Search for the side effects of any research peptide and the results divide into two kinds: seller pages listing mild, transient effects, and warning pages listing serious ones. Neither cites a dataset, because until you go looking it is not obvious that there barely is one.
This page reports what FDA's adverse event database actually contains for these compounds, compound by compound, with the date it was queried. The result is a near-empty record, and the reason it is empty is more useful than the counts.
This is journalism about a public dataset. It is not medical advice, and nothing here describes what anyone should do.
What the database holds
FDA's adverse event reporting system collects reports of events that somebody associated with a drug. It is queryable through the openFDA interface. Queried on 13 September 2026 for reports naming each product:
| Compound | Reports |
|---|---|
| Thymosin alpha 1 | 37 |
| BPC-157 | 18 |
| Ipamorelin | 11 |
| Melanotan | 10 |
| CJC-1295 | 4 |
| GHK-Cu | 2 |
| Semax | 2 |
| TB-500 | 1 |
| Selank | 0 |
| Epitalon | 0 |
| MOTS-c | 0 |
| KPV | 0 |
| AOD-9604 | 0 |
| Elamipretide | 0 |
| Total, fourteen compounds | 85 |
And for comparison, in the same database on the same day:
| Approved drug | Reports |
|---|---|
| Semaglutide | 7,580 |
| Tirzepatide | 3,153 |
| Total, two drugs | 10,733 |
Two approved medicines produce roughly 126 reports for every one produced by fourteen research-market peptides combined. Six of the fourteen produce none at all.
The wrong conclusion, and why it is so tempting
The obvious reading is that research peptides are much safer than GLP-1 drugs. That reading is wrong, and understanding why is the point of the page.
FDA's system is passive. It does not go looking for harm; it receives reports that someone chooses or is required to file. Three groups do the filing, and every one of them is missing from the research-peptide market:
Manufacturers. Mandatory reporting duties attach to the holder of an approved product's marketing authorisation. A seller shipping research-labelled material holds no marketing authorisation, and therefore no duty. Nobody on the supply side is obliged to collect anything.
Clinicians. A prescribed drug has a prescriber — someone who knows what the patient took, sees them afterwards, recognises an event as possibly drug-related, and knows the reporting route exists. Somebody who bought a vial online has no prescriber. If they see a doctor at all, they may not mention it.
Patients who can name the product. Filing a report requires naming what was taken. A buyer of research material frequently cannot: the vial's contents are the seller's claim, not a verified fact, which is the whole subject of our guide to reading a certificate of analysis. "I injected something sold as BPC-157" is not a product name a surveillance system can index.
So the near-zero counts are a measurement of the reporting channel, not of the compound. Absence of reports is absence of a way to report. The error runs in one direction only — the system can undercount badly and cannot overcount — which means these numbers set no upper bound on anything.
The GLP-1 comparison makes the mechanism visible rather than the risk. Those 10,733 reports are what a functioning surveillance system attached to very wide prescribed use looks like. The 85 are what no system looks like.
The most common complaint about BPC-157 is that it did not work
BPC-157 is the most-searched compound in this market. Its 18 reports break down like this:
| Reaction term | Reports |
|---|---|
| Drug ineffective | 3 |
| Erythema | 3 |
| Abdominal discomfort | 2 |
| Balance disorder | 2 |
| Drug intolerance | 2 |
| Dyspnoea | 2 |
| Flushing | 2 |
| Headache | 2 |
| Hypersensitivity | 2 |
| Nausea | 2 |
The joint most frequent coded term is drug ineffective. Of the handful of people who went to the trouble of filing a formal report with a federal agency about this compound, the largest single group did so to say it had no effect.
The year distribution — 2 in 2021, 2 in 2022, 4 in 2023, none recorded in 2024, 3 in 2025, 7 in 2026 — shows counts rising, with 2026 the highest. On a base this small that most plausibly tracks growing use and growing awareness that a reporting route exists, rather than any change in the compound.
Thymosin alpha 1's record has the same character. Its top term is off label use, at 8 of 37 reports — again not a symptom, and again a fact about how the compound reaches people. Its clinical file is set out on our thymosin alpha 1 dosage page.
What is actually known, compound by compound
Building the dosage cluster on this site meant opening the human literature for each compound in turn, and the same pattern recurred:
- Most have no human safety data because they have almost no human studies. Epitalon returns zero registered trials; DSIP returns zero despite 407 papers; IGF-1 LR3's 44 indexed papers include not one human study.
