumulative overload
              — a clean, complete cut               over months or years
  Onset:      instantaneous;            Onset:      gradual; no identifiable
              time zero is KNOWN                    time zero
  Tissue:     healthy tendon, cut       Tissue:     already degenerated before
                                                    symptoms began
  State:      acute; inflammatory       State:      chronic; classic inflammation
              cascade fully engaged                 often minimal or absent
  Structure:  complete discontinuity    Structure:  partial, disorganized
                                                    collagen; matrix changes
  Vessels:    normal architecture,      Vessels:    aberrant neovascularization
              disrupted by the cut                  often ALREADY present
  Subject:    young, healthy, uniform   Subject:    any age, any comorbidity,
              inbred animals                        highly variable
  Loading:    controlled cage           Loading:    the person keeps using it,
              activity                              often for the thing that
                                                    caused it
  Timescale:  days to a few weeks       Timescale:  months to years

Look at that table as a whole rather than row by row. Almost nothing corresponds.

The clinical literature on human tendon problems moved, over the last few decades, away from calling them "tendinitis" precisely because the "-itis" was misleading: chronic tendon disorders frequently show relatively little classic inflammation and considerably more degenerative change — disorganized collagen, altered matrix, changes in cell populations, and often abnormal ingrowth of new vessels and nerves. The preferred term became tendinopathy, and the change of name reflected a genuine change in understanding of what the disease is.

If that is right, then a compound that accelerates the repair of an acute, clean, complete rupture is being asked to do something quite different from what a person with chronic tendinopathy needs. The transection model is a test of repair machinery under maximal activation. Chronic tendinopathy may be closer to a failure of remodeling in tissue where the repair response never properly engaged, or engaged and stalled, or has been repeatedly re-injured before it could finish.

These are arguably different pathologies, not the same pathology at different scales. I say "arguably" because tendon biology is genuinely contested and I am not going to overstate a position in a live scientific argument. But the burden falls the right way: someone claiming a transection result predicts a tendinopathy result owes an argument for why, and "they're both tendon injuries" is not one.

There is a sharper version of the worry, and it deserves stating even though it is speculative, clearly flagged as such. If the compound's proposed mechanism runs substantially through angiogenesis — promoting new blood vessel growth — and if chronic tendinopathy already features aberrant neovascularization as part of its pathology, then the direction of benefit is not obvious in advance. Some clinical approaches to tendinopathy have specifically targeted those abnormal vessels. I am not asserting harm; there is no evidence of harm and no evidence of benefit, which is the whole problem. I am pointing out that a mechanism that sounds unambiguously good in the abstract can have an ambiguous sign once you specify the actual disease, and that this is exactly the sort of thing a trial exists to resolve.

The confounder that makes personal experience nearly uninterpretable

Now the practical part, and I want to handle it carefully because a lot of readers have a personal experience they are weighing against everything in this chapter.

Suppose someone with a six-month shoulder problem starts using BPC-157. Suppose their shoulder gets better over the following two months. What can they conclude?

Almost nothing, and here is why. Consider what else typically changes at the same moment. The person has usually just decided to take the injury seriously — which means they modified their training, stopped the aggravating movement, added rehabilitation exercise, slept more, paid attention. Chronic tendinopathy also has a substantial natural history: many cases improve over months with load management alone, which is why progressive loading programs supervised by a clinician are the best-supported treatment in the human literature. And symptoms fluctuate. People start interventions when they feel worst, which is precisely the condition under which regression to the mean guarantees that the average next measurement is better regardless of what was done.

So the personal experience is real — the shoulder genuinely improved — and it is consistent with the compound having a large effect, a small effect, or none at all. This is not a slight on anyone's observational powers. It is why blinded, controlled, randomized designs were invented: not because individuals are stupid, but because this particular inference is unavailable to any individual, no matter how careful, including me.

💊 In the Clinic — what a clinician is actually weighing

A patient mentions they have been using BPC-157 for a stubborn Achilles problem. What is the clinician thinking?

Not, primarily, "is this compound effective?" They know the honest answer is unknown, and the question in front of them is different: what is this person's tendon doing, and what is the plan?

The best-supported treatment for chronic mid-portion Achilles tendinopathy in the human literature is a structured, progressive loading program, delivered over months, supervised well enough that load is actually progressed. It is unglamorous, it requires sustained effort, its early phase can be uncomfortable, and adherence is the main reason it fails. Nothing in this book is a substitute for it, and a compound with no human trials is certainly not.

The clinician's realistic concerns are: whether an unproven addition is displacing the loading program or delaying appropriate imaging; whether the substance came from a source with any quality assurance at all (§17.9); whether there is anything in the history — a malignancy, proliferative retinopathy, an anticoagulant — that makes an uncharacterized compound with proposed vascular effects a bad idea; and whether the patient will keep telling them the truth.

That last one is the whole ballgame, and it is why the tone of this chapter is what it is. A clinician who reacts to the disclosure with contempt has bought a moment of moral satisfaction and lost the information stream. The patient does not stop using the compound. They stop mentioning it.

