Semax: FDA's Own Reviewers Said No, and the Committee Voted Yes Anyway

On 11 May 2026, FDA's reviewers proposed not adding Semax to the list of bulk substances that compounding pharmacies may lawfully use. On 24 July 2026, the advisory committee considering that document voted 8 to 5, with one abstention, to add it anyway.

The better question is why the reviewers had so little to work with.

What the committee actually voted on

Not approval. This was the Pharmacy Compounding Advisory Committee, meeting 23 and 24 July 2026. The question was whether Semax belongs on the 503A Bulks List, the raw substances a compounding pharmacy may lawfully work with.

Two Semax votes were scheduled, not one. FDA treats semax free base and semax acetate as separate bulk drug substances requiring separate evaluation and separate votes, while noting that both share the same active moiety, semax free base.

The tally is reported from news coverage rather than from the primary record, because FDA has posted no minutes and no transcript. Reuters gave eight in favour, five against, one abstention. RAPS gave 8 to 5, with the indications attached: cerebral ischemia, migraine and trigeminal neuralgia. Neither says which of the two votes it describes, and the official roster lists fifteen eligible voters, one more than the tally accounts for.

What the vote does not do. It is not an approval, it is not binding, and it does not change Semax's unapproved status. Listing requires notice and comment rulemaking, and even success would bring no prescribing information and no restrictions on promoting unapproved uses.

Why the review had almost nothing in it

FDA identified the Russian stroke literature by name and then excluded all of it, because no verified English translation had been submitted, citing 21 CFR 10.20(c)(2). The excluded papers were Miasoedova 1999, Gusev 1997, Gusev 2005 and Gusev 2018, essentially the entire human efficacy base for the stroke indication.

What remained for cerebral ischemia was one conference abstract with no full published article behind it, describing an unknown number of subjects.

The other surviving reference came from the nominators themselves. Koroleva 1996 gave a single 0.5 mg/kg intranasal dose to 37 adults, unblinded and uncontrolled: 12 with migraine, 16 with typical trigeminal neuralgia, 9 with dental plexalgia. In the trigeminal neuralgia group there were no changes in pain characteristics or evoked potentials. In the migraine arm, 4 of 12 reported headache cessation 90 to 120 minutes after dosing, the other 8 weaker but non-zero effects. FDA noted no baseline characterisation, no blinding, no control group, and no standard deviations for any score.

The paper's own conclusion is that semax does not exhibit analgesic activity by itself.

A nomination is an argument and its references are the exhibits. This one filed the letter of recommendation that spends its middle paragraph on the candidate's absences. FDA's verdict was two usable references, both inadequate, for three serious conditions that already have approved therapies, none of whose professional guidelines mention Semax at all.

The Russian literature, handled honestly

The exclusion was procedural. Nothing in the FDA document evaluates the content of those papers. Anyone who tells you FDA reviewed the Russian data and found it wanting is describing something that did not happen. Anyone who tells you that data settles the question is going further than the Russian authors did.

Gusev 1997, the founding stroke study. 12 to 18 mg daily for 5 to 10 days, 30 patients against 80 controls, published in Russian, non randomised. (PMID 11517472) Both figures fall inside the daily dose range the modern 1 percent label now specifies for stroke.

Gusev 2018, the largest and most cited. 110 patients, two 10 day courses at 6000 mcg per day, measuring plasma BDNF, a motor scale and the Barthel index. (PMID 29798983) PubMed indexes it as a Clinical Trial. It is described as comparative and non randomised, without placebo control, in secondary summaries rather than at the primary paper.

The meta-analysis is the most candid document in the file. Shmonin and colleagues found 8 nominally suitable studies covering 654 patients, and could pool only 3, because the rest did not share endpoints. They reported benefits on NIHSS, Rankin and Rivermead. Then they recommended that a proper trial still be run. That was 2018, after two decades of Russian clinical use. The people best placed to defend this drug said the definitive trial had not been done.

Outside Russia there is no clinical replication I could find, and the two English language papers usually cited as replications are commentary and hypothesis.

Three molecules, one name

This is where buyers get hurt, and it is arithmetic rather than pharmacology.

Semax. H-Met-Glu-His-Phe-Pro-Gly-Pro-OH, or MEHFPGP. C37H51N9O10S, average molecular weight 813.9, CAS 80714-61-0. It is the ACTH fragment from residue 4 to residue 7, extended at the tail with the stabilising tripeptide Pro-Gly-Pro. It is loosely called an analogue of the fragment from residue 4 to residue 10, which describes a substitution rather than a shared sequence, since native ACTH there reads MEHFRWG.

