Chapter 31: Peptides in Veterinary Medicine — Where Some "Human" Peptides Were First Used

"Essentially, all models are wrong, but some are useful." — George E. P. Box and Norman R. Draper, Empirical Model-Building and Response Surfaces (1987)

Overview

There is a sentence that appears, in some form, in nearly every conversation about the compounds in Part III. It goes like this: "This has been used in animals for years."

It is usually offered as reassurance, and usually meant sincerely. Sometimes it is even true. And it is one of the most information-poor sentences in the entire field, because it compresses at least four separate questions into a single clause and answers none of them.

This chapter takes that sentence apart. To do that honestly, it has to begin somewhere that may surprise you: with an insistence that veterinary medicine is real medicine. Not a lesser cousin of human medicine, not a regulatory loophole, not a place where standards are lower. Veterinary pharmacology has approved drugs, its own regulators, its own controlled trials, its own adverse-event reporting, and a substantial peer-reviewed literature.

So the argument here is not that veterinary evidence is second-rate. It is narrower, more technical, and much more useful: veterinary evidence answers veterinary questions. It was gathered in a particular species, in a particular condition, against a particular endpoint, for a particular purpose. Move any one of those and the evidence stops covering you.

This is the chapter where you learn exactly how that transfer fails — not as a slogan about mice and men, but mechanically. Receptors differ in sequence and distribution. Clearance differs with body size in a way that is not proportional. Dose does not scale linearly with mass, and the standard correction runs in a direction most gray-market reasoning gets backwards. Disease models differ from the diseases they are named after. Lifespans differ by a factor of forty.

You will also learn where the traffic runs the other way, because it genuinely does. Insulin's development involved dogs. Exenatide came from a lizard. Veterinary practice adopts human drugs constantly. What this chapter objects to is a specific direction of inference — from an animal received it to a human should — not to animal science, which is one of the reasons we know anything at all.

In this chapter, you will learn to:

• Describe the veterinary regulatory and evidence infrastructure accurately, including the peptide drugs genuinely approved for animal use
• Explain why a racing ban is a statement about the rules of a sport rather than about pharmacology
• Read a livestock growth-promotion study and name what it measured — and what it therefore cannot say
• List five specific mechanisms by which species differences break translation, with a concrete example of each
• Explain allometric scaling and predict the direction of error from naive body-weight conversion
• Run the four questions on any "used for years" claim and say what would actually need to be true
• Define extra-label use and withdrawal periods, and name the asymmetry the gray market exploits
• Separate a critique of inference from a critique of animal research

Learning Paths

All five paths should read §31.5 and §31.6. Those two sections contain the reusable machinery; the rest of the chapter is the evidence that makes them stick.

💊 GLP-1 — §31.5 is where the rodent thyroid C-cell finding from Chapter 8 gets its proper explanation. If you have wondered what that boxed warning does and does not mean for you, read it. 🏋️ Performance — one of the two or three most important chapters in the book for you. §31.2, §31.3, and §31.6 dismantle the three arguments you will hear most often; the allometric scaling material in §31.5 is the single most common quantitative error in the space. 🔬 Science — §31.5 and §31.9 are the substance. The model-validity vocabulary (face, construct, predictive) is worth carrying into every paper you read afterward. 💄 Cosmetic — lighter going, but §31.5's point about route bears on Chapter 30, and §31.6's four questions apply verbatim to "clinically proven" cosmetic claims. 🏥 Clinical — §31.7 is the section to read closely. Patients who bring you a veterinary justification are usually making a specific, nameable error, and naming it beats arguing about the compound.

31.1 The veterinary pharmacopeia is real medicine

Let us set the frame properly, because a lot of writing on this topic gets it wrong in a way that makes the actual argument impossible to hear.

Veterinary medicine is not human medicine with the standards relaxed. It is a licensed profession with its own doctoral training, board specialties, malpractice exposure, and regulatory apparatus. In the United States, animal drugs are approved by the Food and Drug Administration's Center for Veterinary Medicine (CVM); veterinary biologics such as vaccines fall under the Department of Agriculture. In the European Union, veterinary medicines run through the European Medicines Agency's Committee for Medicinal Products for Veterinary Use. Approval requires demonstrating that a product is safe and effective for a named species, a named indication, and a labeled dose — and, for animals entering the food supply, that it is safe for the people who will eventually eat them.

That last clause has no analogue in human medicine, and §31.7 returns to it as one of the most revealing features of the whole system.

There are controlled trials in veterinary medicine, randomized and blinded, published in journals such as the Journal of Veterinary Internal Medicine, the Equine Veterinary Journal, and the Journal of Animal Science. There is post-approval pharmacovigilance: adverse events associated with approved animal drugs are reported and collected. In some respects — sample sizes in production-animal research, where a single study may enroll thousands of animals — the veterinary literature is better powered than its human equivalent.

And peptides are genuinely part of it.

💊 In the Clinic — peptide drugs that are actual veterinary medicines

This is not a short list, and it is worth reading slowly, because most people who invoke "veterinary use" as a credential have never encountered any of it.

GnRH analogs for reproductive management. Gonadotropin-releasing hormone agonists — gonadorelin, buserelin, deslorelin — are approved and widely used across veterinary species. In cattle, GnRH agonists are a routine component of estrus-synchronization and timed-breeding programs, which is one of the foundations of modern dairy and beef management. In dogs and in some exotic species, sustained -release deslorelin implants produce reversible suppression of reproductive function through the same downregulation mechanism you met in Chapter 27. This is a real drug class, doing a real job, with real evidence behind it.

Insulin for diabetic dogs and cats. Diabetes mellitus is common in both species and is managed with insulin, including products licensed specifically for veterinary use. This is chronic peptide-hormone replacement therapy in a companion animal, monitored and adjusted over years. It is not a fringe practice; it is standard internal medicine.

Desmopressin. The vasopressin analog you met in Chapter 1's naming table treats central diabetes insipidus in dogs and is also used in some bleeding disorders — the same molecule, the same mechanism, the same clinical logic as in people.

Diagnostic peptides. Synthetic ACTH analogs run adrenal stimulation tests in dogs suspected of Cushing's or Addison's disease. Here the peptide is not a treatment at all; it is a probe.

Appetite stimulation. A ghrelin-receptor agonist — a growth hormone secretagogue in the same family as the compounds in Chapter 15, though a peptidomimetic rather than a peptide — is approved in the United States for appetite stimulation in dogs, with a version approved for cats. Hold onto this one: a compound class that largely stalled in human development and succeeded in veterinary medicine. It matters again in §31.9.

None of these are gray-market compounds. They are labeled products with approved indications, prescribed by licensed veterinarians, with known adverse-effect profiles and reporting channels.

Notice what all of those have in common. Each is scoped. The GnRH implant is approved for particular species and purposes. The insulin product is labeled for dogs and cats, and a good veterinarian knows the two species diverge enough that management strategy differs. The evidence supporting each is excellent evidence — for the population it was gathered in.

That is the whole thesis of this chapter, and it cuts both ways. You should not dismiss a veterinary finding as "just animals." You also cannot pick that finding up and carry it into a human.

