Dihexa: what the studies show, status and safety

Summary

Dihexa is a synthetic molecule derived from angiotensin IV, designed to strengthen a growth-factor signal in the brain. No registered study has ever given it to a person. Its phosphate form, fosgonimeton, was tested in 554 people with Alzheimer's disease and made no difference, after which development stopped. The paper that established the mechanism was retracted in April 2025 for falsified figures, and the first description of Dihexa carries an unresolved Notice of Concern.

Key findings at a glance

• The strongest human evidence is negative. LIFT-AD enrolled 554 people with mild to moderate Alzheimer's disease. After 26 weeks the main measure differed from placebo by -0.08 standard deviations (p = 0.70), and every secondary measure missed as well ^[1], ^[2].
• Dihexa itself has never been in a trial. Searches of the American and European trial registers return nothing ^[3]. The FDA states that it has identified no human exposure data for the substance, by any route ^[4].
• The paper explaining the mechanism was retracted. The journal withdrew it on 29 April 2025. An investigation had found figures 1B and 2A/C to contain falsified or fabricated data ^[5].
• The first description of Dihexa is under an unresolved Notice of Concern, issued in September 2021. Nearly every number that circulates about the molecule comes from that one paper ^[6], ^[7].
• The half-life of about two weeks is a figure from three rats. Measured in people, the same molecule cleared with a half-life of roughly 1.5 hours ^[6], ^[8].
• The developer measured it as unfit to swallow. After one hour, 0.1 percent survived in simulated intestinal fluid. Its maximum solubility in water is under 50 micrograms per millilitre ^[9].
• Sport bans it at all times, without naming it. Two catch-all clauses of the 2026 Prohibited List cover it, and the company behind the molecule paid four million dollars to settle allegations of data falsification ^[10], ^[11].

What it is

Dihexa is a small synthetic molecule, not a peptide in the strict sense. It carries only two real amino acids, tyrosine and isoleucine, capped at one end by a fatty acid and extended at the other by a building block the body does not use in proteins ^[12], ^[6]. The published work calls it a peptidomimetic or simply a small molecule ^[9].

Joseph Harding and John Wright's group at Washington State University built it from angiotensin IV, a signalling peptide of the brain ^[6]. An earlier paper had narrowed the active part down to three amino acids ^[13]. The laboratory spun out a company, first M3 Biotechnology, then Athira Pharma, and since January 2026 LeonaBio ^[14], ^[15].

Dihexa has no international name, only laboratory codes. Development moved early to its phosphate form, fosgonimeton, because Dihexa itself was too insoluble and too fragile in the gut to serve as a medicine ^[9], ^[16].

Quick facts

How it works

The proposed idea is that Dihexa strengthens a growth-factor signal the brain already uses, so that nerve cells build more connections. Whether it does so is unsettled, because the paper that made the biochemical case for it has been withdrawn.

• It may amplify hepatocyte growth factor. The proposal is that Dihexa binds this growth factor and helps two copies pair up. Their receptor MET would then fire at concentrations that are otherwise too low ^[18]. That is the retracted paper ^[5].
• Two further papers from the same group were retracted in April 2025, both about the same target system and the same chemical family ^[19].
• Which receptor angiotensin IV even acts on is disputed. One camp identifies it as an enzyme called insulin-regulated aminopeptidase. The authors of the Dihexa work conceded in 2015 that the identity was controversial, and no paper has settled it ^[20].
• An independent group found a signalling route without testing the binding. In mice bred to develop Alzheimer-like changes, Dihexa improved maze learning and calmed inflammation. A blocker of the PI3K enzyme cancelled those effects ^[21].
• In zebrafish larvae the effect depended on the growth factor, at least indirectly. A blocker of hepatocyte growth factor weakened the protection of sensory cells, as did blockers of three downstream enzymes ^[22]. Two authors of the retracted work are co-authors here.
• In cell culture it substitutes for the growth factor outright. Several independent laboratories use it in place of recombinant hepatocyte growth factor. They do so when turning human stem cells into liver-like cells ^[23], ^[24].

This sets Dihexa apart from the Russian brain peptides it is usually filed beside. Semax and Selank are short chains of natural amino acids with their own separate stories. Dihexa targets a receptor tyrosine kinase, and its parent molecule angiotensin IV worked best in animals when injected directly into the brain ^[25].

What the studies found

Six registered studies in people tested the phosphate form of this molecule. Only three of them asked whether it helped, and all three missed their main measure. The remaining three covered safety, how the body handles the substance, and an open extension that closed with its parent trial ^[3]. Dihexa itself has been studied only in animals and in cells, where the results run both ways ^[1], ^[14].