17.8 Stability, route, and the oral BPC-157 claim

Oral and sublingual BPC-157 products exist and are widely sold. This section evaluates that claim using Chapter 4 and no clinical data at all, which is possible because the chemistry is decisive before evidence is consulted.

A swallowed peptide faces five barriers.

THE FIVE BARRIERS BETWEEN A SWALLOWED PEPTIDE AND YOUR BLOODSTREAM

  1. STOMACH        Hydrochloric acid at roughly pH 1–3, plus PEPSIN — a protease
                    whose entire evolved job is cleaving peptide bonds.

  2. SMALL INTESTINE  Pancreatic proteases: trypsin, chymotrypsin, elastase,
                    carboxypeptidases. Different cleavage preferences, so between
                    them they cover most sequences.

  3. BRUSH BORDER   The intestinal lining carries its own peptidases on its
                    surface — the final stage of protein digestion, designed to
                    reduce short peptides to absorbable amino acids and dipeptides.

  4. THE EPITHELIUM Even an intact peptide must now cross a tight cell layer.
                    At ~1,419 Da and carrying net negative charge, this molecule
                    is too large and too polar for passive diffusion, and there is
                    no established transporter carrying it.

  5. THE LIVER      Anything absorbed from the gut enters the portal vein and
                    passes through the liver BEFORE reaching systemic circulation.
                    First-pass metabolism removes a further fraction.

  EACH BARRIER IS INDEPENDENTLY SUFFICIENT to defeat an oral peptide. They are
  not additive obstacles to be overcome collectively; any one of them alone can
  reduce systemic delivery to effectively zero.

The strongest version of the oral argument

Now the counterargument, at full strength, because it is not stupid.

BPC-157's proline-rich structure has been argued to confer unusual stability. This is a real chemical argument with a real basis. Proline's constrained backbone geometry does make proline-rich sequences poor substrates for many endopeptidases; there are naturally occurring proline-rich peptides notable for their resistance to degradation, and the phenomenon is well recognized in peptide chemistry. A molecule with four prolines including a run of three in a fifteen-residue chain is genuinely an unusual object, and it would not be surprising if it survived proteolytic environments that destroy an ordinary peptide. Some of the animal work reports activity following oral administration, which is not nothing.

Take all of that seriously. Then notice the gap:

Stability is not bioavailability. They are different claims, measured differently, and one does not imply the other.

Stability answers: does the molecule remain chemically intact in this environment? Bioavailability answers: what fraction of an administered dose reaches systemic circulation unchanged? A molecule can be perfectly stable and have zero bioavailability, and BPC-157's own structure shows how — barriers 1 through 3 are degradation barriers, and stability addresses them. Barrier 4 is a permeability barrier, and stability does nothing about it whatsoever. An indestructible molecule that cannot cross the intestinal epithelium is an indestructible molecule that passes through you and is excreted.

This distinction is the single most useful thing in this section, because the substitution of one claim for the other is nearly universal in marketing copy. "Stable in gastric juice" is presented as though it answered the delivery question. It answers one fifth of it.

Two related points, both general chemistry rather than product commentary. Alternative salt forms can genuinely change solubility and solid-state stability, but not the molecule's size or its permeability across an epithelium — the peptide dissociates from its counter-ion in solution, and it is the peptide that has to cross. And sublingual or buccal routes bypass barriers 1, 2, 3, and 5, which is the reason those routes work for a handful of small, lipophilic drugs. But they replace barrier 4 with a different epithelium of comparable difficulty, over a small surface area, with limited contact time, and most of the dose gets swallowed anyway. Sublingual delivery of a 1,419-dalton charged peptide is a hard problem, not a solved one.

What a seller owes you

Here is the standard, and it is the same one Chapter 4 applied to every oral peptide claim:

Which barrier did you solve, how did you solve it, and what is the measured bioavailability in humans?

Three parts, and all three are required. There are real answers to this question in the world. Oral semaglutide exists and works — it required a specific absorption-enhancing excipient, a strict administration procedure, and it achieves roughly 1% bioavailability, with the tablet containing far more drug than the injection precisely because most of it is expected not to arrive. That is what a solved oral peptide problem looks like: a named mechanism, a measured number, and a formulation built around the number.

For an oral BPC-157 product, ask for the equivalent. In practice you will usually get a stability argument, which answers a different question, or an animal citation, which brings us to the last point.

Rodent oral administration does not establish human oral bioavailability, for all four reasons in §17.5 plus one specific to this route: rodent gastric pH, transit time, and intestinal enzymology differ from ours in ways that matter enormously for a molecule whose survival is the question.

And there is a subtler issue that applies even if the rodent oral data is taken entirely at face value. A peptide acting locally in the gut is not evidence that it reaches a tendon. If a compound protects gastric mucosa after oral dosing, the most parsimonious explanation is that it acted where it was — on the tissue it was sitting against — not that it was absorbed, survived first-pass metabolism, distributed systemically, and arrived at a distant hypovascular structure in a meaningful concentration. Those are entirely different pharmacological claims, and the first is much easier to satisfy than the second.