📊 Evidence Rating

Claim: GnRH agonists produce effective, reversible suppression or control of reproductive function in the veterinary species and indications for which they are approved. Rating: ✅ Strong clinical evidence (assessed as of this writing, 2026) Reason: These are approved veterinary drugs with controlled studies, labeled indications and species, decades of practice experience, and post-approval surveillance. The mechanism, the endpoint, the species, and the evidence all match. What would change it: Systematic failure to reproduce the effect in well-controlled veterinary trials, or the emergence of a safety signal substantial enough to withdraw approvals. Neither is in evidence.

This rating is here on purpose. A book that only ever rated veterinary evidence ❌ would be telling you that veterinary science is weak, which is false. The ✅ is the control condition: it shows that the machinery is rating the fit between evidence and claim, not the species the study was run in. When a claim is correctly scoped to the population that was studied, veterinary evidence can be as strong as any evidence in this book.

31.2 TB-500, thymosin β4, and the racehorse argument

Now to the compound that generates more veterinary hand-waving than any other.

Chapter 18 covered this in full, and one point from it needs restating because everything here depends on it: thymosin β4 and the compound marketed as TB-500 are not the same molecule. Thymosin β4 is a 43-residue endogenous peptide with well-studied roles in actin regulation, cell migration, and wound repair. What is sold as "TB-500" is typically a short fragment corresponding to the actin-binding region — a handful of residues, not the parent. The fragment is marketed on the parent's biology, which is a substitution of one molecule for another, and Chapter 18 works through what does and does not survive the swap.

🧬 The Molecule — the fragment and the parent, one more time

Thymosin β4: 43 residues, a genuinely interesting endogenous peptide, studied in wound healing, corneal repair, and cardiac tissue, with some legitimate clinical investigation behind it.

"TB-500": as sold, a short synthetic fragment built around the actin-binding motif. Vendors routinely present it as thymosin β4 or as "the active portion" of it.

The phrase the active portion is doing enormous unearned work. A fragment containing a functional motif is not automatically a functional molecule. It has different size, different charge distribution, different conformational behavior, different clearance, different tissue distribution, and no guarantee of the parent's binding partners. Sometimes fragments retain activity. Sometimes they retain one activity out of several. Sometimes they do nothing.

Which of those is true for this fragment is an empirical question, and Chapter 18 reports the state of the answer. What matters here is that when someone cites a horse study, or a thymosin β4 paper, or an equine practitioner's experience, you have to ask which molecule they are talking about — and the answer is frequently "not the one in the vial."

With that established, the veterinary argument. Thymosin β4 and its fragment have been used in equine practice, and the compound is well known in racing contexts. That is true. People will tell you so, and they are not lying.

Two distinct points follow, and they are usually run together into a single bad inference.

First: use in an animal is not evidence of efficacy in a human. This is the whole of §31.5 and §31.6, so we will not spend it here. Note only that "used in horses" does not even establish efficacy in horses. Use establishes that people administered something. It does not establish that anything happened.

Second, and this is the subtler error: racing jurisdictions have prohibited these substances. Prohibited-substance lists are maintained by national and international racing authorities, and thymosin β4 and related compounds appear on them. In human sport, the World Anti-Doping Code likewise prohibits growth factors affecting muscle, tendon, and ligament, and separately prohibits substances with no current approval for human therapeutic use — a category sweeping in many Part III compounds, including BPC-157, which was added to the Prohibited List explicitly.

Here is what people do with that fact:

"They banned it in racing. You don't ban something that doesn't work."

⚠️ Hype Check — "it's banned, so it must work"

This is one of the most persuasive-sounding bad arguments in the field, and it deserves a careful answer rather than a dismissal, because there is something real inside it.

What's true in it. Regulators do not usually write rules about nothing. If a substance appears on a prohibited list, somebody thought it worth prohibiting — which usually means people were using it and that a plausible pharmacological rationale existed. That is genuine information: the compound is in circulation and a rationale exists.

Where it fails. It confuses a regulator's decision with an efficacy finding, and those come from completely different processes optimizing completely different things. A racing or sporting authority prohibits a substance for reasons that include, at minimum:

Uncertainty. A compound with unknown effects in a competing animal is a hazard by default. Prohibition is often the response to not knowing — precisely the opposite of a finding that it works.

Animal welfare. If a substance might mask injury, alter pain perception, or let an animal run on a compromised limb, that is sufficient reason to prohibit it whether or not it heals anything. §31.8 returns to this.

Regulatory status. Many lists prohibit unapproved substances as a class. The compound is not listed because a regulator found it potent; it is listed because it is not an approved medicine and nobody can vouch for what is in the syringe.

Integrity of competition. Sport regulates perception as well as physiology. A substance widely believed to confer advantage damages confidence in results even if it confers none.

Administrability. It is far easier to prohibit a whole pharmacological class than to adjudicate each member, and class-based prohibitions sweep in compounds nobody has ever studied.

Verdict: a prohibition tells you a compound is being used and that an authority decided the risk of permitting it exceeded the benefit. It tells you nothing about whether it repairs tissue in a horse, and less than nothing about whether it does so in you. The list is a risk-management document, not a pharmacology textbook. Never upgrade a rating because something is banned — that is rule 3 of the rating system (Chapter 5) wearing a disguise.

One more thing about the equine argument, uncomfortable for both sides. The horse cannot tell you whether its shoulder feels better. Objective equine endpoints — imaging, lameness scoring, gait analysis, return to racing — are real and can be measured well. But "the horse improved" is frequently an unblinded judgment by a human with a financial stake, about a nonverbal patient who was also rested, iced, shod differently, and worked back gradually over the same period. That is not a knock on veterinarians; it is precisely why careful veterinary researchers run controlled trials instead of relying on impressions, exactly as careful human researchers do.

📊 Evidence Rating

Claim: TB-500 (the marketed thymosin β4 fragment) accelerates tendon, ligament, or muscle repair in humans, on the basis of its use in horses. Rating: ❌ Hype outpaces evidence (assessed as of this writing, 2026) Reason: Two independent failures. Veterinary use is not human evidence — it does not control for species, indication, endpoint, or measurement — and, per Chapter 18, the compound in circulation is a fragment rather than the parent peptide whose biology is being cited. The argument therefore transfers the wrong evidence about the wrong molecule. What would change it: Adequately powered, randomized, controlled human trials in a defined injury population with objective functional and imaging endpoints, using a characterized compound of stated identity and purity. A better-designed horse study would not change this rating; it would improve the equine evidence, which is a different rating.

Note carefully what this ❌ does not say. It does not say thymosin β4 biology is uninteresting — it is interesting. It does not say the compound cannot work. It says the argument being made for it does not support the conclusion being drawn. That is rule 2: the ❌ describes the evidence.

31.3 GHRPs, livestock, and a literature that is large, rigorous, and silent

Growth-hormone-releasing peptides and related growth-axis compounds have been studied in food animals. So have growth hormone itself, growth-hormone-releasing hormone analogs, and a long list of other interventions aimed at making an animal convert feed into muscle more efficiently.

This is not a marginal literature. It is decades deep, generously funded, methodologically careful, and often very well powered, because the economics are enormous and the questions are answerable. A production system feeding tens of thousands of animals can detect a small improvement in feed efficiency and act on it. The studies are frequently randomized, blinded where blinding is possible, and analyzed by people who are very good at mixed models.

So this is a case where "there's a big animal literature" is completely true. Now look at what it measures.