Alzheimer's disease

Phase 2/3 RCT LIFT-AD randomised 554 people with mild to moderate Alzheimer's disease to a daily injection under the skin or to placebo, for 26 weeks. The primary analysis covered the 287 participants who took no cholinesterase inhibitor alongside ^[1], ^[2].

It missed. The main measure combined thinking and daily function into a single standardised score, so it is counted in standard deviations rather than in points. It differed from placebo by -0.08, with a standard error of 0.10 (p = 0.70). Memory testing differed by -0.70 points (standard error 0.77, p = 0.35) and daily function by 0.67 points (standard error 0.92, p = 0.61) ^[1].

All three differences were small, and all three leaned the same way, nominally towards the drug. None of them could be told apart from chance, and the authors record that the trial achieved neither its primary nor its secondary endpoints ^[1].

Only one blood marker of Alzheimer's disease moved, phosphorylated tau 217, and only nominally ^[1]. Development stopped in September 2024, and the open extension study of 423 participants was terminated with the parent trial ^[1], ^[26].

Earlier trials of the same molecule

Phase 2 RCT ACT-AD tested 77 people with Alzheimer's disease and missed its main measure, the timing of a brain-wave response called P300 ^[27]. Later analyses suggested that cholinesterase inhibitors may have interfered, which is a hypothesis rather than a finding ^[14].

Phase 2 RCT SHAPE enrolled 28 people with Parkinson's dementia or Lewy body dementia and was stopped early. The register gives the reason as study design limitations, and the main measure was missed ^[28].

Phase 1 The first human study covered 88 people ^[8], ^[29]. They were 48 healthy young men given a single dose, plus 29 healthy older volunteers and 11 patients with Alzheimer's disease, the latter two groups dosed daily for nine days. It holds the only human figures for Dihexa itself, which appears in the blood as the active form of the injected prodrug.

Exposure rose in step with the dose, nothing accumulated over nine days, and both peak levels and total exposure of the active form ran three to twelve times above those of the prodrug ^[8]. An exploratory brain-wave measure shifted towards normal in the patients, in a comparison of 7 people on the drug against 4 on placebo (p = 0.027). Those group sizes are far too small to carry weight ^[8].

Memory in rodents

Animal data In the first description, rats given a drug that blocks memory formation recovered in a water maze after Dihexa, whether it was injected into the brain, into the belly cavity or given by stomach tube ^[6]. Twenty-four-month-old rats also learned the same maze faster (p < 0.03 by Mann-Whitney U test), though only six animals per group were tested, three male and three female ^[6].

Animal data An independent group in Nanjing treated Alzheimer-model mice for three months. Learning improved, more nerve cells survived, and inflammatory messengers fell while a calming one rose ^[21]. The paper gives no group sizes and names two different injection routes in the same passage.

A blinded animal study that found nothing

Animal data The cleanest animal study of this molecule is negative. Forty male Wistar rats received a toxin that mimics Huntington's disease, and the treated group did no better on any measure ^[30].

Body weight rose least in the treated group, 13.31 percent below the healthy controls (p = 0.0078) and 5.72 percent below the untreated toxin group (p = 0.3486). Maze latency changed by 1.78 seconds (p = 0.7941). Against the untreated toxin group, every p value lay between 0.34 and 0.94 ^[30].

The work was blinded, the group size was calculated in advance, and the dose was taken straight from the earlier papers as their most effective one ^[30].

Cells in a dish

In vitro Hippocampal nerve cells grown with Dihexa at one picomole per litre for five days carried nearly three times as many dendritic spines, the small protrusions on which synapses sit: 41 against 15 per 50 micrometres of dendrite with vehicle, from 200 dendritic segments (p < 0.001) ^[6]. A second, much shorter exposure read on day 12 in culture gave a far narrower gap, 23.9 against 17.4 spines over the same length, from 60 measurements (p < 0.0001) ^[6]. Such effects at so low a concentration were one reason for later doubts.

In vitro The reproducible use of Dihexa is as a laboratory reagent. Independent groups in Edinburgh and Beijing employ it instead of recombinant hepatocyte growth factor. The resulting liver-like cells store glycogen and make serum proteins ^[23], ^[24].