⚠️ Hype Check — "the oral version works because it's stable in the gut"

The claim, in its usual form:

"You don't need to inject it. BPC-157 is stable in gastric juice — that's literally where it comes from — so the oral capsules work fine. Studies show oral administration is effective."

What's true in it. The stability argument has a genuine chemical basis: proline-rich sequences really are resistant to many proteases, and this molecule is unusually proline-rich. Some animal work really does report activity after oral administration. Neither point is invented.

Where it fails. Three places.

First, "it comes from gastric juice, so it survives gastric juice" is not an argument. Where a sequence was originally identified says nothing about the stability of a synthetic peptide administered in a capsule. Many things are found in gastric juice, including partially digested proteins, which are found there precisely because they are being destroyed.

Second, stability answers barriers 1–3 and says nothing about barrier 4. Permeability is a separate physical property. This is the load-bearing error and it survives because most readers have never been given the distinction.

Third, "studies show oral administration is effective" is doing the species substitution from §17.6. The studies are in rodents, and gut-local action in a rodent does not establish systemic delivery in a human.

Verdict: the claim conflates stability with bioavailability, and rodent local effect with human systemic delivery. The question that settles it is not rhetorical — it is a request for a measured human bioavailability figure and the method used to obtain it. If it exists, it should be easy to produce. If it does not, the honest description of an oral product is that its delivery has never been characterized in humans.

🔍 Check Your Understanding

1. A product claims lab testing shows its peptide is 90% intact after two hours in simulated gastric fluid. Which of the five barriers does that address, and which does it leave completely untouched?
2. Why is "it protected the stomachs of rats given it by mouth" not evidence that a swallowed capsule delivers the compound to a tendon?
3. Oral semaglutide achieves about 1% bioavailability and is still a successful product. Explain how both halves of that sentence can be true, and what it implies about the standard of proof for other oral peptide claims.

17.9 Safety: "no reported harm" and "shown to be safe" are different sentences

BPC-157 has a reputation for being remarkably well tolerated. That reputation is not based on human safety data, because there is no published human safety data.

There is no completed, published Phase I trial of BPC-157. No dose-escalation study, no human pharmacokinetics, no systematic collection of adverse events under observation, no laboratory monitoring panel, no cardiac safety assessment, no defined denominator of exposed subjects.

Where, then, does the reputation come from? Two sources. The animal literature, which does not describe dramatic toxicity in the models used — a genuine observation, and one of the reasons a development program would be plausible. And the absence of reports of harm from human use.

That second one requires taking apart carefully, because it is the most load-bearing and least examined claim in the entire consumer peptide space.

Why absence of reports is close to uninformative here

"No reported harm" is a statement about a reporting system. When there is no reporting system, it is a statement about nothing.

Consider what would have to happen for a serious adverse event in a person using gray-market BPC-157 to become a "report."

The person would have to attribute the event to the compound — but symptoms are nonspecific, and attribution to an unlabeled substance is the last thing most people think of. They would have to disclose the use to a clinician — but the compound is typically not prescribed, is sometimes purchased under a "not for human consumption" label, and disclosure invites exactly the reaction §17.7 warned about. The clinician would have to recognize the association and know that a reporting pathway exists — but pharmacovigilance systems are built around approved drugs with labels and marketing authorization holders, and an unapproved research chemical fits awkwardly into every field on the form. And someone would have to aggregate those reports and analyze them.

None of that infrastructure exists for this compound. So the absence of reports is not the finding it appears to be. Compare: for an approved drug, "no signal in postmarketing surveillance across two million patient-years" is a genuinely informative statement, because there is a surveillance system, a denominator, and a legal obligation to report. For BPC-157, "no reports" means there is nowhere for a report to go.

There is a further problem that is specific and underappreciated. We do not know the denominator. Even if reports existed, we would not know how many people have used the compound, at what exposures, for how long, or in what state of health, because nobody is counting. A safety statement without a denominator is not a rate; it is an anecdote about anecdotes.

The specific open questions

Being precise about what is unknown is more useful than gesturing at "unknown risks," so here are the particular questions a trial would be designed to answer.

Anything with proposed angiogenic activity raises a question about tissue where new vessel growth is not wanted. The obvious cases are occult malignancy — a tumor too small to have been detected — and proliferative retinopathy. I want to state this with exactly the right force, so let me say it in both directions in a single breath: there is no evidence that BPC-157 causes or accelerates cancer in humans, and there is no evidence that it does not, because the studies that would look have not been done. Someone who tells you it is carcinogenic is fabricating. Someone who tells you the question has been ruled out is also fabricating. This is a live unknown, and it is the kind of thing that appears in a Phase I protocol's exclusion criteria and a longer trial's follow-up plan.

Chronic exposure is entirely uncharacterized. The animal work is largely short-term by nature — rodent studies run days to weeks. People use this compound for months. Chapter 3's lesson applies generally: sustained artificial signaling is a different pharmacological situation from acute exposure, and receptor downregulation, tolerance, and adaptive responses are the normal consequences of chronic stimulation of any pathway. Whether any of that occurs here is unknown.