WHAT LIVESTOCK GROWTH RESEARCH ACTUALLY MEASURES

  PRIMARY PRODUCTION ENDPOINTS
    Average daily gain (ADG) ............ kilograms of body mass added per day
    Feed conversion ratio (FCR) ......... kilograms of feed per kilogram of gain
    Dry matter intake ................... how much the animal eats

  CARCASS ENDPOINTS (measured after slaughter)
    Dressing percentage ................. carcass weight as a fraction of live weight
    Backfat thickness ................... fat depth at a standard anatomical site
    Loin eye / ribeye area .............. cross-sectional muscle area
    Marbling score ...................... intramuscular fat, a grading criterion
    Lean-to-fat ratio ................... carcass composition

  METABOLIC / ENDOCRINE ENDPOINTS
    Circulating GH, IGF-1 ............... did the axis respond as designed
    Nitrogen retention .................. a proxy for protein deposition

  SAFETY ENDPOINTS
    Tissue residue depletion ............ how long until the meat is safe TO EAT
    Withdrawal interval ................. the regulatory consequence of the above

  WHAT IS NOT ON THIS LIST
    Strength. Athletic performance. Injury incidence. Tendon or ligament healing.
    Recovery between exertions. Pain. Sleep. Mood. Wellbeing. Function.
    Quality of life. Anything the animal would report if it could.

Read that list twice, because it contains the entire lesson of this section.

Every endpoint on it is economically meaningful and scientifically legitimate. Feed conversion ratio is a genuine measure of a genuine thing. Carcass composition determines the grade and therefore the price. None of this is junk science.

And not one of those endpoints is the endpoint a human user cares about. Nobody in that literature was asking whether the compound helps a rotator cuff heal, or improves recovery between training sessions, or increases strength as opposed to mass, or makes anyone feel better. Those questions were not answered badly. They were not asked, because they were irrelevant to the purpose of the research and, in most cases, unmeasurable in the animals being studied.

A literature can be large, rigorous, and completely silent on the question you are asking.

That sentence is worth memorizing, because the failure mode it names is not limited to livestock. It is the general shape of misused evidence: someone points at a substantial body of work, correctly notes that it is substantial, and never checks whether it measured the thing under discussion. Call it endpoint mismatch. It is the most polite form of misinformation, because everything said is true and the conclusion is still unsupported.

🔬 Read the Study — the shape of a livestock growth trial

The design below is the standard architecture of this literature rather than a summary of any single paper. It is described this way deliberately: the point is the shape, not a citation.

Population. Several hundred to several thousand animals of a single species, breed, sex, and approximate age, entering a defined production phase — weaned pigs, feedlot steers, broiler chickens. Genetically far more uniform than any human population, on an identical diet, in identical housing.

Design. Randomized to treatment and control, often with pen or pen-group as the unit of randomization, sometimes factorial against diet composition. Blinding of personnel where feasible.

Duration. Weeks to a few months — the length of the production phase, not of the animal's life.

Primary endpoints. Average daily gain and feed conversion ratio. That is usually it.

Secondary endpoints. Carcass measures at slaughter, hormone concentrations, sometimes morbidity or mortality within the production period.

Analysis. Mixed models accounting for pen effects, adequately powered to detect small differences because the sample size is large.

What this design is excellent at. Detecting whether the compound changes growth rate and feed efficiency, precisely, at economically relevant effect sizes.

What this design cannot detect, ever. Anything requiring the subject to report an experience. Anything on a timescale longer than the production phase. Anything about function under load, because these animals are not asked to perform. Anything about injury recovery, because the animals are not injured. Anything about long-term consequences, because the animals do not live long enough to have any — they are slaughtered at the end of the study, which is not a limitation of the design but its context.

The honest summary. This is good science answering the question it was designed to answer. If your question is different, this study is not evidence about your question. It is not weak evidence. It is not evidence at all, in the way that a very precise thermometer is not weak evidence about your blood pressure.

There is a second observation buried in that endpoint list, and it may be the most useful thing in this section.

Look at the safety endpoints: tissue residue depletion, withdrawal interval. The safety question being asked in production-animal research is how long must we wait before this animal's tissue is safe for a person to eat? The protected party is a consumer who will ingest trace residue, orally, in cooked muscle, months later.

That is close to the exact opposite of the exposure a person has when they inject a compound at a bioactive dose. The livestock safety literature is designed to establish that almost none of the compound remains — not to establish what happens when a great deal of it is deliberately introduced into a person. Citing it as human safety data inverts the question it was built to answer.

📊 Evidence Rating

Claim: Growth-hormone-releasing peptides improve strength, body composition, recovery, or athletic performance in humans, on the basis of livestock growth-promotion research. Rating: ❌ Hype outpaces evidence (assessed as of this writing, 2026) Reason: The livestock literature's endpoints are average daily gain, feed conversion ratio, and carcass composition in young food animals on controlled diets over a production phase. It contains no measurements of strength, function, injury, recovery, or wellbeing, and its safety endpoints concern dietary residue in human consumers rather than administered exposure. The literature is silent on the claim, and silence is not support. What would change it: Randomized controlled human trials with functional endpoints — strength, performance, body composition by validated methods, injury incidence — in a defined population, at characterized exposures. Chapter 15 tracks the human growth-secretagogue evidence that does exist and rates it on its own terms; this rating concerns the livestock inference specifically.

31.4 BPC-157: animal research is not veterinary use

Chapter 17 covers this compound thoroughly, including its origin story, its mechanistic claims, the shape and quality of its literature, and the current state of human investigation. This section makes one narrow point and then refers you back.

The narrow point is a distinction people collapse constantly. There is a difference between "used in animals" and "used in laboratory animals."

When someone offers "used in animals for years" as reassurance, the sentence sounds like it means veterinarians have prescribed it to patients — a licensed clinician, a labeled product, a treated dog with an owner and a follow-up appointment.

For BPC-157, that is not what happened. Essentially all of the evidence is animal research — conducted as research, in laboratory animals, under experimental protocols, with a substantial portion of it in rodents. Those animals were not patients. They were subjects: purchased, housed, randomized, treated, and in many studies euthanized so that tissue could be examined. The compound was not prescribed to them; it was administered as an independent variable.

This is not a criticism of the research. That is what preclinical work is, and it is a necessary stage of drug development — every approved drug in this book passed through something like it. The criticism is of the rhetorical slide, in which "used in animals" borrows the warmth of veterinary practice for a body of work that is nothing of the kind.

A further point follows directly. BPC-157 is not an approved veterinary drug either. It has no labeled indication in any species. A veterinarian cannot dispense it as an approved product for a dog any more than a physician can prescribe it for a person. So the reassurance being offered — animals get this all the time — is not merely mis-transferred. In the sense the listener understands it, it is not true.

The second thing to hold from Chapter 17 is this: the compound never had a human development program. Not one that stalled at Phase III, not one that failed on a safety signal, not one that was abandoned for commercial reasons after showing promise. Chapter 1 already gave you the tell, before any of this evidence was on the table: a compound still known by a laboratory code, decades after its first description, is telling you something about its regulatory history. Chapter 17 supplies the detail and the current state of the exceptions.

Chapter 17 also supplies the specific model-mismatch example that §31.5 is about to use, so if you have not read it recently, the transected-tendon material is the part to refresh.

31.5 Species differences: how translation actually breaks

This is the technical core of the chapter — the part that turns "animals are not people" from a truism into a checklist.

There are at least five distinct mechanisms by which a finding in an animal fails to reach a human. They are independent — a study can survive one and fail another — and each is checkable in about ninety seconds.

