By the Peptevity Research Desk — vendor-neutral, evidence-graded. Scientific review: independent reviewer pending — how our review works.

Animal vs Human Evidence for Peptides: Why "Promising in Rats" Isn't "Works in People"

Direct answer

A finding in rats, mice, or a cell dish does not transfer to humans by default — it transfers rarely, and only after independent human trials confirm it. Across drug development, roughly 90% of compounds that enter human trials fail, despite having cleared animal testing first (Hartung, Front Drug Discov 2024). When researchers directly compared animal results to later human trials for six well-studied interventions, half were discordant — the animals said one thing, people did another (Perel et al., BMJ 2007, PMC1781970). A systematic scoping review found animal-to-human concordance rates ranging anywhere from 0% to 100%, with no reliable way to predict which (Leenaars et al., J Transl Med 2019, PMC6631915). This is why Peptevity grades animal data C–D and reserves the higher grades for human evidence. This page is research and educational information, not medical advice.

This is the methodological backbone of how Peptevity reads the peptide literature. Almost every peptide we cover — BPC-157, TB-500, GHK-Cu, MOTS-c, and most of the catalog — has an evidence base that is overwhelmingly rodent and in-vitro, with thin or absent human trials. The single most important question you can ask about any peptide claim is "in what species was this shown?" The rest of this page explains why that question decides everything, and connects it to our A–F evidence-grading methodology.

The headline number: most things that work in animals fail in people

The cleanest way to see the animal–human gap is to follow what happens after a compound clears animal testing and enters human trials. It has already passed the rodent and large-animal screens. And still:

  • Roughly 90% of drug candidates that enter clinical trials fail to reach approval; one widely cited industry analysis put the clinical-stage failure rate at about 95% (Hartung, Front Drug Discov 2024). These are compounds the animal data said were worth taking into people.
  • Of those failures, roughly 20–40% fail because of toxicity or side effects that the animal studies did not predict (Hartung, Front Drug Discov 2024). Most of the rest fail for lack of efficacy — the effect seen in animals simply did not show up in humans, which across the wider literature is the single largest reason clinical-stage candidates fail.

Put plainly: the animal stage is a filter that removes obviously toxic or useless compounds, and that is genuinely useful. But "passed the animal filter" is the start of human evidence, not a substitute for it. The 90% that fail had all "worked in animals" too.

There is a fair counter-framing worth stating, because we do not deal in scare numbers. Research-advocacy bodies note that the "9 out of 10 fail" figure is sometimes quoted out of context, and that animal screening does usefully remove a large fraction of dangerous candidates before they ever reach a person (Understanding Animal Research). Both things are true at once: animal models earn their place in early screening, and a positive animal result is a weak predictor of a positive human result. The honest reading is the second clause — which is exactly the clause peptide marketing drops.

When animal and human results were compared head-to-head

The most damning evidence is not the failure rate in the abstract; it is what happens when someone takes a specific animal finding and checks it against the human trial that followed.

  • Six interventions, half discordant. A landmark BMJ systematic review compared animal experiments with the human trials for six interventions where the clinical answer was already known. Three were concordant; three were not (Perel et al., BMJ 2007, PMC1781970). Corticosteroids reduced damage in animal head-injury models but showed no benefit in clinical trials; tirilazad helped in animal stroke models but was ineffective or harmful in people (Perel et al., 2007).
  • Concordance is unpredictable. A later systematic scoping review of 121 studies found reported animal-to-human concordance rates spanning the entire range from 0% to 100%, and concluded the evidence was too biased and too dated to make translation reliably predictable in advance (Leenaars et al., J Transl Med 2019, PMC6631915). Within that review, animal toxicology predicted human adverse events in some datasets only 37% of the time, and rodent positive-predictive value for human toxicity ranged from 0% to 54% (Leenaars et al., 2019).
  • Older syntheses agree. Reviews of the translational literature have long found that a substantial share of animal findings are contradicted in humans, with agreement often near a coin-flip across compared studies (Bracken, J R Soc Med 2009, PMC2746847).

A coin-flip is the right mental image. When you read "study showed [peptide] healed tendons / reduced inflammation / extended lifespan" and the study was in rats, the base rate that the same thing happens in humans is closer to a coin-flip than to a promise — and often worse, because the peptide field's animal studies are smaller and less rigorous than the drug-development studies these reviews analyzed.

Why the gap exists — four mechanisms, not bad luck

The animal–human gap is not random noise. It comes from identifiable, well-documented sources. Understanding them tells you what to look for in a claim.