N-Acetyl Semax. Ac-Met-Glu-His-Phe-Pro-Gly-Pro-OH. Acetylated at the head, tail still a free acid. C39H53N9O11S, 856.0. The acetyl group adds 42.04 daltons and nothing else.

N-Acetyl Semax Amidate. Ac-Met-Glu-His-Phe-Pro-Gly-Pro-NH2. Acetylated at the head, amidated at the tail. C39H54N10O10S, 855.0, CAS 2920938-90-3.

Now the trap. Amidation swaps one oxygen, 15.995, for an NH, 15.011, a net loss of 0.984 daltons. The two acetylated forms sit about 1 dalton apart. The parent sits 42 daltons below both.

Mass spectrometry across these three is a ruler with fine gradations at one end and a blank stretch in the middle. It tells you instantly that you are not holding plain Semax, because 42 daltons is enormous. It cannot tell you which of the two acetylated forms you do have, because a certificate reporting nominal or low resolution mass will not resolve one dalton. You can prove what it is not. You cannot prove what it is. Our guide on how to read a certificate of analysis covers what a genuine match looks like.

The naming is worse than the chemistry. Vendors and at least one chemical database assign the same CAS number, 2920938-90-3, to products labelled "N-Acetyl Semax" and to products labelled "N-Acetyl Semax Amidate." PubChem assigns that CAS to the amidate, so a product sold as N-Acetyl Semax carrying it is labelled, by its own identifier, as something else. FDA hit the same failure in a different pair: both nomination packages carried a certificate titled "Semax Acetate" that identified the free base.

Nothing is published on the derivatives. I could find no PubMed indexed study of N-Acetyl Semax Amidate as a compound, in vivo or in humans. The one published study treating N-terminal acetylation as a distinct compound found that acetylation abolished a protective effect of the parent: Ac-Semax failed to protect neuroblastoma cells from copper induced toxicity, attributed to loss of the free N-terminal amine. (PMID 27586814) The opposite direction from the marketing.

What the mechanism rests on

The BDNF story is real, small, and Russian.

Dolotov and colleagues gave rats intranasal Semax at 50 and 250 mcg/kg and measured BDNF protein three hours later. It rose in the basal forebrain and not in the cerebellum, regionally selective rather than global. (PMID 16635254) The fullest paper reports rat hippocampus after a single 50 mcg/kg dose: BDNF protein up 1.4 fold, trkB phosphorylation up 1.6 fold, exon III BDNF messenger RNA up 3 fold, trkB messenger RNA up 2 fold. (PMID 16996037)

Notice the split. Protein moved by less than half. Transcript moved by two and three times. That distinction collapses in marketing copy, hardest around the number everyone quotes: an eight fold rise in BDNF messenger RNA, measured in rat glial cell cultures and peaking at 30 minutes. (PMID 11457573) An in vitro transcript reading, sold as an in vivo protein effect.

Every primary BDNF paper I could locate traces to one Moscow institution, the Institute of Molecular Genetics of the Russian Academy of Sciences. And in humans, nobody has measured brain BDNF at all. The only human BDNF measurement in the record is plasma BDNF in Gusev 2018, and plasma BDNF is overwhelmingly platelet derived, a poor proxy for what is happening centrally.

FDA's own read was that the mechanism is not established, and that Semax does several unrelated things in rodents. The strongest independent work points elsewhere entirely, to copper chelation and to opioid receptor signalling. (PMID 40692165)

The route nobody studied, and where the doses come from

Every approved Semax product in the world is nasal drops. Peptogen markets two in Russia, 0.1 percent at 50 mcg per drop and 1 percent at 500 mcg per drop. There is no approved injectable anywhere, no monograph in the European, Japanese or International Pharmacopoeias and none in USP or NF, and FDA states its foreign recognition is limited to Russia. The widely repeated line about approval "in Russia and Ukraine" appeared only on vendor pages in my searches.

FDA's finding on route is flat. It was unable to find literature discussing subcutaneous administration; the only route discussed anywhere was intranasal. The American market runs on the other one. FDA documented sites offering a 2.5 mg/mL injectable and 30 mg vials sold "for research purposes only" listing subcutaneous injection as a common route.