FIVE PLACES THE INFERENCE BREAKS

  animal study result
        │
        ├─ (1) RECEPTOR ────── does the human receptor have the same sequence,
        │                       the same affinity, and the same tissue distribution?
        │
        ├─ (2) CLEARANCE ───── does the compound persist for the same relative
        │                       time in a body with a different metabolic rate?
        │
        ├─ (3) DOSE SCALING ── is the exposure comparable, given that dose does
        │                       NOT scale linearly with body mass?
        │
        ├─ (4) MODEL ───────── is the animal's condition the same disease as
        │                       the human condition it is named after?
        │
        ├─ (5) TIME ────────── is the study duration the same FRACTION of a
        │                       lifespan in both species?
        │
        ▼
  human claim

  Every arrow is a place the inference can fail silently. None of them announce
  themselves in an abstract. All five must hold for the transfer to be sound.

Take them in order.

(1) Receptors differ in sequence and in distribution

A peptide acts by binding a receptor. If the receptor is not the same protein, or is not in the same places, the same molecule does a different thing.

The standing example in this book is the rodent thyroid C-cell finding from Chapter 8. Long-acting GLP-1 receptor agonists produced thyroid C-cell tumors in rodents, which is the origin of the boxed warning on that drug class. The reason this may not transfer is specific and mechanistic: rodent thyroid C-cells express GLP-1 receptors far more abundantly than human C-cells do. The rodent tissue is, in effect, listening to a signal that the human tissue barely hears. A finding driven by abundant receptor expression in one species is not automatically a finding about a species where that expression is sparse.

Note what that argument is and is not. It is a mechanistic reason to doubt transfer, and it is a good one. It is not proof of human safety, which is why the warning remains and why long-term human surveillance continues. Chapter 8 handles the current state of that question properly. What matters here is the shape: receptor distribution is a species-specific fact, and it can create a signal in one species that has no counterpart in another.

Sequence differences matter just as much, and Chapter 11 gives the cleanest illustration. Insulin is not the same molecule in every mammal. Porcine insulin differs from human insulin by a single amino acid; bovine differs by three. Those differences mattered clinically in the era of animal-sourced insulin, for both potency and immunogenicity. And here is the detail that makes the point: porcine insulin is identical in sequence to canine insulin, which is a substantial part of why porcine-derived products have worked so well in diabetic dogs. Feline insulin is closer to bovine.

So the same word — insulin — names slightly different molecules across species, with clinically relevant consequences. If the ligand differs by species, expect the receptor to as well. It does.

(2) Metabolic rate and clearance differ

Small mammals run hot and fast. Metabolic rate does not scale in proportion to body mass; it scales to roughly the three-quarters power of mass, a relationship known since the 1930s and usually called Kleiber's law. A mouse's mass-specific metabolic rate is many times a human's. Heart rate, respiratory rate, renal filtration relative to body size, and hepatic enzyme activity all follow.

The consequence for a peptide is direct. Half-life, time to peak concentration, total exposure, and duration of receptor occupancy are all different in a mouse than in a person, and not by a factor you can guess. A compound producing a brief pulse in a rodent may produce a much longer exposure in a human, or the reverse, depending on whether clearance is dominated by proteolysis, renal filtration, or receptor-mediated uptake — and those three scale differently from one another.

This is why an animal finding that a compound "works" tells you very little about what exposure a human would need to reproduce it. The dose is not the exposure, and the exposure is what the receptor sees.

(3) Dose does not scale linearly with body weight

Now the single most common quantitative error in gray-market reasoning about animal data. It is worth being blunt.

You cannot convert a rodent dose to a human dose by multiplying milligrams per kilogram by body weight. Doing so systematically overestimates the human-equivalent dose, and often by roughly an order of magnitude.

The standard regulatory approach — used when estimating a maximum safe starting dose for a first-in-human study — normalizes not to body mass but to body surface area, which tracks metabolic rate better than mass does. In practice this is implemented with species-specific correction factors. The conversion from an animal dose in milligrams per kilogram to a human equivalent dose in milligrams per kilogram runs, approximately, as follows: divide a mouse dose by about 12, a rat dose by about 6, a rabbit or monkey dose by about 3, a dog dose by about 2.

Those divisors are the whole story. Consider what happens when someone skips them.

NAIVE CONVERSION vs. ALLOMETRIC SCALING     [constructed teaching example]

  A hypothetical compound is studied in mice at 10 units per kilogram.
  (Units, deliberately — this is arithmetic about study design, not a dose
  for anything, and no number here applies to any real compound or person.)

  NAIVE mg/kg REASONING
    "10 per kg in a mouse, so 10 per kg in a human."
    For a 70 kg person:  10 × 70 = 700 units total.

  BODY-SURFACE-AREA (ALLOMETRIC) SCALING
    Mouse-to-human divisor ≈ 12.3
    10 ÷ 12.3 ≈ 0.81 per kg human-equivalent
    For a 70 kg person:  0.81 × 70 ≈ 57 units total.

  THE GAP
    700 vs. 57 — the naive figure is roughly TWELVE TIMES higher.
    The error is not random. It runs in ONE direction: naive conversion
    always overestimates, because it ignores that small animals clear
    faster per unit mass.

  AND THE SECOND ERROR, WHICH IS WORSE
    This method estimates a STARTING dose for a first-in-human SAFETY
    study. It is a floor-finding tool with a safety factor built in.
    It is not a calculator for an effective dose, and treating it as one
    imports a conservatism that was designed for a completely different
    purpose.

Two things about that example deserve emphasis.

First, the direction. This is not a coin flip. Naive milligram-per-kilogram conversion from a small rodent always produces a larger number than surface-area scaling does, because rodents clear faster relative to their size.

Second, even allometric scaling is a rough instrument, and for peptides rougher than usual. The method was built around compounds cleared by hepatic metabolism; peptides are cleared by proteolysis and renal filtration, and a peptide engineered for protease resistance or albumin binding (Chapter 33) may scale in yet another way. Surface-area scaling is a defensible default when you have nothing better. What you actually want is measured pharmacokinetics in the species you care about — which, for a compound with no human program, is exactly what does not exist.

🩺 Safety and Risk — why the scaling error is a safety problem and not just a math error

It is tempting to file allometric scaling under "interesting technicality." It is not one.

A person reads a rodent paper reporting an effect at some dose per kilogram. They multiply by their own body weight. They arrive at a number roughly an order of magnitude above what a regulatory conversion would produce for a starting dose in a first-in-human safety study — a figure that is itself deliberately conservative.

Three things have gone wrong at once. The arithmetic is wrong in a known direction. The tool has been repurposed from safety-floor estimation to efficacy targeting, which it was never designed for. And the human pharmacokinetics of the compound are unknown, so there is no measurement anywhere in the chain to check the answer against.

A fourth problem sits underneath all of them: none of this addresses what is actually in the vial. Concentration, identity, purity, and sterility are separate questions with their own failure modes, and Chapters 19 and 34 document where the real documented harm in the unregulated market has come from. A person can do the scaling arithmetic perfectly and still be injecting something other than what the label says, at a concentration nobody measured.

The general principle: when a calculation's inputs are all estimates and its purpose has been changed, its output is not a number. It is a guess wearing a number's clothes. Decisions about any of this belong with a clinician who can order labs and see you, not with arithmetic derived from a mouse.