1. Species physiology and metabolism genuinely differ

Different species metabolize the same molecule differently, express receptors differently, and respond to the same dose differently. The textbook cautionary tales run in both directions: thalidomide is famously not teratogenic in many animal species but is in humans, while corticosteroids are widely teratogenic in animals but not in people (Bracken, J R Soc Med 2009, PMC2746847). A peptide's stability, half-life, receptor binding, and breakdown products can all differ between a mouse and a person. For peptides specifically, this matters acutely: many are rapidly degraded by peptidases, and how fast that happens — and what fragments result — is species-dependent.

2. Dose doesn't scale one-to-one

A dose that produces an effect in a 25-gram mouse cannot be copied milligram-for-milligram to a 70-kilogram human, and it cannot be scaled by simple bodyweight either. Allometric scaling (adjusting for metabolic rate and body-surface area, not just mass) routinely changes the human-equivalent dose by large factors, and pharmacokinetic prediction from animals to humans is itself unreliable — one analysis found only about 52–65% of drugs had clearance predicted within a two-fold error from animal data (Leenaars et al., J Transl Med 2019, PMC6631915). This is one reason Peptevity never publishes human dosing: an animal-study dose is not a human dose, the conversion is genuinely uncertain, and translating it is exactly the error this page warns against. (For the lab-procedure framing of concentration math — not dosing — see how to reconstitute peptides.)

3. Animal models don't mimic the human condition

Animal disease models are simplified by design. A young, genetically uniform, otherwise-healthy rodent with a surgically induced injury is not a middle-aged human with comorbidities, other medications, and a different disease trajectory. The BMJ review attributed much of the observed discordance precisely to "the failure of animal models to mimic clinical disease adequately" — different timelines, absent comorbidities, simplified injuries (Perel et al., BMJ 2007, PMC1781970). An effect on a clean model wound in a healthy young rat tells you little about a chronic injury in a real patient.

4. The animal studies themselves are often weak

This one is uncomfortable but central. A large fraction of published animal studies are methodologically fragile — frequently not randomized, not blinded, and not adequately powered — and these weaknesses systematically inflate the apparent effect. Animal experiments conducted without randomization or blinding have been shown to be roughly five times more likely to report a positive treatment effect (Perel et al., BMJ 2007, PMC1781970). Reviews repeatedly find that the methodologically weakest studies show the strongest effects, and the strongest studies show weak or no effect (Bracken, J R Soc Med 2009, PMC2746847). So the glowing rodent result you are reading may be glowing partly because the study was poorly controlled.

Publication bias: the studies you never see

Even if every individual animal study were rigorous, the body of published animal evidence is skewed, because positive results get published and null results often do not. Systematic reviewers examining translational research have found "widespread publication bias" across the animal literature, including strong evidence of it in specific fields such as thrombolysis for stroke (Bracken, J R Soc Med 2009, PMC2746847; Perel et al., BMJ 2007, PMC1781970).

The practical consequence for a peptide reader: when you see five rodent papers all reporting that a peptide "worked," you are not seeing the experiments that found nothing and were never written up. The visible literature is the tip of an iceberg whose underwater mass is disproportionately negative. A handful of positive rat studies is therefore much weaker evidence than the same number of registered, pre-specified human trials, where null results are far likelier to reach publication. This is compounded in the peptide space, where many animal studies are small, old, and concentrated among a few research groups.

How this maps to Peptevity's A–F grades

This page is why our grading scale looks the way it does. The full rubric lives on the evidence-grading methodology page; here is the short version, anchored to everything above:

Grade What it means Why the gap on this page sets the ceiling
A Consistent, high-quality human RCTs / meta-analyses Human outcome data, independently replicated — clears the translation gap entirely
B Limited human evidence (small / single / surrogate-endpoint trials) Human, but thin — the direction is real but unreplicated
C Animal-only evidence Capped here because the animal→human base rate is near a coin-flip
D In-vitro / mechanistic only A cell-dish or gene-expression signal, furthest from a human outcome
F Contradicted, or claim unsupported by any credible study —

The hard line is between B and C: the presence or absence of human data. As we put it on the retatrutide monograph, a human trial is what earns a B; rodent data, however consistent, stays at C. This is not pedantry — it is the single most reliable filter for separating a peptide that might do something in people from one that has only ever done something in a mouse.

What to demand before you believe a human claim

Use this as a checklist against any peptide claim you encounter — including, especially, ours:

  1. Species. Was the result in humans, animals, or a cell culture? If the source doesn't say, treat it as animal/in-vitro until proven otherwise. This is the first question, always.
  2. Study type. A controlled human trial outranks a case series, which outranks an animal study, which outranks a cell-dish experiment. "Anecdotes" and influencer testimonials are below all of these.
  3. Replication. One positive study — animal or human — is a signal, not a settled fact. Independent replication is what moves a claim up the scale.
  4. Methodology. Was the animal or human study randomized and blinded? Unblinded, non-randomized work inflates effects roughly five-fold (Perel et al., BMJ 2007, PMC1781970).
  5. The missing studies. Are you seeing only the positive results? Publication bias means null findings are often invisible (Bracken, J R Soc Med 2009, PMC2746847).
  6. The dose. If a claim cites an animal dose as if it were a human dose, the conversion is unreliable and the claim is mishandling the evidence — animal-to-human pharmacokinetics are predicted within two-fold only about half to two-thirds of the time (Leenaars et al., J Transl Med 2019, PMC6631915).