Immunogenicity and drug interactions have not been assessed in humans. Anything peptide-like can in principle provoke an antibody response, and repeated administration is the circumstance in which that happens. Interactions are unstudied too — including the ones a proposed vascular mechanism would make you want to check first, such as anticoagulants. Nobody has looked at either.

🩺 Safety and Risk — two categories of risk, and they are independent

When people ask "is BPC-157 safe," they are usually asking one question. There are two, and conflating them causes real harm.

Risk category one: the molecule. Summarized above. No human safety data of any kind. Several specific open questions. The honest statement is unknown, and unknown is neither "dangerous" nor "fine."

Risk category two: the product. This risk is entirely independent of the molecule and, on the current evidence about the gray market, it is the one that has actually harmed people. From Chapters 6 and 19:

• Identity — is the peptide in the vial the peptide on the label? A wrong sequence, a truncated sequence, or a different compound entirely are all documented outcomes of unregulated peptide supply.
• Purity — synthesis byproducts, deletion sequences, and residual solvents from manufacturing.
• Potency — the stated quantity versus the actual quantity, which can be anywhere.
• Sterility — anything injected must be sterile. Non-sterile injection causes abscesses and systemic infections, and this is not a theoretical risk.
• Endotoxin — bacterial cell wall fragments survive sterilization, are not removed by filtering, and produce fever and systemic inflammatory responses. Endotoxin testing is a specific assay that unregulated suppliers frequently do not perform.

Notice that every one of these would apply to a vial of sterile saline mislabeled as a peptide. They have nothing to do with BPC-157's pharmacology. This is why "peptides are broken down into amino acids, so they're safe" (Chapter 1) misses the point so completely: the documented harm in this market has come overwhelmingly from what else was in the vial, not from exotic peptide toxicity.

The one practical thing this chapter will say. Whatever you decide about this compound, your clinician cannot account for what they do not know about. If you are using it, tell them — not so they can approve, which they will not, but so that the person interpreting your labs, prescribing your other medications, and evaluating your next unexplained symptom is working with complete information. Chapter 39 is entirely about making that conversation productive, including how to raise it with a clinician who you expect will react badly.

17.10 The honest ❌, written out in full

Here is the rating.

📊 Evidence Rating

Claim: BPC-157 accelerates healing of tendon and soft-tissue injuries in humans.

Rating: ❌ Hype outpaces evidence. (as of 2026)

Why: As of this writing there is no completed, peer-reviewed, randomized controlled human trial of BPC-157 for this or any indication. The supporting evidence is a substantial body of rodent work — genuine, often well conducted, and on the animal rung of the ladder.

What would change it: a completed, adequately powered, randomized controlled human trial with a pre-specified functional endpoint would move this to ⚠️ or ✅ depending on the result — or further toward ❌ if it failed. Even a well-conducted Phase I safety study would be a meaningful addition, since human safety data does not currently exist either.

What this rating does NOT say: that BPC-157 does nothing. That is not established either. The animal data is real and the human question is genuinely open. ❌ describes the state of the evidence, not the state of the molecule.

Now the part that this section exists for.

Writing an honest ❌ is harder than writing either a ✅ or a dismissal, and it is worth understanding why, because you are going to have to write several of them.

Why a ✅ is easy

A ✅ is a bibliography. The trials exist, you cite them, you note the effect size and the population and the safety profile, and the reader can check every claim. The work is in the reading, not in the writing. Chapter 8's rating of semaglutide for weight loss in obesity took a long time to research and almost no time to phrase, because when the evidence is there, describing it accurately and describing it favorably are the same act.

Why a dismissal is easy

A dismissal is even easier, because it requires no research at all.

Pick the weakest version of the claim — an influencer's most extravagant sentence, a badly written product page, someone's confident post about a shoulder. Quote it. Note that there are no human trials. Gesture at the gym. Conclude. The whole thing can be written in twenty minutes by someone who has read no primary literature, and it will be factually defensible in every particular.

It will also fail completely at its job, for three reasons.

It persuades nobody who matters. The reader who most needs the argument is the one who has actually read some of the animal work. When they encounter a dismissal that does not engage with it, they conclude — correctly — that the author has not read it either, and they discount everything else the author says. The dismissal reaches only people who already agreed.

It teaches nothing transferable. "This one is bunk" is a verdict, not a method. The reader cannot apply it to the next compound.

And it is often, on its own terms, an overclaim. "There's no evidence it works" slides into "it doesn't work," and that second sentence is not supported. It is the same error as the marketing claim, pointed the other way. Chapter 5's rule 4 — never downgrade a rating with distaste — exists because this failure is so easy and feels so much like rigor.

What an honest ❌ has to do

Four things simultaneously, and they pull against each other.