(4) The model is not the disease

Animal models of human disease are evaluated on three kinds of validity, and the vocabulary is worth owning because it makes an entire genre of overclaiming visible.

Face validity — does the model look like the human condition? Does the animal display similar signs?

Construct validity — does the model arise from the same underlying mechanism as the human condition?

Predictive validity — do interventions that work in the model also work in humans? This is the only one that matters for the inference this chapter is about, and it is the one least often established.

A model can have excellent face validity and no construct validity at all, and that combination is where most translational failure lives. Two examples already in this book:

Chapter 17's tendon example. Rat models of tendon healing frequently use a surgically transected tendon — the tendon is cut and the repair is measured. That is a clean, reproducible, quantifiable injury with a well-defined healing trajectory. The human condition people want to treat is usually overuse tendinopathy: a chronic, degenerative, largely non-inflammatory condition developing over months of repetitive loading, with disorganized collagen and often no acute rupture at all. These are not the same injury. Face validity — it is a tendon, and it heals — is high. Construct validity is low, because the mechanisms differ. Predictive validity is unestablished.

Chapter 13's ob/ob mouse. The ob/ob mouse is profoundly obese and was the animal in which leptin was discovered — genuinely one of the great successes of model-organism research. But the mouse is obese because it cannot make leptin. It has a single-gene deficiency. Common human obesity is not a leptin-deficiency state; most people with obesity have high circulating leptin and reduced responsiveness to it. Giving leptin to an animal that has none corrects a deficiency, dramatically. Giving it to a person who already has plenty does much less. The model was right about the molecule and wrong about the market, and Chapter 13 tells that story in full.

There is a general form here. A model built to be tractable is often built to be different. The features that make an animal model reproducible — acute rather than chronic, genetic rather than multifactorial, uniform rather than heterogeneous, young rather than old — are frequently the same features that separate it from the human condition.

Two related mismatches are routinely ignored. Route: much rodent work uses intraperitoneal injection, oral gavage, or the compound dissolved in drinking water, so "effective in mice" sometimes means "effective by a route no human would use." Subject characteristics: laboratory rodents are young, genetically near-identical, pathogen-free, unmedicated, and caged on a fixed diet, while the humans interested in these compounds are older, diverse, and often taking other medications. Each difference is a place where an effect can shrink or reverse.

(5) Lifespans differ by a factor of forty

The last mechanism is the easiest to state and the most often forgotten.

A laboratory mouse lives roughly two to three years. A standard rodent carcinogenicity study runs about two years — which is to say, most of the animal's adult life. That is a genuinely long-term study in that animal. A two-year study in a human covers about two and a half percent of a lifespan.

This cuts both ways. Downward: when a rodent study reports no adverse findings over its full duration, that is meaningful for the rodent and says nothing about human exposures lasting decades. The compounds in Part III that people take continuously for years have no evidence base of comparable relative duration anywhere.

Upward: when a two-year rodent study does find something — a tumor signal, an organ change — it found it over a substantial fraction of a life, under conditions designed to be sensitive. That is why regulators take such findings seriously even while acknowledging they may not transfer, and it is exactly the posture taken with the C-cell finding in (1). The finding is taken seriously and its transferability is questioned. Holding both at once is most of what separates careful reading from motivated reading.

🔍 Check Your Understanding

1. A rodent study finds a tissue effect at a given dose in milligrams per kilogram. Someone multiplies that figure by their own body weight to estimate a human dose. Name the direction of the resulting error and explain, in one sentence, why it runs that way.
2. Which type of model validity — face, construct, or predictive — is the one that actually licenses an animal-to-human inference? Which is the one most often demonstrated in the papers cited by vendors?
3. The rodent thyroid C-cell finding (Chapter 8) is used by some people to argue that GLP-1 drugs are dangerous and by others to argue that rodent findings are meaningless. Explain why both readings are wrong, using the receptor-distribution argument from this section.
4. Give one reason a two-year rodent study is simultaneously long-term evidence and short-term evidence.

31.6 "It's been used for years" — four questions

This is the most portable section in the chapter, and if you retain one page of it, retain this one.

The claim appears constantly, in these and similar forms:

"Vets have been using this in horses for twenty years." "This has been given to animals since the 1990s with no problems." "It's been used in livestock for decades — it's one of the most studied compounds out there."

Each of these is offered as a conclusion. Each is actually an unanswered question, or rather four of them. Here they are, in the order that most efficiently kills a bad claim.

Question 1 — In whom?

Which species, of what age, in what condition?

A horse is not a person. Neither is a rat, a beagle, a broiler chicken, or a rhesus macaque. §31.5 gives five specific reasons why, and any one of them is sufficient to break the inference. Ask this first because it is the cheapest question and it fails fastest.

Question 2 — For what?

What was the indication, and is it the indication you have in mind?

Growth promotion is not injury recovery. Estrus synchronization is not longevity. Appetite stimulation in a cat with kidney disease is not appetite regulation in a healthy adult. A compound used for decades in one indication has generated exactly zero evidence about a different one.

This question catches the endpoint mismatch from §31.3, and people find it the most surprising, because "it's been studied a lot" feels like it should be transitive across indications. It is not.

Question 3 — Measured how?

What was the outcome measure, and could it have captured what you care about?

Weight gain is not function. Carcass composition is not strength. Return-to-racing is not tendon histology. And in animals, nobody reports pain, fatigue, mood, or wellbeing, because the subject cannot report and the observer is inferring.

This matters enormously for the compounds in Part III, because a very large fraction of the human claims made for them are about subjective experience: it feels better, I recover faster, my joints hurt less. Those are legitimate outcomes — patient-reported outcomes are real endpoints in real trials — but they are categorically unmeasurable in an animal. An animal literature cannot support a subjective human claim even in principle. Not weakly. Not at all.

Question 4 — And collected how?

Was anyone recording?

This is the question almost nobody asks, and it is frequently the one that ends the conversation.

Veterinary practice, like human practice, generates enormous quantities of clinical experience and comparatively little systematic outcome data. A compound can be administered in a practice for twenty years without any structured record of who received it, what their baseline was, what happened, or how many did badly. Approved veterinary products do have pharmacovigilance channels — that is one of the things approval buys — but those channels capture adverse events for labeled products used through licensed routes. They do not capture outcomes for a compound obtained outside that system, and they cannot capture efficacy at all.

So "used for years" may mean, and often does mean, "administered for years without anyone measuring anything."

A track record is not a dataset unless somebody was recording.

That is the sentence to keep. It generalizes far beyond peptides, and it is the reason that decades of confident clinical practice have periodically been overturned the first time somebody ran a trial.

A fifth question is implied by the fourth: compared to what? Clinical experience has no control group. The animals that recovered are visible; the ones that did not are attributed to the injury. Compounds that were quietly dropped generate no stories, because nobody tells stories about the thing they stopped using. Absence of reported harm is not absence of harm when there is no reporting mechanism — it is absence of a reporting mechanism.

Running the four questions

Here is the whole procedure applied to a real-sounding claim.

CLAIM: "TB-500 has been used in racehorses for twenty years with no problems."

  Q1  IN WHOM?  Horses. Different receptor distribution, clearance, lifespan,
      and musculoskeletal loading. Also: which molecule — the 43-residue parent
      or the marketed fragment? (Ch 18)

  Q2  FOR WHAT?  Typically musculoskeletal injury in an animal that must return
      to racing. If your question is chronic tendinopathy, that is not it.

