“Safe” isn’t a property. It’s a sample size.

This is the written version of Are Peptides Safe? The Question Nobody Answers Straight — watch it instead if you’d rather.

Every week somebody asks me the same four words. Are peptides safe? And every week I start typing an answer in a comment box and stop, because the honest version does not fit in one.

Nobody answers it straight, and it is not because everyone is hiding something. It is because in a regulator’s hands “safe” is not a verdict at all. It is an arithmetic result — and once you see the arithmetic, the shrugging starts to look like the only honest answer available.

What an approval actually proves

When a drug is approved, nobody has proven it is safe. What has been proven is something much narrower, and much more useful: we exposed this many people, for this long, and this is the smallest problem we would have been able to see.

If you want proof that this is really how it works, read the safety section of any drug label in your bathroom cabinet. Take tesamorelin — a peptide, an FDA-approved one, so about as good as this ever gets. Its Adverse Reactions section opens by telling you not to trust its own numbers: because trials are run under widely varying conditions, the rates observed cannot be directly compared and may not reflect the rates observed in practice.

That is what you get on an approved drug, with a named manufacturer, under a legal obligation to report harm. That is the good version. So when somebody tells you a peptide is safe, the only sensible follow-up is: compared to what, measured how, and in how many people? That question breaks apart into four separate ones, which is precisely why nobody answers it in a sentence.

Question one: is the molecule itself a problem?

This is the one everybody thinks they are asking. Does it cause tumours, wreck an organ, do something horrible at the doses people actually use?

There is a boring industrial answer to how you find that out. It is called a carcinogenicity study, and the international guideline is blunt about when you need one: if people will take something continuously for six months or more, you do the study, and you finish it before you apply to sell it. Not after.

Now look at what exists for the most-used repair peptide on the internet. There are well over a hundred papers indexed on it — more than plenty of approved drugs had when they went to market. Then filter that same search by publication type for controlled studies in humans, and watch what comes back.

A large literature is not the same thing as evidence about you. Almost all of it is animals, and an animal study is a hypothesis generator. It is not a safety file.

Be fair to it, though. Zero trials is not evidence of harm. Nobody has shown this stuff is dangerous either. Unmeasured is not the same as unsafe, and anyone who lets you walk away believing it is has done the same thing to you the sellers do, pointing the other way.

Free guide · goes with this article Peptides, Plainly

The free beginner’s guide: the chemistry, maths and evidence ideas everything else rests on, explained from scratch. Comes with the other three free guides, for one email.

Get the free guides

Question two: is that actually what is in the vial?

People skip this one because it feels like a different subject. It isn’t. A molecule with a spotless safety record is no use to you if it arrives as something else. Two examples changed how I think about it, and neither involves a dodgy website or a bloke in a garage. Both happened inside proper facilities, with proper paperwork, making the correct molecule.

• The first was a stopper. A widely used biologic began causing a rare and serious blood disorder in patients who had been fine on it for years, peaking in 2002. When it was taken apart, the molecule had not changed at all. A stabiliser had been swapped out and a different one swapped in, and material began leaching from uncoated rubber stoppers. The drug was still the drug. The stopper was the problem.
• The second was a clean room. In 2012 a compounded injectable made in an American facility was contaminated with fungus. The final count was 753 cases across 20 states and 64 deaths. The detail that ought to stop you cold is the timing: the median gap between injection and diagnosis was 47 days. Six and a half weeks of feeling completely fine.

Whether the molecule is safe and whether your vial is safe are two different questions. The second one is answered by a factory you have never seen, in a month you were not watching.

Nothing you can read off a label touches that. The long version of what a certificate of analysis can and cannot tell you about the vial in your hand is its own article.

Question three: does it collide with what else you take?

Here is the good news, and it is real. Most drug interactions happen in the liver, at a family of enzymes that chew up medicines. It is a shared bit of plant, and if two chemicals both want it, one backs up. The FDA’s own position is that peptides are not expected to significantly modulate those enzymes or the transporters that move drugs around the body. On that one axis they genuinely look low-risk, and pretending otherwise would be dishonest.

But the same guidance names one mechanism that does bite. Anything that slows your stomach down changes how fast every tablet you swallow gets absorbed. That is not a side effect. That is a process change sitting upstream of everything else you take. So question three is a partial yes: mostly quiet, one loud exception — and for the compounds most people are actually buying, not one study anybody has run.

Question four: what does five years of it do?

Not what one injection does. What a habit does.

This is where fialuridine belongs. It was trialled for hepatitis B in fifteen patients. The early weeks were fine. Bloods fine, patients fine, everybody fine. Then, weeks in, it turned, and the trial was stopped for severe multisystem toxicity. Five of the fifteen died.

The mechanism matters, because it was not hiding. It was mitochondrial — damage that accumulates. It could not appear early, because there had not been enough time for it to add up. A short study did not miss it. A short study is not built to see it.

Now scale that up. Somebody looked at 222 new medicines the FDA approved over a decade — full trials, full labels, full surveillance — and asked how many later picked up a new safety problem. The answer was 32%, and the median time to find it was 4.2 years after approval. That is the system working. It still needs four years and a functioning alarm to catch up with itself.