Subcutaneous Semax has zero published human literature of any kind. No efficacy, no safety, no pharmacokinetics. The gap between the community route and the published route is total, not partial.

1

2

3

4

1,000

10,000

mcg per day, intranasal · log scale

derived from researchcommunity convention

1. 1Community convention · 500 to 3,000 mcg per day, intranasalRoughly 250 to 1000 mcg per administration, one to three times daily.
2. 2Russian 0.1 percent label · 800 to 8,000 mcg per day, intranasalSingle dose 200 to 2000 mcg, four times daily for 10 to 14 days.
3. 3Russian 1 percent label, moderate stroke · 6,000 to 12,000 mcg per day, intranasal2 to 3 drops per nostril, 3 to 4 times daily, for 10 days.
4. 4Russian 1 percent label, severe stroke · 12,000 to 20,000 mcg per day, intranasal3 to 4 drops per nostril, 4 to 5 times daily. The highest labelled dose anywhere.

For once the community number is not the outlier. The convention that circulates sits at the low end of what the Russian label already specifies, and it remains vendor and forum consensus rather than a regimen any trial produced. The 0.1 percent label also contraindicates it in acute psychotic states, disorders accompanied by anxiety, and a history of seizures.

One figure will not reconcile. FDA's summary of published human doses includes "600 mg daily for 10 days," which is 30 times the highest labelled Russian daily dose. It comes from the Cherkasova 2002 meeting abstract, behind which no full paper exists. I report it as printed and flag it rather than quietly correcting it.

Cycling is pure convention, and reported practice varies wildly: continuous daily use, 5 days on and 2 days off said to prevent irritability, 3 weeks on with long off periods. No study has compared any schedule, and among the human studies FDA located, the longest adult course was 10 days.

What nobody knows

What it does in people under controlled conditions, in a language the reviewing agency was permitted to read. That is the shape of the gap, and it is not the same thing as an absence of data.

What the two safety signals mean. Semax at 0.15 or 0.6 mg/kg intraperitoneally potentiated amphetamine induced striatal dopamine release in mice, which FDA flagged as an abuse potential concern. FDA drew a second, antithrombotic concern from the same Cherkasova 2002 abstract that was its only cerebral ischemia effectiveness reference. The document carrying the strongest effectiveness signal also carries a safety signal.

Regulatory status

WADA. Semax does not appear in the 2026 Prohibited List. I extracted all 26 pages and the string is absent.

The common assertion that class S0 catches it does not survive the literal text. S0 requires no current approval by any governmental regulatory health authority for human therapeutic use, and Semax holds a current Russian marketing authorisation for exactly that. The hook is contestable rather than automatic. S0 names BPC-157 explicitly. It does not name Semax.

The alternative hook, S2.2.2, covers corticotrophins and their releasing factors, and its examples all raise cortisol, while Semax is characterised in the literature as noncorticotropic and devoid of hormonal activity. (PMID 39442746) Whether that clause reaches it is interpretation rather than a published ruling, and I found no WADA, USADA or CAS decision naming Semax.

FDA. Unapproved for anything in the United States. The July 2026 vote does not change that.

My take

The thing worth carrying out of this is not that the evidence is weak. It is that the record was emptied by a rule, and then a vote was taken on what was left.

FDA's reviewers did their job correctly. The translation requirement exists for a reason, and an agency cannot evaluate papers it cannot verify. The result is still a formal federal assessment of a drug that never touched the four studies holding nearly all of its human data. A committee read that assessment, disagreed, and voted the substance onto the list anyway.

Both moves are defensible. Neither produces knowledge. Nobody translated the papers. Nobody ran the trial the Russian meta-analysts asked for in 2018. Nobody has published anything about the route Americans actually use, or about the two acetylated molecules that make up most of what is sold. July was a procedural event wearing the shape of a scientific one, and underneath it sit three compounds with nearly the same name, two separated by a single dalton, sold to people who were taught to check the mass.

If you have bought Semax in the past year, I would like to know which of the three names was on the vial, and whether the certificate named the same molecule as the CAS number printed beside it.

---

This content is for educational and informational purposes only and is not medical advice. Peptides discussed are research compounds and may not be approved for human use. Nothing here should be used to diagnose, treat, cure, or prevent any disease. Always consult a qualified healthcare professional before starting any peptide, supplement, or protocol. Individual responses vary. Do not self-administer compounds without proper medical supervision.