  Q3  MEASURED HOW?  Return to racing, lameness scoring, sometimes imaging.
      Often unblinded human judgment about a nonverbal patient who was ALSO
      rested, re-shod, and worked back gradually. No animal reported how
      anything felt.

  Q4  COLLECTED HOW?  In most cases, not systematically. No registry, no
      denominator, no control group, no record of the horses it did not help.

  AND: COMPARED TO WHAT?  Nothing. No counterfactual anywhere in this claim.

  WHAT THE CLAIM ACTUALLY SUPPORTS
      That people have administered something to horses for a long time.
      That is the entire content. Every further inference is supplied by
      the listener.

Note that running the four questions did not require knowing anything about thymosin β4 biology. It required knowing what kind of statement was being made. That is the point — this is a procedure you can apply to a compound you have never heard of, which is exactly when you most need it.

📊 Evidence Rating

Claim: "This compound has been used in animals for years, therefore it is safe in humans." Rating: ❌ Hype outpaces evidence (assessed as of this writing, 2026) Reason: This rating attaches to the inference, not to any molecule. The claim fails all four questions of §31.6: the species is different (§31.5 gives five mechanisms by which that matters); the indication is usually different; the outcome measures used in animals cannot capture human subjective or functional endpoints; and veterinary or laboratory administration frequently generates no systematic outcome data at all. A track record is not a dataset unless somebody was recording. What would change it: Nothing about the animal record. This particular inference cannot be repaired by better animal data, because the defect is structural rather than evidentiary. What would change the underlying safety question for any specific compound is human safety data — dose-ranging studies, defined exposures, systematic adverse-event capture in people. That is a different evidence base, and for most Part III compounds it does not exist.

This is a rating of a claim form. It is included to make rule 1 concrete: ratings attach to claims, and a claim can be an argument as easily as it can be a molecule. When you meet this sentence in the wild, you do not have to research the compound to know the argument is unsound.

31.7 How veterinary drugs are actually regulated — and what the gray market borrows

If veterinary use is going to be invoked as a credential, it is worth knowing what that credential consists of. It turns out to be quite a lot, and the details matter.

Approval

A veterinary drug approval in the United States requires the sponsor to demonstrate, to the Center for Veterinary Medicine, that the product is safe and effective for a specified species, a specified indication, and a specified dose regimen, and that it can be manufactured consistently. For food-producing animals, two additional arms apply: human food safety, which establishes residue tolerances and withdrawal periods, and an environmental assessment.

The label that emerges is a legal document. It names the species, the indication, the route and regimen, the warnings and contraindications. Departing from it has a name and a set of rules, which brings us to the second concept.

Extra-label use

Extra-label use is the veterinary analogue of off-label prescribing, and in the United States it is governed by the Animal Medicinal Drug Use Clarification Act. It is permitted — but conditionally, and the conditions are the interesting part:

• It must be by or on the lawful order of a licensed veterinarian, within a valid veterinarian-client-patient relationship — meaning the veterinarian has actually assumed responsibility for the animal, has seen it, and is available for follow-up.
• The animal's health must be threatened, or suffering or death may result from failure to treat.
• There must be no approved animal drug labeled for the intended use that is clinically adequate.
• For food-producing animals, additional constraints apply: the veterinarian must establish an extended withdrawal period, and certain drugs and drug classes are prohibited from extra-label use entirely because residues are unacceptable at any level or because resistance concerns override.

Read that list again with the gray market in mind. Every clause is a control: a licensed professional, an established relationship, a clinical necessity, an exhaustion of approved options, and a residue safeguard. Extra-label use is not a loophole. It is a regulated exception with named preconditions and an accountable person attached.

Withdrawal periods

A withdrawal period is the interval that must elapse between the last administration of a drug to a food-producing animal and the point at which that animal's meat, milk, or eggs may enter the human food supply. It exists because the patient becomes food, and it has no counterpart in human medicine.

THE WITHDRAWAL PERIOD — a concept human medicine has no need for

  treat the animal ──────── withdrawal period ────────► food supply
        │                                                   │
   drug present in                                 residues below the
   edible tissue                                   established tolerance

  HOW IT IS SET      Residue depletion studies track drug and metabolite
                     concentrations in edible tissue over time; a tolerance
                     is set from toxicology on the residue itself; the
                     withdrawal period is the time to fall below it, plus
                     a safety margin.
  HOW IT IS ENFORCED Residue testing in the slaughter and milk supply chains.
  WHO IS PROTECTED   A person who never met the animal, will never know it
                     existed, and faces a trace oral exposure months later.

Sit with that last line. The veterinary regulatory system contains an entire apparatus — years of toxicology, analytical chemistry, and enforcement infrastructure — devoted to protecting a stranger from residues measured in parts per billion in a piece of meat.

The asymmetry

Now the point this section has been building toward, and it is the sharpest thing in the chapter.

Compounds sold to humans as "research chemicals" frequently point to veterinary use as reassurance. Vets use this. It's used in animals all the time. The invocation borrows the authority of everything described above: the approval standard, the labeled indication, the licensed prescriber, the residue science, the enforcement. And the veterinary framework that would actually make that use accountable is precisely what the gray market lacks.

WHAT VETERINARY MEDICINE PROVIDES        WHAT THE GRAY MARKET PROVIDES

  A licensed prescriber who has            A website
  examined the patient and is
  responsible for follow-up

  A labeled product with a stated          A vial with a printed name and no
  species, indication, dose, and           independent verification of identity,
  manufacturer                             concentration, or purity

  A demonstrated manufacturing             "Research grade," which is a marketing
  standard                                 term rather than a specification

  An adverse-event reporting channel       Forum posts

  A withdrawal period, where the           Nothing analogous, because there is no
  question of downstream exposure          regulator asking the question
  is asked at all

  Legal accountability if any of           A "not for human consumption" label,
  the above fails                          which exists to transfer liability
                                           rather than to protect anyone

The last row deserves a sentence of its own. The phrase "not for human consumption" or "for research use only" is not a safety statement. It is a regulatory posture — a way for a product to avoid being classified and regulated as a drug. Chapters 19 and 34 cover what that does to the supply chain. Here, note only the irony: the same label that invokes veterinary and research use as reassurance also disclaims, in writing, any intention that a human use it.

So the argument is: the gray market borrows the credibility of a system while operating outside every one of its controls. It cites veterinary use and omits the veterinarian.

🔍 Check Your Understanding

1. Name three conditions that must be satisfied for extra-label use of a drug in a veterinary patient in the United States. Which of them, if any, is satisfied when a person buys a compound online?
2. What question does a withdrawal period answer, and who is it protecting? Why does human medicine have no equivalent?
3. A vendor's site says its product is "used by veterinarians" and also carries a "not for human consumption" label. Explain why those two statements are in tension, and what each one is actually for.

31.8 Performance animals and consent

A short section, because the point is simple and does not improve with elaboration. Performance animals cannot consent, and the incentive structure around them is not aligned with their interests.

A racehorse does not choose to race. It cannot decline a treatment, report that something hurts, weigh a short-term gain against a long-term risk, or seek a second opinion. Every decision about what enters its body is made by a person, and most of those people have a financial relationship to the animal's performance.

That is not an accusation. Many owners, trainers, and veterinarians care deeply about the animals in their charge, and the veterinary profession's ethical obligations run to the patient. But the structure is worth naming plainly: when the party who benefits from an intervention is not the party who bears its risk, and the party bearing the risk cannot object, you have an arrangement requiring external safeguards rather than good intentions.