When a vendor or influencer says a peptide "is proven to heal injuries" and the proof is rodent tendon studies, items 1, 3, and 5 alone are usually enough to downgrade the claim from "proven" to "preliminary signal in animals." That is the whole game.

Honest bottom line

Animal and in-vitro data are the beginning of a scientific story, not the end. They generate hypotheses worth testing in humans — and roughly nine times out of ten, the human test doesn't confirm them (Hartung, Front Drug Discov 2024). Half of head-to-head comparisons came out discordant (Perel et al., BMJ 2007, PMC1781970), concordance is unpredictable across the literature (Leenaars et al., J Transl Med 2019, PMC6631915), and the animal studies themselves are often unblinded, underpowered, and published only when positive (Bracken, J R Soc Med 2009, PMC2746847). None of this means animal research is worthless — it means "promising in rats" is a starting line, and the marketing that prints it as a finish line is making a category error.

This is why nearly every compound monograph on this site grades its animal findings C and its mechanism D, and reserves B and A for the rare peptide with real human trials. If you read one principle into Peptevity's whole catalog, read this one: ask what species it was shown in, and grade accordingly. Most "research peptides" remain not FDA-approved and are sold "research use only — not for human consumption" — a status that is itself shifting through 2026 and into 2027, tracked on our dated 2026 regulatory tracker. This page is research and educational information, not medical advice; Peptevity sells nothing and publishes no human-dosing protocols.

Frequently asked questions

Do animal studies on peptides apply to humans? Not by default. Across drug development, about 90% of compounds that pass animal testing and enter human trials still fail (Hartung, Front Drug Discov 2024), and when animal results were compared directly with the human trials that followed, half were discordant (Perel et al., BMJ 2007, PMC1781970). An animal result is a hypothesis to be tested in humans, not evidence of a human effect. That is why Peptevity grades animal-only findings C and in-vitro findings D, per our evidence-grading methodology.

Why doesn't "it worked in rats" mean it works in people? Four documented reasons: species differ in metabolism and receptors (thalidomide is teratogenic in humans but not many animals; corticosteroids the reverse) (Bracken, J R Soc Med 2009, PMC2746847); animal doses don't scale cleanly to humans and pharmacokinetics are predicted within two-fold only about half to two-thirds of the time (Leenaars et al., 2019, PMC6631915); animal disease models don't reproduce the human condition (Perel et al., 2007, PMC1781970); and many animal studies are unblinded and non-randomized, which inflates apparent effects roughly five-fold (Perel et al., 2007, PMC1781970).

What is the translational failure rate from animal studies to human trials? Roughly 90% of drug candidates that enter human clinical trials fail despite having cleared animal testing, with one industry analysis putting clinical-stage failure near 95%; about 20–40% of those failures are due to unanticipated toxicity (Hartung, Front Drug Discov 2024), with most of the remainder attributed across the wider literature to lack of efficacy — the animal effect not reproducing in humans. A systematic scoping review found animal-to-human concordance rates ranging from 0% to 100% with no reliable predictor (Leenaars et al., 2019, PMC6631915).

Does publication bias affect peptide research? Yes. Systematic reviewers have documented widespread publication bias in the animal literature — positive results get published while null results often do not (Bracken, J R Soc Med 2009, PMC2746847). So a cluster of positive rodent papers overstates the true picture, because the experiments that found nothing are largely invisible. This is one reason a handful of positive animal studies is much weaker evidence than registered human trials.

Are animal studies useless then? No, and this page does not say that. Animal screening usefully removes many toxic or ineffective candidates before they reach people, and some of the "9 out of 10 fail" framing is quoted out of context (Understanding Animal Research). The accurate position is narrower: animal data generate hypotheses and are a reasonable early filter, but a positive animal result is a weak predictor of a positive human result — so it should be graded as preliminary, not presented as proof.

How we graded this page

This is a methodology explainer, not an efficacy claim about a specific compound. Every translational, statistical, and methodological statement is tied to a peer-reviewed primary source per our evidence-grading methodology and sourcing and citation policy. Peptevity carries no advertising, no affiliate links, and sells nothing — see our conflict-of-interest and funding statement, and our medical disclaimer and RUO statement.

References

External references appear as citations only; none of the cited institutions endorse, review, or are affiliated with Peptevity.

Related on Peptevity

Every claim above is cited inline to a primary source. See how we grade evidence and our sourcing & citation policy.