Evidence review: Open Peptide Project (verified against primary PubMed, ClinicalTrials.gov & FDA records) · Last checked 2026-10-27. ← Back to Evidence Library

Why rodent peptide studies often fail in humans (and how to read preclinical claims)

Name: Why rodent peptide studies often fail in humans (and how to read preclinical claims)
Published: 2026-10-27

Species differences, dose-scaling myths and publication bias, and why this library grades animal-only evidence as No human data

Evidence verdict: Based on systematic reviews of animal-to-human translation, FDA first-in-human guidance and this library's published grading rules. [1, 3, 7, 4]

Last reviewed: Oct 27, 2026

In 30 seconds

Treatments that work in animals often fail, or even harm, in human trials. A systematic review found animal and human results agreed for some interventions and clearly disagreed for others. [1] [Controlled study, review]
Mouse and human biology can differ in ways that matter. In one landmark analysis, gene responses in mouse models of inflammation were close to random at matching the human conditions they were meant to model. [2] [Lab]
Animal results are skewed by publication bias. In animal stroke research, it may have inflated reported efficacy by about a third. [3] [Lab, review]
This library grades compounds only on published human evidence. Compounds with only cell or animal data get No human data, whatever the rodent results show. [4] [Policy]

Animal results often do not carry over

Perel and colleagues compared animal studies with human trials for six interventions whose human effects were already known. Corticosteroids for head injury helped in animals but not in human trials. Tirilazad for stroke looked strongly protective in animals but was linked to worse outcomes in patients. Other interventions, such as clot-busting drugs for stroke, agreed. The authors attributed the gaps to bias in the animal studies and to models that do not mimic human disease. [1] [Controlled study, review] Reviews of the field reach similar conclusions about study design, model choice and reporting. [5] [Review]

Biology differs between species

Receptors, metabolism and disease processes can differ between rodents and humans. In a genome-wide comparison, human responses to trauma, burns and sepsis were highly similar to each other, while the matching mouse models correlated poorly with the human conditions and with each other. [2] [Lab] Drug labels give concrete examples. Semaglutide causes thyroid C-cell tumors in rodents, but its label says human relevance "has not been determined". [6] [Regulatory] A rodent finding, good or bad, is a reason to study something in people, not a prediction of what will happen in people.

The dose-scaling myth

Online posts often convert a rat dose to a "human equivalent" and present it as a usable human dose. That is not what the conversion is for. FDA's guidance describes converting animal doses by body-surface area only as one step in finding a conservative maximum safe starting dose for a first-in-human trial in healthy volunteers. A safety factor is then applied to protect those volunteers. [7] [Regulatory] The method does not predict an effective dose, and it does not replace human trials. This page gives no dose conversions.

Publication bias and small studies

Studies with positive results are more likely to be published. In a review of 525 animal stroke publications, only about 2% reported no significant effect on the main outcome. Statistical adjustment suggested publication bias accounted for around one-third of reported efficacy. [3] [Lab, review] Many peptide claims rest on a few small rodent studies, often from one lab, so expect the true effect to be smaller than reported, or absent.

Questions to ask of any preclinical claim

Was it cells, animals or people? A headline rarely says.
Was the study randomized and blinded, and how many animals were used? [1]
Has an independent group replicated it?
Is the dose, route and duration remotely like what people use? (This page does not give doses.)
Has anyone tested it in a controlled human trial? Search ClinicalTrials.gov and PubMed.
Who funded it, and would a negative result have been published? [3]

How this library grades preclinical-only compounds

Every compound in the Evidence Library is graded on published human clinical evidence: Strong, Moderate, Limited, Very limited or No human data. Compounds with only rodent or cell-culture papers carry a No human data badge, because those papers do not prove human safety or efficacy. [4] [Policy] Under a standing editorial rule, these pages still explain what lab and animal work suggests, say plainly that no human treatment studies exist yet, and flag the question as worth future study. [Policy]

Evidence levels used on this page

[Regulatory] FDA, USP or WADA primary document or an FDA-approved label. [Controlled study] randomized trial or systematic review of trials. [Observational] surveillance, pharmacovigilance or cohort data. [Lab] analytical, cell or animal work. [Guidance] professional-society guidance. [Policy] Open Peptide Project published policy.

Evidence tiers (FAQ), Immunogenicity and anti-drug antibodies, N-of-1 tracking and observational registries

References

Perel P, et al. Comparison of treatment effects between animal experiments and clinical trials: systematic review. BMJ. 2007. PMID 17175568.
Seok J, et al. Genomic responses in mouse models poorly mimic human inflammatory diseases. Proc Natl Acad Sci U S A. 2013. PMID 23401516.
Sena ES, et al. Publication bias in reports of animal stroke studies leads to major overstatement of efficacy. PLoS Biol. 2010. PMID 20361022.
Open Peptide Project. FAQ (evidence tiers), read Oct 27, 2026. https://openpeptideproject.org/faq/
van der Worp HB, et al. Can animal models of disease reliably inform human studies? PLoS Med. 2010. PMID 20361020.
WEGOVY (semaglutide) injection and tablets, prescribing information, Novo Nordisk (NDA 215256). DailyMed setid ee06186f-2aa3-4990-a760-757579d8f77b, published Jun 30, 2026.
US FDA. Guidance for industry: Estimating the maximum safe starting dose in initial clinical trials for therapeutics in adult healthy volunteers. July 2005. https://www.fda.gov/media/72309/download

Disclaimer: This page is for education only. It is not medical advice, and it does not endorse using, buying or preparing any unapproved product. Talk to a licensed clinician before you start, stop or change any treatment.