One: concede everything that is true. Not grudgingly, not in a subordinate clause, but at full strength and at length. §17.3 and §17.4 of this chapter are longer than §17.5 and §17.6 combined, and that is deliberate. If the case for a compound cannot be stated in its strongest form by the person rating it, the rating has not been earned. This is the steelman requirement, and it is the part most often skipped.

Two: refuse to overstate in the negative direction. The rating says the claim is unsupported. It does not say the claim is false. Those are different propositions with different truth conditions, and collapsing them is the exact error the rating system was built to prevent. Some ❌ compounds in this book will be ⚠️ or ✅ in a decade. Some will be shown not to work. The current state of the evidence does not tell you which, and pretending it does is a failure of nerve dressed as skepticism.

Three: name the falsifier. A rating that cannot be moved by any specified observation is not a scientific claim; it is a position. §17.11 is the falsifier for this rating, and it is written to be specific enough that a trialist could take it and design a study.

Four: be readable by the person using the compound. This is the constraint that makes the other three hard. Everything above has to be true and has to be written so that someone who injected this morning can read it without being insulted. Not because their feelings outweigh accuracy — they do not — but because a chapter they stop reading has conveyed nothing, and the alternative sources are worse and friendlier.

The strongest objection, and the answer

Let me put the best counterargument in its own words, because it is a good one and it deserves better than the strawman version.

"Your ❌ isn't a scientific finding. It's a report on funding. Nobody has run the trial because nobody can patent the molecule, not because anyone tested it and it failed. You're describing an economic fact and calling it an evidence rating. Meanwhile people are getting better."

Most of that is correct.

The trial has not been run largely for economic reasons (§17.11). The absence of evidence here really is a funding artifact rather than a scientific verdict. And I have already granted that the animal work is real, that the mechanism is coherent, and that the human question is open.

Here is why none of it moves the rating.

A rating is a statement about what is known, and "nobody looked" is not knowledge. The rating system does not claim to describe reality; it describes the state of the evidence about reality, and it says so in its own definition. That is not a weaseling distinction — it is the entire design. Chapter 5's rule 2 exists precisely so that ❌ can be used honestly for compounds that may well work.

And notice what the objection would license if accepted. If "the trial wasn't run for funding reasons" upgraded a rating, then every unfunded compound would be upgraded, including the ones that do not work — and there are far more of those, because nine in ten fail. The objection is a reason to want the trial run. It is not a reason to act as though it had been.

The last sentence of the objection — meanwhile people are getting better — is the one I take most seriously, and §17.7 is my answer. People with chronic soft-tissue injuries do get better, in large numbers, for reasons that include natural history, load management, rehabilitation, regression to the mean, and expectation. All of those operate whether or not the compound does anything. That is not a dismissal of anyone's experience. It is a description of why that particular experience cannot distinguish between the two worlds.

What ❌ does not license, in either direction

Since this rating gets misused by both camps, both misuses:

It does not license "it's a scam." The animal literature is real, the researchers are doing science, and the compound may well have activity in humans. Calling it a scam is a claim about intent and about efficacy, and neither is established.

It does not license "the evidence just hasn't caught up." That phrasing smuggles in the conclusion. It presumes the trial, when run, would be positive. Nine times in ten it would not be. The honest phrasing is that the evidence has not been gathered, and its direction is unknown.

And it does not license contempt for anyone using it. People with persistent injuries that conventional care has not fixed are not fools for trying something with a plausible mechanism and a real preclinical file. They are making a decision under uncertainty with incomplete information, which is what everyone does about everything. What this chapter offers is a more accurate picture of the uncertainty — not a verdict on the person.

17.11 What would change this rating

A rating that cannot be falsified is a preference. So here is the falsifier, specified tightly enough that you could hand it to a clinical trialist.

The trial

THE STUDY THAT WOULD SETTLE IT

  DESIGN
    Randomized, double-blind, placebo-controlled, parallel-group.
    Central randomization with concealed allocation. Pre-registered protocol
    and statistical analysis plan filed BEFORE the first participant.
    Intention-to-treat as the primary analysis.

  POPULATION
    Adults with chronic mid-portion Achilles tendinopathy — symptoms present
    at least 3 months, diagnosis confirmed clinically and on imaging.
    Chosen deliberately: it is common, it is disabling, it has a validated
    outcome instrument, it has a defined standard of care, and it is one of
    the conditions people actually use this compound for.
    Excluded: complete rupture, inflammatory arthropathy, recent
    corticosteroid injection, and the conditions §17.9 flags as open questions.

  CO-INTERVENTION
    ALL participants, both arms, receive the same structured progressive
    loading program with the same supervision and the same adherence
    monitoring.
    This is essential. The real-world question is not "does it beat nothing" —
    it is "does it add anything to what already works." A trial against
    nothing would answer a question no one is asking.

  PRIMARY ENDPOINT
    Change in a validated patient-reported tendinopathy function score
    (for Achilles tendinopathy the standard instrument is the VISA-A) at a
    single pre-specified timepoint.
    Patient-reported and functional — NOT tendon thickness, NOT a marker,
    NOT an imaging change. The endpoint has to be the thing that matters to
    the person: can they use the leg.