What is still unknown

• What Dihexa does in a person. No registered study has ever administered it ^[3], ^[4].
• How it behaves when swallowed. The only human measurements follow an injection of the prodrug under the skin ^[8], ^[9].
• Whether it causes cancer. No carcinogenicity study, no long-term toxicology, no reproductive or genetic toxicity data exist ^[4], ^[31].
• Which receptor it works through. The identity has been disputed for more than a decade ^[20].
• Why the rat and human half-lives differ two-hundredfold. No paper has reconciled them ^[6], ^[8].
• Whether the rodent memory findings hold. They rest largely on one paper under a Notice of Concern, and the one blinded replication attempt in another model failed ^[7], ^[30].

Side effects and safety

There is no safety record for Dihexa itself. PubMed held no case report and no adverse event report for the substance on 15 September 2026, and the FDA states that it has identified no human exposure data at all ^[32], ^[4]. What exists comes from the prodrug, given by injection rather than by mouth.

• Injection site reactions were the commonest event over 26 weeks. They affected 128 of 224 on the lower dose and 78 of 107 on the higher, against 31 of 218 on placebo ^[2].
• Eosinophilia, a rise in one type of white blood cell, appeared in 16 of 224 and 8 of 107, and in none of the 218 on placebo. The signal rose with the dose ^[2].
• Withdrawals from side effects reached 14.2 percent against 4.6 percent on placebo, mostly because of the injection sites ^[1].
• Serious events were not raised: 11 of 224 and 3 of 107, against 15 of 218 on placebo, and no deaths in any group ^[2].
• In the 88-person phase 1 study no serious events occurred and no maximum tolerated dose was reached. Two people stopped, one with moderate neutropenia and one with a skin allergy, both judged unlikely to be related ^[8].

Two points deserve more than a bullet.

The pathway itself is an oncogene. MET drives tumour growth when it is amplified, mutated or fused ^[31]. Its amplification is an established route by which lung cancers escape targeted treatment ^[33]. Dihexa is meant to push that same pathway harder.

The contradiction is sharpest inside the originating laboratory. The same group developed closely related molecules expressly to block this growth factor and treat pancreatic cancer ^[34]. Those papers have since been retracted ^[19]. No carcinogenicity study of Dihexa exists, and 26 weeks of observation in LIFT-AD says nothing about cancer risk ^[4], ^[2].

Nothing about it dissolves or survives normally. Maximum solubility in water is under 50 micrograms per millilitre, and animal work required pure dimethyl sulfoxide as the carrier ^[9]. After one hour in simulated digestive fluid, almost nothing is left.

No contraindications and no interaction warnings are on record, because no authority has ever reviewed a product containing it ^[4].

Doses used in studies

The doses below are simply what the cited studies gave, listed so that the results can be understood. They are not advice on how to use anything, and they do not translate to human use.

Every line comes from an animal or from cells in a dish, and no human dose for Dihexa exists. Most lines trace back to the 2013 paper that carries a Notice of Concern ^[7].

Animal doses in milligrams per kilogram cannot be scaled to people. The human trials dosed the prodrug, which is a separate molecule with a failed programme, so their dose arms are not Dihexa doses and are left out of this table ^[1], ^[28].

Development and approval status

Dihexa is approved nowhere, was never in a registered trial under its own name, and the programme built on it has been shut down ^[3], ^[36].

• 2011 — the procognitive part of angiotensin IV is narrowed to three amino acids ^[13].
• 2013 — Dihexa is described in rats and cell culture ^[6].
• 2017 to 2019 — the phosphate form enters human testing in 88 volunteers and patients ^[8].
• 2021 — four papers from the originating laboratory receive Notices of Concern ^[37].
• September 2024 — LIFT-AD fails and the programme is discontinued ^[36].
• January 2025 — the company settles data falsification allegations for four million dollars ^[11].
• April 2025 — the paper behind the mechanism is retracted ^[5].
• January 2026 — the company becomes LeonaBio and works on other targets ^[15].

Every register query behind this table was run on 15 September 2026. Dihexa is not a controlled substance in the United States ^[46].

The only assessment any authority anywhere has published is the American compounding page. It lists dihexa acetate among substances nominated and then withdrawn. The agency notes there that it has identified no human exposure data, and that it lacks important information on whether the substance would cause harm ^[4].

That entry is often read backwards. Without a place on the 503A or 503B list, an unapproved active ingredient may not lawfully be compounded in the United States at all ^[4].

Anti-doping

Dihexa is banned in sport at all times, in and out of competition, although the 2026 Prohibited List does not name it. A full-text search of the list returns neither Dihexa nor PNB-0408 ^[10].

Two catch-all clauses reach it. Section S2.3 names hepatocyte growth factor, and the heading of section S2 extends the ban to "other substances with similar chemical structure or similar biological effect(s)" ^[10].