This is a substantial part of why racing medication rules exist at all. Integrity of competition is one stated purpose; welfare is another, and it is not decorative. A compound that alters pain perception or masks injury in an animal that will then be asked to run at speed on that limb is a specific and serious hazard, because the animal cannot decline. The catastrophic-injury problem in racing sits exactly at the intersection of a compromised limb, a loaded gait, and a patient who cannot say no.

There is a version of this argument that becomes moralizing, and this book is not going to make it. Working, sport, and food animals exist; the ethics of that are outside this chapter's scope and readers will differ. The narrow, in-scope observation is this: the consent asymmetry is a reason to be more careful about extrapolating from performance-animal practice, not less. Practices persist in that setting partly because the subject has no voice in whether they persist, which means their persistence carries less information than it would in a population that could object.

Chapter 43 takes up the human version of this problem — the athlete, the service member, the employee, the person whose "choice" to use something is shaped by what happens if they do not. The animal case is the cleaner illustration precisely because the consent is not merely compromised but absent.

31.9 The genuine two-way street

Everything to this point has been about a failure of inference in one specific direction. This section is about the directions where the traffic is legitimate. It is not a consolation paragraph.

Animal science gave human medicine some of its foundations.

Insulin's development involved dogs — the pancreatectomized dog preparations of the early 1920s are inseparable from the discovery, and the first therapeutic insulin came from bovine and porcine pancreas. Chapter 11 tells that story. Millions of people lived because of work done with animals.

Comparative physiology has produced real drugs, repeatedly.

Exenatide came from a lizard. Chapter 4 covers it: exendin-4, isolated from Gila monster venom, is a GLP-1 receptor agonist that resists the enzyme destroying human GLP-1 within minutes. Nobody designed that. A reptile did, and a comparative endocrinologist noticed. The entire GLP-1 therapeutic class owes a debt to that observation.

Nor is that isolated. The first orally active ACE inhibitor descends from a peptide found in pit viper venom. An approved analgesic potent enough to require delivery into spinal fluid comes from a cone snail. Salmon calcitonin — a fish peptide — is more potent in humans than human calcitonin is, a lovely inversion of everything this chapter has been saying: sometimes the species difference is the gift. Venoms in particular are peptide libraries refined by selection for exactly the receptor specificity drug designers spend careers trying to engineer.

Model organisms have revealed mechanisms that were then confirmed in humans.

Leptin is the case in point, and it is worth stating carefully because §31.5 used it as a cautionary tale. Leptin was discovered through the ob/ob mouse, and that discovery was correct, important, and transformative: it established that adipose tissue is an endocrine organ, which reoriented the entire field of metabolic research. The mouse was right about the biology. What it was wrong about was the therapeutic market. The model told the truth about the mechanism and misled about the application. Both halves of that sentence are the lesson.

And veterinary medicine adopts human drugs constantly.

This is the direction people forget entirely. A very large fraction of the pharmacopeia used in companion-animal practice originated in human medicine: analgesics, antibiotics, cardiac drugs, oncology agents, and — as §31.1 noted — insulin and desmopressin, frequently under extra-label rules, because the veterinary market is too small to support independent development for every indication. The veterinary direction is often the harder one, since the veterinarian must reason across species with less data than a physician has.

Recall the appetite-stimulant example from §31.1. A ghrelin-receptor agonist in the same family as the compounds in Chapter 15 is an approved veterinary medicine doing real clinical work, while the human development path for that class largely did not produce approved products for the indications people hoped for. That is not veterinary medicine getting a hand-me-down. That is a compound class finding the population where its risk-benefit profile actually works — this chapter's argument pointed in the friendly direction: evidence belongs to the population it was gathered in, and sometimes that population is not us.

So let us be precise about the target.

This chapter's argument is about the direction of inference — from animal use to human claim — not about the value of animal science. Animal research is how hypotheses are generated, mechanisms identified, safety signals first detected, and dose ranges first bounded. It is a necessary stage and frequently a brilliant one. What it is not, and was never designed to be, is a substitute for the human trial that comes after it.

The failure mode is not "using animal data." The failure mode is stopping there and calling it evidence for a human claim. Preclinical work earns a compound the right to be tested in people. It does not earn it a conclusion.

📋 Your Evidence Dossier

This chapter adds a species-translation block to Field 2 for any compound whose supporting evidence is animal or veterinary. Appendix C has the workbook pages.

Field 2 is where you recorded what each of your compounds is claimed to do and by what mechanism. For most entries, some or all of the support behind that field comes from animals. This chapter gives you the tool to say so precisely rather than vaguely.

The species translation block

For each compound whose Field 2 rests on animal or veterinary evidence, record the following. If you cannot fill a line, write "not stated" — and treat that as a finding, because a citation that does not specify its species or endpoint is not usable.

FIELD 2 — SPECIES TRANSLATION BLOCK

  Species .................. which animal, and is it a lab model or a veterinary patient?
  Model .................... what condition was induced or present, and how?
  Endpoint ................. what was measured, in what units, by whom, blinded or not?
  Duration ................. how long, and what fraction of that species' lifespan?
  Route .................... how was it administered, and does that route exist in humans?
  Dose-scaling basis ....... mg/kg as reported? surface-area corrected? not stated?
  Molecule identity ........ the parent peptide, or a fragment/analog? (see Ch 18)

  WHAT HUMAN QUESTION THIS EVIDENCE ADDRESSES
    ........................................................................

  WHAT HUMAN QUESTION IT DOES NOT ADDRESS
    ........................................................................

Those last two lines are the point of the exercise. Everything above them is bookkeeping; the two sentences at the bottom are where bookkeeping becomes judgment. Write them as full sentences — a fragment lets you avoid committing.

Worked demonstration — two entries, deliberately opposite

SPECIES TRANSLATION — TB-500 / thymosin β4 fragment     [worked demonstration]

  Species          Rodents (laboratory research); horses (veterinary/racing use)
  Model            Rodent: induced or surgical injury. Equine: naturally
                   occurring musculoskeletal injury in performance animals
  Endpoint         Rodent: histology, tissue markers, healing scores. Equine:
                   largely return-to-work and clinical impression
  Duration         Weeks (rodent); variable and often unrecorded (equine)
  Route            Injection; rodent work includes routes without human analogues
  Dose scaling     Reported as mg/kg. Human-equivalent conversion NOT established;
                   no human pharmacokinetic data to calibrate to
  Molecule         CRITICAL: marketed product is a short fragment, not the
                   43-residue parent whose literature is being cited (Ch 18)

  ADDRESSES        Whether the parent peptide's biology is plausible and worth
                   investigating. It is — a real answer to a real question.
  DOES NOT ADDRESS Whether the marketed fragment, at any exposure, improves
                   healing, function, pain, or recovery in a human with an
                   overuse injury. No human trial has asked. → ❌ (§31.2)

SPECIES TRANSLATION — GnRH agonist, veterinary use       [worked demonstration]

  Species          The approved species named on the label — and only those
  Model            Not a model. Actual patients with the labeled indication
  Endpoint         Objective and physiological: ovulation timing, conception
                   rates, suppression of reproductive function, hormone levels
  Duration         Matched to the clinical purpose and labeled duration of effect
  Route            Labeled route, in the labeled formulation
  Dose scaling     Not applicable — dose was established IN the target species,
                   which is the only way dose is ever established well
  Molecule         The approved product, identity verified, made to a standard

  ADDRESSES        Whether this drug does this job in this species. Fully. → ✅
  DOES NOT ADDRESS Anything about humans — not because the evidence is weak but
                   because it was never pointed at a human question. GnRH analogs
                   in humans have their own evidence base (Ch 27), gathered in
                   humans, which is the point

Notice the structure of the contrast. The second entry is rated ✅ and the first ❌, and the difference has nothing to do with which species was involved. It has to do with whether the claim matches the population, indication, and endpoint the evidence was gathered in. The veterinary entry earns its ✅ by being modest about its scope; the other earns its ❌ by overreaching twice, across species and across molecules.