  SECONDARY ENDPOINTS
    Pain on loading; time to return to prior activity; imaging change;
    adverse events collected by a pre-specified schedule; and long-term
    follow-up to 6–12 months, because a treatment that accelerates recovery
    but does not improve the endpoint is a real and interesting result that a
    12-week trial alone cannot detect.

  DURATION
    12 weeks of intervention with the primary endpoint at 12 weeks;
    follow-up to at least 6 months, preferably 12.

  MAGNITUDE
    The between-group difference must exceed the established minimal
    clinically important difference (MCID) for the instrument — not merely
    reach statistical significance — and the trial must be POWERED for that
    difference, with the power calculation published in the pre-registration.
    A statistically significant improvement smaller than the MCID is a real
    finding about biology and not a reason for anyone to do anything.

  BEFORE ANY OF THIS
    A Phase I dose-escalation study in healthy volunteers with human
    pharmacokinetics, defined adverse event collection, and laboratory
    monitoring. This does not exist. It would be a meaningful addition on its
    own, independent of any efficacy question, because it would convert
    "unknown safety" into "characterized short-term safety" — and it would
    finally produce the human exposure data that §17.5's dose problem needs.

How each outcome would move the rating

The virtue of a specified falsifier is that you know in advance what you will conclude.

That last row is worth dwelling on, because it is the most likely near-term scenario and the one readers are least prepared for. The appearance of a human trial is not the same as the appearance of evidence. When the first one shows up, check the registration date against the enrollment date, check whether the endpoint was pre-specified, check the power calculation, check who funded it, and check whether the primary endpoint reported is the one that was registered. Chapter 5 gave you that checklist and this is the compound it was built for.

Why the trial has not been run

Now the uncomfortable structural fact, stated once and carefully.

No party has an incentive to fund the trial that could show BPC-157 does not work.

Consider each potential funder.

A pharmaceutical company would have to spend a large sum — trials of this size cost millions of dollars — on a compound whose sequence has been public for decades and which is therefore effectively unpatentable in its native form. A positive result would produce an approved product that competitors could copy; a negative result would produce nothing. The expected return does not justify the expense, and no company is behaving badly by declining. This is exactly the situation Chapter 5 described as the orphan compound problem: cheap, unpatentable molecules do not attract sponsors, regardless of merit.

The existing sellers are in an even clearer position. They currently sell the compound into an established market without any obligation to demonstrate efficacy. A trial has two possible outcomes from their perspective: a positive result that they cannot exclusively monetize because competitors sell the same molecule, or a negative result that destroys the market they already have. The rational choice is not to fund the study, and — this is the part worth being precise about — nobody has to be acting in bad faith for this to be the outcome. It falls out of the incentive structure automatically. Accusations of conspiracy are unnecessary and, more importantly, they distract from the actual problem, which is structural and would persist even if every individual involved were scrupulous.

Academic and government funders could run it, and this is the realistic route. Public research funding exists precisely to answer questions the market will not. So do philanthropic foundations, sports medicine research bodies, and institutional grants. Compounds do get rescued this way — the Chapter 5 case study of CAST is the canonical example, where a widely used practice was overturned only when a trial that nobody commercially wanted finally got run, and the result reversed a belief that had been considered obvious. It took a funder with no stake in the answer.

That is what is missing here. Not evidence that the compound fails. Not a conspiracy. A funder with no stake in the answer.

So the honest summary of the situation is: the question is open, the experiment is entirely feasible, the design above is not exotic, and the reason it has not happened is that the money to answer questions and the money to sell things flow through different channels — and this compound sits in the gap.

And the rating stays ❌ until it is run. Not because the compound has failed. Because the state of the evidence is what a rating describes, and this is what the evidence is.

📋 Your Evidence Dossier

This chapter fills the hardest field in the project: writing an honest ❌.

Every previous dossier entry has been descriptive — what a molecule is, what it does, what was tested. This one is a piece of writing, and it is the one people get wrong most consistently, in both directions. The hedge that refuses to conclude anything is as useless as the sneer that concludes too much.

Here is the procedure, and then a demonstration on BPC-157.

The six moves

WRITING AN HONEST ❌

  MOVE 1 — STATE THE STRONGEST TRUE CASE FOR
    In full. Not in a subordinate clause. If you cannot make the case for the
    compound better than its advocates can, you have not earned the rating.
    Test: would a knowledgeable advocate recognize their own position here?

  MOVE 2 — STATE THE EVIDENCE SITUATION IN ONE SENTENCE, WITH A DATE
    One sentence. Flat tone. Dated, because it is a claim about a moment.
    If it takes a paragraph, you are arguing rather than reporting.

  MOVE 3 — NAME THE RUNG
    Cell culture / animal / uncontrolled human / controlled human /
    replicated controlled human. Say which one, and say that rungs are not
    reached by accumulating studies on the rung below.

  MOVE 4 — SAY WHAT THE RATING DOES NOT CLAIM
    Explicitly, in its own sentence. ❌ is about evidence, not efficacy.
    This move is what separates a rating from a verdict, and it is the one
    that gets cut for space. Do not cut it.

