Disclosure: PeptideOnline earns commissions from some links on this page, at no extra cost to you. See our affiliate disclosure.
The short answer
To judge a peptide study, answer four questions before you read the conclusion: what was studied (cells in a dish, animals, or people), how many subjects there were, whether there was a comparison group, and who ran or paid for it. A positive result in a cell culture or a rat is a reason for further research, not evidence that a product works in people. Most claims for unapproved peptides rest on exactly that kind of early evidence.
PubMed is the free National Library of Medicine search engine for biomedical literature. It lists citations and abstracts; the PubMed user guide notes that search results do not include the full text of articles. That limitation is important, because the abstract is where marketing usually stops reading.
- Identify the study type first: in vitro, animal, case report, observational, or randomized controlled trial.
- Check the subjects: species, number, and whether they resemble the people the claim is aimed at.
- Look for a control group, blinding, and a clinically meaningful outcome.
- Read the funding and conflict-of-interest statements, and check whether anyone else has replicated the result.
The evidence ladder, bottom to top
| Study type | What it is | What it can tell you | What it can't |
|---|---|---|---|
| In vitro (cell or tissue) | Peptide applied to cells in a dish | Possible mechanisms worth testing | Whether it works or is safe in a living body |
| Animal (rodent, other) | Peptide given to animals, often in induced injury or disease models | Whether an effect appears in a whole organism | Human dose, human safety, human effect size |
| Case report / case series | Description of one or a few patients | That something happened in someone; useful for safety signals | Whether the peptide caused it; how often it happens |
| Observational study | People followed or compared without assigning treatment | Associations in real populations | Cause and effect, because users differ from non-users |
| Randomized controlled trial (RCT) | People randomly assigned to peptide or comparison, ideally blinded | Whether the peptide causes the effect, and how large it is | Results outside the population and duration studied |
| Systematic review / meta-analysis | Structured synthesis of multiple studies | The overall weight of evidence | Anything better than the studies it includes |
FDA-approved peptide drugs such as semaglutide and tirzepatide reached approval through large randomized trials. Many research-labeled peptides have mostly cell and animal data, with few or no controlled human trials. Our nootropic peptides evidence ladder applies this ranking to Semax, Selank and others, and recovery peptides evidence by tissue does it for the injury-recovery category.
Using PubMed's own tools
Article type filters
PubMed's left-hand filters include article types such as Clinical Trial and Review. The user guide explains that these filters rely on each citation's Publication Type data. Filtering a search like "BPC-157" by Randomized Controlled Trial is one of the fastest ways to see how much human trial evidence exists.
Species filter
The Humans and Other Animals filters limit results to human or animal studies. The guide notes they usually give precise results but can exclude some relevant papers, so treat them as a quick check, not a final answer.
MeSH terms and similar articles
PubMed indexes citations with MeSH, the National Library of Medicine's controlled vocabulary, and builds its "Similar articles" list by comparing words from the title, abstract and MeSH terms. Following similar articles from a strong human trial is a quick way to find related trials; following them from a rodent study mostly finds more rodent studies.
Read past the abstract
Abstracts compress. Use the full-text links on the record when available. In the methods section, look for: species and strain, number of subjects per group, how the peptide was given, duration, the comparison group, and how outcomes were measured.
Check for registered trials
Human trials are typically registered before they start. Searching ClinicalTrials.gov for a peptide shows trials that are planned, running, finished or stopped, including ones whose results were never published.
Ten questions to ask of any study
- Cells, animals or people? The single most important fact.
- How many? A handful of animals per group can produce striking results that don't replicate.
- Compared to what? No control group, no conclusion about cause.
- Randomized and blinded? Without these, expectation and selection bias creep in.
- What outcome? A lab marker or tissue measurement is not the same as a person feeling or functioning better.
- How was it given, and how much? Animal studies often use routes and amounts that don't translate to people; this is a reason not to treat study doses as instructions.
- How long? Short studies can't speak to long-term safety.
- Who funded it and who wrote it? Check the funding and conflict-of-interest statements.
- Has anyone else found the same thing? Results from one lab or one group of authors deserve more caution.
- What do the authors say the limitations are? The limitations paragraph is often more candid than the conclusion.