And that alarm is the part worth holding on to. Mandatory reporting sits on whoever holds the approval. No approval holder, no mandatory reporting. What is left is people volunteering — and across 37 studies, the median under-reporting rate to voluntary systems was 94%.

“Withdrawn” is not “cleared”

Someone always raises this, so let me raise it myself. The FDA took several of these compounds off its significant-safety-risks list. Doesn’t that mean they were cleared? It is a reasonable thing to think, because that is what the page looks like. But what it says is withdrawn, and withdrawn means the nomination was withdrawn. Somebody stopped asking. It does not mean the answer came back good. It means the form got pulled off the desk.

To be even-handed in the other direction: where is the ceiling? Probably thymosin alpha-1 — approved in a good number of countries, decades of use behind it, and a randomised trial in 361 patients with severe sepsis. By the standards of this corner of the internet, that is a serious file. When the FDA reviewed it, it still found insufficient information on its use and a lack of information about immunogenicity risks, and recommended against adding it. That is the best one in the space, and it is still a no.

The bottom line

Nobody can tell you a gray-market peptide is safe, and the reason is not conspiracy or cowardice. It is that the sentence would be arithmetically empty.

Here is the number. A guideline the FDA adopted says what a safety database for a long-term drug should look like: about 1,500 people exposed, 300 to 600 of them for six months, 100 of them for a full year. And it says why, in a sentence I think is quietly magnificent. A hundred patients for a year with nothing going wrong gives reasonable assurance that the true rate is no greater than 3%.

Read that again. A hundred people, a full year, everything fine, and the strongest thing a regulator will let you say is probably rarer than one in thirty-three. That is the floor — the least a regulator will accept.

So do the arithmetic on the studies these compounds actually have. A dozen people. A few weeks. Nothing went wrong. That bounds the risk at about one in four. Not one in a thousand — one in four. To rule out something that hits one person in a thousand, you would need around three thousand exposures.

“Is it safe” is a question about sample size, and here the sample size is somewhere between tiny and nobody counted. Which does not make it dangerous. It makes it unmeasured — and now you know how big that difference is.

Sources

1. ICH E1A, adopted by FDA — the safety-database guideline: about 1,500 exposed, 300 to 600 for six months, 100 for a full year, and what a clean year of 100 patients actually buys — https://www.fda.gov/media/71180/download
2. Hanley & Lippman-Hand, JAMA 1983, “If nothing goes wrong, is everything all right? Interpreting zero numerators” — the rule of three — https://pubmed.ncbi.nlm.nih.gov/6827763/
3. ICH S1A, adopted by FDA — when a carcinogenicity study is required, and when it has to be finished — https://www.fda.gov/media/71921/download
4. Tesamorelin prescribing information — the Adverse Reactions paragraph that warns you not to compare its own numbers — https://www.accessdata.fda.gov/drugsatfda_docs/label/2013/022505s004lbl.pdf
5. PubMed — run the search yourself, then filter publication type to “Clinical Trial” and look at what comes back — https://pubmed.ncbi.nlm.nih.gov/?term=BPC+157
6. Locatelli, Del Vecchio & Pozzoni, Peritoneal Dialysis International 2007 — the pure red-cell aplasia cluster, the stabiliser swap, and leachates from uncoated rubber stoppers — https://pubmed.ncbi.nlm.nih.gov/17556324/
7. Smith et al., New England Journal of Medicine 2013 — the contaminated compounded injectable, and the median 47 days from injection to diagnosis — https://doi.org/10.1056/NEJMoa1213978
8. CDC — the final case count for that outbreak: 753 cases across 20 states and 64 deaths — https://archive.cdc.gov/www_cdc_gov/hai/outbreaks/meningitis-map-large.html
9. FDA, Clinical Pharmacology Considerations for Peptide Drug Products, December 2023 — peptides and CYP enzymes, and the gastric-emptying exception — https://www.fda.gov/media/171901/download
10. Kleiner et al., Modern Pathology 1997 — the fialuridine trial, the five deaths among fifteen patients, and the mitochondrial mechanism — https://pubmed.ncbi.nlm.nih.gov/9071726/
11. Downing et al., JAMA 2017 — 222 novel therapeutics approved between 2001 and 2010, and how many picked up a new safety problem afterwards — https://doi.org/10.1001/jama.2017.5150
12. Hazell & Shakir, Drug Safety 2006 — the median under-reporting rate to voluntary adverse-event systems, across 37 studies — https://doi.org/10.2165/00002018-200629050-00003
13. 21 CFR 314.80 — postmarketing adverse-experience reporting, and whose obligation it is — https://www.ecfr.gov/current/title-21/chapter-I/subchapter-D/part-314/subpart-B/section-314.80
14. FDA — the bulk drug substances page, including the entries withdrawn by their nominators — https://www.fda.gov/drugs/human-drug-compounding/certain-bulk-drug-substances-use-compounding-may-present-significant-safety-risks
15. FDA Pharmacy Compounding Advisory Committee briefing, 4 December 2024 — the thymosin alpha-1 review, the 361-patient sepsis trial, and the conclusion — https://www.fda.gov/media/183892/download

Educational and research purposes only — not medical advice. Peptide Corner does not recommend any vendor, source, or dose. Keep safe, keep skeptical.