  MOVE 5 — NAME THE FALSIFIER
    Specific enough to hand to someone who could run it. Design, population,
    endpoint, magnitude. "More research is needed" is not a falsifier; it is
    a way of not having one.

  MOVE 6 — READ IT BACK AS THE PERSON USING THE COMPOUND
    Literally. Read your own text imagining you injected this morning.
    Anywhere you flinch, ask whether the flinch came from a fact or from
    your tone. Facts stay. Tone gets rewritten.

Worked demonstration — BPC-157

DOSSIER ENTRY — BPC-157, tendon/soft-tissue healing    [worked demonstration]
                                                        [dated: 2026]

  FIELD 1 — IDENTITY
    15 amino acids (pentadecapeptide), GEPPPGKPADDAGLV, ~1,419 Da.
    Laboratory code only; no generic name ever assigned. Not approved in any
    major jurisdiction.

  FIELD 3 — MECHANISM
    NOT ESTABLISHED. No receptor definitively identified. Proposed pathways
    (nitric-oxide-related signaling, angiogenic and growth-factor signaling)
    are inferred from downstream observations in animal models.
    Write "not established." Do not write the hypothesis as though it were
    the finding.

  MOVE 1 — THE CASE FOR
    A substantial animal literature spanning tendon and ligament healing,
    gastrointestinal protection, vascular and angiogenic responses, and
    neuroprotection, from more than one research group, in peer-reviewed
    journals, using standard models with control arms and blinded outcome
    assessment. The direction of reported effect is consistent across all
    four model families. The proposed unifying mechanism — improved perfusion
    to injured tissue — is coherent and addresses a real bottleneck in
    poorly vascularized tissue such as tendon. This is a preclinical package
    that would plausibly justify a Phase I.

  MOVE 2 — THE EVIDENCE SITUATION
    As of 2026, there is no completed, peer-reviewed, randomized controlled
    human trial of BPC-157 for any indication in the published literature,
    and no published Phase I safety study.

  MOVE 3 — THE RUNG
    ANIMAL. The literature is large and it is entirely on one rung. Volume
    within a rung does not promote a claim to the rung above.

  MOVE 4 — WHAT THIS DOES NOT CLAIM
    It does not claim BPC-157 is ineffective in humans. That is unestablished
    in the same way and to the same degree. The human question is open, and
    the rating describes the evidence, not the molecule.

  MOVE 5 — THE FALSIFIER
    Randomized, double-blind, placebo-controlled trial in adults with chronic
    mid-portion Achilles tendinopathy, all participants on a standardized
    progressive loading program, primary endpoint a validated patient-reported
    function score at 12 weeks, powered for a difference exceeding the
    instrument's MCID, pre-registered. Result exceeding MCID → ⚠️;
    independently replicated → ✅; adequately powered null → ❌ of a better
    kind. A Phase I with human PK would be a meaningful addition on its own.

  MOVE 6 — THE READ-BACK
    Checked. Nothing here calls anyone credulous, and nothing here softens
    the evidence situation to be pleasant. The one sentence in MOVE 2 is the
    hardest sentence in the entry and it is stated once, plainly, without
    adornment, and then not repeated for emphasis.

Your turn

Do this for one compound in your own dossier that you currently rate ❌ — and make it one you wanted to work. That constraint is the whole exercise. The move that fails first is almost always MOVE 1, because it is genuinely hard to write the strongest case for something you have already concluded against.

Then do it for one you have no feelings about, and compare. If your ❌ for the compound you liked is noticeably gentler than your ❌ for the one you did not, you have found your bias — and Chapter 5's rules 3 and 4 exist because everyone has that bias and almost nobody can name theirs. Date both entries. Chapter 40 comes back for them.

Conclusion

BPC-157 is a fifteen-residue peptide, roughly 1,419 daltons, with four prolines that make part of it rigid, three glycines that make part of it flexible, a net negative charge, and a laboratory code instead of a name. It has a substantial animal literature spanning tendon, gut, vascular, and neural models, from more than one research group, reporting consistent directional benefit. It has no identified receptor. It has never been tested in a completed, published, randomized controlled human trial, for any indication, and it has no published human safety data.

Both halves of that paragraph are true, and this chapter's whole job was to show you why they do not contradict each other.

They do not contradict because animal results and human results are different kinds of knowledge, separated by five specific gaps — biology, disease, dose, endpoint, and a filtered publication record — that no amount of further animal work can close. Because roughly nine in ten compounds that reach human trials fail, which makes an impressive preclinical file the ordinary starting condition of a compound that will not work as much as of one that will. Because a cut rodent tendon and a worn human tendon may not be the same disease. And because the trial that would answer the question is entirely feasible, costs money, and sits in a gap where no one whose money it is has a reason to spend it.

The rating is ❌, and the fourth line of the rating box is not a courtesy. ❌ describes the state of the evidence, not the state of the molecule. BPC-157 may turn out to do something useful in humans. It may turn out to do nothing. Both futures are live, and the only instrument that can distinguish them has not been used.