Red flags in the paper itself
- Retractions and corrections. PubMed records show linked retraction notices and errata. A retracted paper should not be cited as evidence, but marketing pages often keep doing so.
- Preprints. Papers posted before peer review can be valuable and fast, but they haven't been checked by independent reviewers yet. Treat them as preliminary.
- Journal quality. Some journals publish with little meaningful review. An unfamiliar journal is not automatically bad, but check whether it is indexed in MEDLINE and who publishes it.
- Conference abstracts. Short summaries presented at meetings often lack full methods and may never become full papers.
- Mismatch between results and conclusion. If the results section shows a small or non-significant effect but the conclusion is enthusiastic, trust the results.
- Surrogate endpoints only. A change in a blood marker or tissue score may or may not translate into anything a person would notice.
How marketing stretches the evidence
These patterns come up constantly in peptide marketing. None requires fabricating a study; each just stretches a real one.
- Species swap. "Shown to heal tendons" when the study was in rats with surgically injured tendons.
- Mechanism as outcome. "Boosts growth factors" from a cell culture is presented as if it means faster recovery in people.
- Review counted as proof. A narrative review summarizing animal work gets cited as if it were a human trial.
- Citation count as evidence. "Over 100 studies" when nearly all are from one lab or in animals.
- Old or foreign-language trials without detail. Claims rest on trials that are hard to access or evaluate. Our piece on epitalon's Russian trials and marketing is a case study.
- Dose conversion. Animal doses converted into human "protocols." Animal studies were not designed to set a human dose, and research-use-only products have no approved human dose.
Our broader guide to separating science from marketing on longevity peptides shows these patterns in action.
A worked approach, without the hype
Say a product page claims BPC-157 "repairs the gut." Search PubMed for BPC-157, apply the Humans filter and then the Randomized Controlled Trial filter, and see what remains. Then remove the filters and read the abstracts of the top results, noting species and study type for each. You will quickly see whether the claim rests on human trials or on animal models. For background on the compound itself, see our BPC-157 profile, and compare its claims against what your own search finds.
Do the same for any compound before you take a claim at face value. Keep a simple note: compound, claim, best evidence found (type and size), and any human safety data. That note is more useful than any product page.
For a recent example of a non-trial publication that is still useful, the February 2026 Annals of Pharmacotherapy editorial we discussed in our piece on compounded semaglutide add-ins catalogued products and reviewed the rationale for each ingredient. It is not a clinical trial, and it does not claim to be; reading it correctly means using it for what it is.
Frequently asked questions
What is the strongest type of evidence for a peptide?
Randomized controlled trials in people, and systematic reviews that combine several good trials. Cell and animal studies are early-stage evidence that a compound is worth testing, not proof it works in people.
Does PubMed have the full text of studies?
No. PubMed lists citations and abstracts. The record often links to full text from the publisher or another source when it is available.
How can I tell if a peptide study was done in animals?
Check the abstract and methods for the species, and use PubMed's Humans or Other Animals species filter. The filter is a quick check, since it can occasionally exclude relevant studies.
Why don't rat study results apply directly to people?
Animals differ in metabolism, physiology and lifespan, and animal studies often use injury models, routes and amounts that do not translate. Many effects seen in animals fail to appear in human trials.
Can I use doses from a peptide study?
No. Study doses describe what researchers used under controlled conditions, not what a person should do. Prescription dosing is set by a prescriber, and research-use-only peptides have no approved human dose.
Sources and primary references
- PubMed User Guide (National Library of Medicine): article type and species filters, full text
- ClinicalTrials.gov: registry of clinical studies
- Belcourt J, Sapowadia A, White CM. Annals of Pharmacotherapy editorial on compounded semaglutide and tirzepatide formulations (online February 14, 2026)
Regulatory status can change. PeptideOnline date-stamps regulatory summaries and links to primary agency sources so readers can verify the current position.
Medical disclaimer: This article is for educational and research-information purposes only. It does not provide diagnosis, prescribing, dosing, injection, or individualized treatment advice. It has not been reviewed by a clinician. FDA-approved drugs, compounded prescription products, and research-use chemicals are different product categories and should not be treated as interchangeable. See our medical disclaimer and editorial policy.