How we rate peptide safety

Peptide writing online tends to fail in one of two directions: uncritical enthusiasm, or a blanket warning that treats an approved drug and an unstudied research chemical as the same thing. Both are useless to someone trying to decide. This page sets out exactly how each entry is built, so you can check our reasoning rather than trust it.

What each verdict means

The verdict answers one question: where does this peptide stand with regulators and with the human evidence? It is a classification, not a recommendation — with one deliberate exception, noted below.

Regulatory status as of October 2026. Every verdict rests on the regulatory record on that date, and each entry shows when it was last checked.

Approved Approved

Approved by the FDA and/or Health Canada for at least one human indication. Safety has been characterised in clinical trials and is described on an official product label. Approved does not mean safe for everyone, or safe outside the indication it was approved for.

Investigational Investigational, not approved

Unapproved, with substantial published trial data. No regulator has authorised it for any human use, so legally it sits exactly where an unapproved compound sits — but unlike most unapproved peptides it has been through registered human trials that have reported, so there is a real body of evidence on what it does and what goes wrong. What is missing is the long-term and post-approval record: the trials that surface uncommon or delayed harms have not finished, and nothing has been reviewed by a regulator.

Unapproved Unapproved

Not approved for human use by the FDA or Health Canada. The human evidence is limited, animal-only, or absent. Material sold outside regulated channels also has no verified identity, purity or stability, which is a separate risk from the molecule itself.

Avoid Avoid

Carries a documented serious safety signal, an explicit regulator warning, or published reports of serious harm. This is the one verdict on the scale that is a recommendation, and it is only used when a named authority has said so in writing.

Avoid is the only verdict that carries a recommendation, and we use it only when a named regulator has published a warning or the literature contains reports of serious harm. We do not use it to express unease about a compound that simply has not been studied — that is what Unapproved is for.

Approval is also not a safety certificate. An approved peptide has a known risk profile for the population and indication it was tested in; used outside that, it is back in untested territory, which is why several entries carry an approved verdict and a thin evidence level at the same time.

The four evidence levels

The evidence level is assigned separately from the verdict, and it describes the claimed uses rather than the risks. It answers: how well has this been tested in humans?

Strong

Multiple randomised controlled trials in humans, or a systematic review of them, reporting consistent results on an outcome that matters.

Limited

Some human data — small trials, single trials, open-label or observational studies — enough to describe an effect but not enough to settle it.

Animal-only

Studied in animals or in cell culture, with no completed trial in humans. Animal results routinely fail to reproduce in people, so this level supports no human conclusion at all.

None

No published studies we can find that test the claim, in humans or animals. The claim is unverified rather than disproved.

Two rules govern how the level is set. First, animal work never raises the level above animal-only, no matter how many studies there are or how striking the results — rodent findings fail to reproduce in humans often enough that treating them as weak human evidence would be misleading. Second, an absence of reported harm is never counted as evidence of safety; for most unapproved peptides, nothing has been reported because nothing has been measured.

What counts as a source

Every factual statement traces to a published regulatory record or to peer-reviewed literature. We accept citations from a fixed allowlist of publishers, and the build fails if a link anywhere on the site points outside it:

Regulators and official labels

• DailyMed (NIH)
• European Medicines Agency
• FDA
• FDA (Drugs@FDA)
• Health Canada
• Health Canada (Drug Product Database)
• Health Canada (Recalls & Safety Alerts)
• MHRA / UK Government

Peer-reviewed literature and trial registries

• ClinicalTrials.gov
• Cochrane Library
• NCBI
• PubMed
• PubMed Central

Public-health and anti-doping reference bodies

• MedlinePlus (NIH)
• NIH
• US Anti-Doping Agency
• World Anti-Doping Agency
• World Health Organization

Trial sponsors, for their own unpublished results

• Eli Lilly (sponsor)

A sponsor reporting its own phase 3 topline is often the only source that exists until peer review and regulatory review catch up. We cite it rather than drop the result or state it unsourced — and the entry says on its face that the figures are company-reported. Never accepted for a safety conclusion a regulator has also published on, and never counts toward the minimum below.

Secondary reporting, for dating a regulatory event

• ABC News (secondary)
• AJMC (secondary)

Advisory-committee votes and draft policies are frequently reported before the agency publishes a citable record. These sources may establish that something happened and when — never a medical claim, and never the regulatory position itself once a primary record exists. Also counts toward no minimum.

Every published entry must cite at least one regulator source. It must also cite a research source whenever we claim human evidence exists — that is, whenever the evidence level is strong or limited. Atanimal-only or none the finding is the absence of human data, and what attests to that is the regulatory record, so no trial citation is demanded; requiring one would block exactly the entries most worth publishing. Both minima are checked at build time, not by hand.

We link to top-level, long-lived pages where we can, because deep links into regulatory databases rot quickly and a dead citation is worse than a general one.

We do not cite forums, social posts, personal accounts, clinic marketing or anyone offering the compound being discussed — not as evidence, and not as colour. The two qualified groups above are the only exceptions to a primary-source-only rule, each is labelled as such on the entry that uses it, and neither can satisfy a minimum. Where the only available information about a peptide comes from places we will not cite, the honest answer is that the evidence level is none, and that is what the entry says.

What we will not publish

Five rules are enforced mechanically by a check that runs before every build. If any page or data entry breaks one, the build fails and the site does not deploy. They exist because a safety reference that doubles as a how-to guide is not a safety reference.

• No amounts, schedules or administration instructions. Not for unapproved compounds, and not for approved drugs either — that belongs on the official label and with the prescriber. Where a route of administration bears on risk, we describe it and link the label.
• No purchasing information. No storefronts, no marketplaces, no sourcing advice, no pricing, no commentary on quality by seller.
• No affiliate, referral or commercial links. The site earns nothing from any entry, which is enforced by the allowlist above rather than promised.
• No personal accounts or quoted experiences. An individual report tells you what one person believes happened; it cannot separate a drug effect from everything else going on. Published case reports are different and are cited as literature.
• No unsourced medical claims, and no absolutes. Nothing on this site is described as safe, risk-free, proven or guaranteed. Claims are attributed and qualified, or they are not made.

How entries are kept current

Every entry shows the date it was last checked against its sources, and that date drives the sitemap rather than the build timestamp — so a date on this site means someone looked, not that a deploy ran.

• Approved and investigational entries are rechecked when a regulator updates a label, issues a safety communication, or a registered trial reports — and reviewed at least twice a year regardless.
• Unapproved entries are rechecked at least annually, and sooner if a human trial is registered or published, which is the event most likely to change both the verdict and the evidence level.
• Avoid entries are rechecked whenever a new regulator warning appears.
• A verdict change is never silent: the summary says what changed and the entry's date moves.

Corrections are welcome and we would rather hear about them than not — if an entry misreads a source or misses one, write to corrections@ispeptidesafe.comwith the citation, and please include a link to the primary record.

Current coverage

16 peptides have entries — 6 written and 10 still in progress. Entries being written say so on the page rather than presenting an outline as an answer, and they stay out of the sitemap until the content is in.