If you take one thing from this chapter, take the ten-minute check in §17.6. Not because it settles BPC-157 — it does, but that is the small prize. Take it because it works on every compound in Part III, it costs you almost nothing, and it moves you permanently out of the population that has to be told what the evidence says and into the population that can look.

Chapter 18 turns to TB-500 and thymosin β4, where you will find a structurally different molecule, a partially overlapping set of claims, and a very similar evidence situation — and where you will get to run this chapter's whole procedure yourself, with much less help.

Key Terms

Pentadecapeptide — a peptide of exactly fifteen amino acids. BPC-157 is one; the term appears in the literature as part of its formal description.

Preclinical evidence — research conducted before human testing: cell culture, tissue, and animal studies. It generates candidates and cannot adjudicate them.

Translational gap — the difference between a result in a laboratory model and a result in human patients, spanned only by human trials. The reason roughly nine in ten compounds entering human trials never reach approval.

Transection model — an experimental injury created by surgically cutting a structure such as a tendon, ligament, or nerve. Acute, complete, standardized, and with a known time zero.

Tendinopathy — the preferred term for chronic tendon disorders, replacing "tendinitis" because these conditions typically show degenerative change and often relatively little classic inflammation.

Angiogenesis — the growth of new blood vessels. Central to healing in poorly vascularized tissue, and the core of BPC-157's proposed unifying mechanism.

Allometric scaling — the non-linear conversion of doses between species based on body size and metabolic rate. Necessary because a dose per kilogram in a rat does not correspond to the same dose per kilogram in a human.

Publication bias — the tendency for positive results to be published and null results not to be. Especially severe in preclinical work, which is largely unregistered, leaving the published record a filtered sample from a population of unknown size.

Trial registration — the public filing of a clinical trial's design before it begins. A registration is a plan, not a result; registered, run, completed, results-posted, and published are five distinct states.

Case report — a written account of a single patient. Valuable for flagging unexpected events, close to worthless for establishing efficacy: no control, no blinding, no randomization, no denominator.

Regression to the mean — the statistical tendency for an extreme measurement to be followed by a less extreme one. Because people begin treatments when symptoms are worst, it guarantees apparent improvement independent of any treatment effect.

First-pass metabolism — the processing of an absorbed substance by the liver, via the portal vein, before it reaches systemic circulation. The fifth of the five oral barriers.

Bioavailability — the fraction of an administered dose that reaches systemic circulation unchanged. Distinct from stability: a molecule can be entirely stable and have zero bioavailability if it cannot cross an epithelium.

Phase I trial — the first study of a compound in humans, typically a dose-escalation in healthy volunteers, producing pharmacokinetics and systematic safety data. None exists for BPC-157 in the published literature.

Minimal clinically important difference (MCID) — the smallest change in an outcome measure that patients perceive as meaningful. A statistically significant difference below the MCID is a finding about biology, not a reason to act.

Falsifier — the specific, pre-specified observation that would change a rating. A rating without one is a position rather than a scientific claim.

Orphan compound — a molecule that cannot be profitably developed because it is unpatentable, cheap, or off-patent, regardless of scientific merit. The structural reason many open questions stay open.

Gray market — the trade in compounds sold without approval for human use, often labeled "research use only." Carries identity, purity, potency, sterility, and endotoxin risks that are entirely independent of the molecule's pharmacology.

Spaced Review

1. (Ch 5 + Ch 17) Chapter 5 gave five reasons animal results do not transfer to humans. Choose the two that apply most forcefully to BPC-157 specifically — not to animal research in general — and defend your choice. Then state what could be done about each, and notice that only one of your answers involves running more animal studies.

2. (Ch 4 + Ch 17) A product is sold as an oral BPC-157 capsule, and the seller offers laboratory data showing the peptide remains intact after prolonged exposure to simulated gastric fluid. Using Chapter 4's five barriers, explain exactly what that data establishes and what it leaves entirely unaddressed. Then write the single question you would ask the seller next, and say what an acceptable answer would look like.

3. (Ch 6 + Ch 17) Chapter 6 described the pipeline by which a preclinical finding becomes a consumer claim. Trace one sentence — "BPC-157 accelerates tendon healing" — backwards through that pipeline to the experiment it originally described. At which stage did the word "rats" disappear, and why is that stage almost never the work of someone who intended to deceive?

4. (Ch 17) Write the ❌ rating for BPC-157 in four lines — claim, rating, reason, falsifier — without using the words "unproven," "bro science," "snake oil," or "just." Then read it back imagining you have been using the compound for six months. Which sentence, if any, would make you stop reading? Rewrite that sentence without changing what it asserts.

5. (Ch 5 + Ch 17) Someone tells you the ❌ is unfair because the trial was never run for financial reasons rather than scientific ones, so the compound is being punished for being cheap. State the strongest version of their argument in your own words, identify the part of it that is correct, and explain in two sentences why it nonetheless does not change the rating. Then name the one thing that would.