One instruction for your own entries. Go back through the compounds you chose in Chapter 1 and mark, for each, whether its Field 2 mechanism claim rests on human data, animal data, or mechanism alone. Then count. Most readers find the proportion uncomfortable, and that discomfort is the most useful thing this project produces.

Conclusion

Veterinary medicine is real medicine. It has approvals, regulators, controlled trials, pharmacovigilance, and a serious literature, and peptides are genuinely part of its pharmacopeia — GnRH analogs, insulin for diabetic dogs and cats, desmopressin, diagnostic peptides, appetite stimulants. That is good evidence. It is good evidence about the animals it was gathered in.

The problem is never the veterinary data. The problem is the transfer.

Five things break on the way from an animal to a human. Receptors differ in sequence and distribution, which is why a rodent thyroid signal may not transfer. Clearance differs with body size, so exposures do not correspond. Dose does not scale linearly with mass, and naive milligram-per-kilogram conversion overestimates the human equivalent in a predictable direction and by a large factor. Disease models differ from the conditions they are named after — a transected tendon is not overuse tendinopathy, and a leptin-deficient mouse is not a person with common obesity. And lifespans differ by enough that "long-term" means two different things.

When someone tells you a compound has been used in animals for years, you now have four questions that convert that sentence back into the inquiry it always was. In whom? For what? Measured how? And collected how? Most such claims fail on the fourth, because a track record is not a dataset unless somebody was recording.

Meanwhile the traffic runs the other way too, legitimately. Insulin came partly from dogs. Exenatide came from a lizard. Salmon calcitonin outperforms the human molecule in humans. Leptin was found in a mouse. Veterinary practice adopts human drugs constantly.

The quarrel is with one specific move: treating an animal received this as a human should. Preclinical work earns a compound the right to be tested in people. It does not earn it a conclusion.

Chapter 32 turns to how these molecules are actually made — because once you have stopped accepting animal use as evidence, the next question is what is in the vial, and that turns out to be a manufacturing question before it is a clinical one.

Key Terms

Veterinary pharmacopeia — the drugs approved and used in veterinary medicine, including a substantial set of peptide medicines with labeled species and indications.

Center for Veterinary Medicine (CVM) — the U.S. FDA division that approves and regulates animal drugs, including human food safety for food-producing species.

Extra-label use — use of a drug in a way not described on its approved label; the veterinary analogue of off-label prescribing, permitted under defined statutory conditions.

AMDUCA (Animal Medicinal Drug Use Clarification Act) — the U.S. statute setting those conditions, including a licensed veterinarian and, for food animals, an extended withdrawal period.

Valid veterinarian-client-patient relationship — the veterinarian has assumed responsibility for the animal, knows it well enough to diagnose, and is available for follow-up. A precondition of lawful extra-label use.

Withdrawal period — the interval after treatment before a food animal's meat, milk, or eggs may enter the human food supply. No analogue in human medicine.

Residue tolerance — the maximum permissible concentration of a drug or its metabolites in edible tissue, from which the withdrawal period is derived.

Allometric scaling — scaling a quantity by a power of body mass rather than in proportion to it. For dose, the standard normalization is to body surface area.

Human equivalent dose (HED) — an animal dose divided by a species-specific correction factor. A tool for setting maximum safe starting doses, not effective doses.

Body surface area normalization — the basis of cross-species dose conversion; tracks metabolic rate better than mass and yields far lower figures than naive milligram-per-kilogram conversion.

Feed conversion ratio (FCR) — kilograms of feed per kilogram of body mass gained. A primary livestock production endpoint.

Average daily gain (ADG) — body mass added per day; the other primary livestock growth endpoint.

Carcass composition — the distribution of lean tissue, fat, and bone in a slaughtered animal.

Growth promotion — intervening to raise growth rate or feed efficiency in food animals. An economic objective, not a clinical one.

Endpoint mismatch — when evidence is real and rigorous but measured something other than the outcome under discussion, and is therefore silent rather than supportive.

Model organism — a species studied as a stand-in for another, chosen for tractability and therefore partly for the ways it differs from what it stands in for.

Face validity — how far a model superficially resembles the human condition it represents.

Construct validity — how far a model arises from the same underlying mechanism as that condition.

Predictive validity — how far interventions effective in a model prove effective in humans. The only validity that licenses this chapter's inference, and the least often established.

Prohibited substance list — a sporting or racing authority's list of substances barred from competition. A risk-management document reflecting uncertainty, welfare, integrity, and regulatory status — not a finding of efficacy.

Pharmacovigilance — systematic collection of adverse events after approval. It requires a reporting channel, which products sold outside regulated systems do not have.

Comparative physiology — the study of physiological variation across species; a productive source of therapeutic leads, and a reminder that species differences are sometimes an opportunity.

Thymosin β4 — a 43-residue endogenous peptide involved in actin regulation and tissue repair; the parent of the fragment marketed as TB-500 (Chapter 18).

Actin-binding domain — the region of thymosin β4 that interacts with actin, and the basis of that fragment. A motif is not a molecule.

Translational gap — the failure of preclinical findings to reproduce in humans; this chapter enumerates five of its mechanisms.

Spaced Review

1. (Ch 31 + Ch 18) Someone shows you an equine case series describing improvement after treatment with "TB-500." Before evaluating the outcomes at all, name the two independent problems with using this as support for a human tissue-repair claim — one from §31.2 and one from Chapter 18's account of what the marketed compound actually is.

2. (Ch 31 + Ch 17) Chapter 17 describes rodent tendon studies using surgically transected tendons. Using the face/construct/predictive vocabulary from §31.5, classify that model's validity against human overuse tendinopathy, and explain which type of validity a vendor citing those studies is implicitly claiming.

3. (Ch 31 + Ch 5) Write a complete four-line Evidence Rating — claim, rating, one-sentence reason, what would change it — for the claim "BPC-157 is safe for long-term human use because it has been given to animals for decades without reported problems." Your "reason" line must invoke at least two of the four questions from §31.6, and your "what would change it" line must name a specific, obtainable kind of evidence.

4. (Ch 31) A rodent study reports an effect at a given number of milligrams per kilogram. Show, qualitatively, why multiplying that number by a person's body weight overestimates the human-equivalent dose, name the approximate mouse-to-human divisor used in standard regulatory conversion, and state the second, more serious error involved in using that method to target an effective dose.

5. (Ch 31 + Ch 5) Explain, in a paragraph you could say aloud to a skeptical friend, why this chapter rates GnRH analogs in veterinary species ✅ and the veterinary argument for TB-500 in humans ❌, without saying anything that implies veterinary evidence is inferior to human evidence.