- Where human studies exist they are small and short. The longest exposure in the DSIP sleep literature is seven consecutive nights. The entire clinical experience section of the FORZINITY label — the approved elamipretide product, covered on our SS-31 dosage page — rests on twelve patients.
- Injection-site reactions are the one consistent finding. Across the compounds with any trial record, local irritation at the injection site is what gets reported. It is also the least informative possible finding, since it says nothing about systemic effects.
- The approved compounds in this catalogue have real safety sections, and they are specific. The FORZINITY label carries a warning against use in neonates because of benzyl alcohol. The VYLEESI label for bremelanotide carries limitations of use excluding men. These exist because a regulator required them, and their absence elsewhere is not reassurance.
What FDA has said, and what it reasoned from
On 29 September 2023, FDA placed GHRP-2, GHRP-6, ipamorelin acetate and kisspeptin-10 in category 2 of its compounding lists — substances identified as presenting significant safety risks. The stated concerns centre on immunogenicity together with insufficient safety data, and for the secretagogues, endocrine effects. The whole record is indexed on our page covering FDA's category 2 list.
Immunogenicity is the risk that the body raises an immune response against an injected peptide. It matters twice over: the response can neutralise the substance, and it can cross-react with the body's own version of a similar molecule.
Note what that reasoning is not based on. There was no adverse event record to reason from — ipamorelin's eleven reports and kisspeptin's none could not support any conclusion. FDA reasoned from structure and from the absence of data, which is what a regulator does when the evidence it would normally use does not exist.
It also describes exactly the kind of harm this dataset would never capture. An immune response developing over months, producing fatigue or an endocrine shift nobody connects to a vial bought two years earlier, generates no report from anyone.
What this page cannot tell you
It cannot give a side-effect profile for any of these compounds, because no dataset capable of producing one exists.
It cannot tell you that a compound with zero reports is dangerous. The argument here is symmetrical: these counts support no conclusion in either direction. A zero can even sit beside a substantial clinical literature — tesofensine's trial record runs to thirteen registered studies and publishes a dose-dependent heart-rate effect at every step, and it holds no US adverse-event record at all, because it holds no US approval.
It cannot substitute for a clinician. Anyone experiencing a symptom they associate with anything they have taken needs a person who can examine them, and this is a publication.
What it can do is replace a guess with a number, and name the reason the number is unusable. A seller citing the absence of reported side effects is citing the absence of a reporting system, and now you can check that yourself.
How to check this in five minutes
The openFDA adverse event endpoint is public and needs no key. Query the drug/event.json endpoint with search=patient.drug.medicinalproduct:"BPC-157" and read meta.results.total. Swap the product name for any compound above. Add count=patient.reaction.reactionmeddrapt.exact to see the coded terms, or count=receivedate for the distribution over time.
Run it for semaglutide alongside, in the same session, and the contrast in the two totals is the entire argument on this page.
Related on this site
- FDA's category 2 list, indexed — what FDA has actually restricted, and on what grounds
- How to read a dosing claim — separating a sourced figure from a repeated one
- How to read a certificate of analysis — why naming what was in a vial is harder than it sounds
- What the rules actually say about peptides — the regulatory position of research-labelled material
- FDA warning letters to peptide sellers — the enforcement record, letter by letter
Sources and dates
- openFDA drug adverse event endpoint (
api.fda.gov/drug/event.json), total counts bypatient.drug.medicinalproductand reaction-term and receive-date frequency counts. All queries run 13 September 2026. Product-name matching is against free-text reported product names, so counts index reports naming that string. - FDA, Certain Bulk Drug Substances for Use in Compounding That May Present Significant Safety Risks. Content current as of 22 April 2026; read 13 September 2026.
- Ancell CD, Phipps J, Young L. Thymosin alpha-1. Am J Health Syst Pharm. 2001;58(10):879–85. PMID 11381492, for the injection-site tolerability finding.
- Schneider-Helmert D. Effects of delta-sleep-inducing peptide on 24-hour sleep-wake behaviour in severe chronic insomnia. Eur Neurol. 1987;27(2):120–9. PMID 3622582, for the seven-night exposure.
- DailyMed SPL records for FORZINITY (elamipretide) and VYLEESI (bremelanotide), for the labelled warnings and limitations cited. Previously read 11 September 2026 and 8 September 2026 respectively for this site's dosage pages.