Section S0 covers "any pharmacological substance which is not addressed by any of the subsequent sections of the List and with no current approval by any governmental regulatory health authority for human therapeutic use" ^[10]. Which clause applies changes the penalty range, not the ban.

German anti-doping law is the exception people miss. Its annex names only insulin-like growth factor 1 and its analogues, mechano growth factors and thymosin beta-4 derivatives, and it has no open similarity clause ^[41]. No validated detection method for Dihexa has been published ^[32]. The wider picture is at peptides banned in sport.

Compared with related peptides

Three things separate Dihexa from the substances beside it. It is the only entry here with a published record of research fraud behind its stated mechanism ^[5], ^[11].

It is also the only one whose active principle has been through a properly powered trial in people and come out with nothing ^[1]. The Russian peptides beside it are untested by that standard rather than disproven.

And it is the only one banned in sport, through clauses that do not name it ^[10]. A peer-reviewed review in 2026 nonetheless files Dihexa alongside Semax and Selank as a neuroactive peptide circulating in the consumer market, while noting that clinical trials are lacking ^[48].

Common misconceptions

• "The half-life is about two weeks, so it lasts for days." That figure of 12.68 days comes from three rats, with a margin of error of several days. Its source is the paper under a Notice of Concern. In people the same molecule cleared in roughly 1.5 hours ^[6], ^[8].
• "It works when swallowed." The developer measured 7.8 percent surviving in simulated stomach fluid and 0.1 percent in intestinal fluid. It called oral bioavailability poor and variable, and built a prodrug for that reason ^[9].
• "The growth-factor mechanism is established." The paper that made the biochemical case was retracted for falsified figures, two companion papers with it. What remains is sponsor data and one independent mouse study that never tested the binding ^[5], ^[19], ^[21].
• "Fosgonimeton is a different molecule, so its failure says nothing." It is the phosphate ester of Dihexa and splits to Dihexa in the blood. Levels of that active form ran three to twelve times above the prodrug's own. The 554 participants of LIFT-AD were exposed to Dihexa daily for 26 weeks ^[16], ^[8], ^[1].
• "It is listed under German anti-doping law." It is not. The annex names neither Dihexa nor hepatocyte growth factor, and it has no open similarity clause. The sport ban comes from the international list instead ^[41], ^[10].
• "No side effects have been reported, so it is safe." Nothing is reported because nothing collects reports: PubMed held no case report at all ^[32]. Meanwhile the pathway it targets is a known cancer driver, and no carcinogenicity study exists ^[31], ^[4].
• "Its wide use in stem cell research shows it is well characterised." That use is in dishes, replacing a growth factor in culture medium ^[23], ^[24]. It says nothing about a living person, and the reason it works there is precisely that it drives cells to grow and differentiate.

Frequently asked questions

Has Dihexa ever been tested in people?

No. No registered study has given Dihexa to a person, under that name or as PNB-0408. The trial registers of the United States and the European Union both come back empty, and the FDA states that it knows of no human exposure data for the substance in any form. What was tested in people is its phosphate form, fosgonimeton, given as an injection under the skin.

What is Dihexa?

It is a laboratory molecule built from a fragment of angiotensin IV, a signalling peptide the brain makes itself. Only two of its building blocks are real amino acids, so chemists call it a peptidomimetic rather than a peptide. Joseph Harding and John Wright's group at Washington State University designed it to be fat-soluble enough to reach the brain.

Did the Alzheimer's trial of Dihexa's prodrug work?

No. LIFT-AD enrolled 554 people with mild to moderate Alzheimer's disease and ran for 26 weeks. The main measure, a combined score counted in standard deviations, differed from placebo by -0.08, which is nowhere near a real effect (p = 0.70). Every secondary measure missed as well, and the company stopped the programme in September 2024.

Why was the research behind Dihexa retracted?

An investigation at Washington State University found that figures 1B and 2A/C of the 2014 paper explaining the mechanism contained falsified or fabricated data. The journal retracted it on 29 April 2025 and named two researchers as solely responsible. Two further papers from the same group were retracted, and the company later paid four million dollars to settle allegations of data falsification.

How long does Dihexa stay in the body?

The figure of 12.68 days circulating online comes from three rats in the paper that now carries a Notice of Concern, with a margin of error of several days. When the same molecule was measured in people, it cleared with a half-life of roughly 1.5 hours. That is a difference of about two hundredfold.

Can Dihexa be taken by mouth?

The company that developed it measured the opposite of what the first paper claimed. After one hour, 7.8 percent survived in simulated stomach fluid and 0.1 percent in simulated intestinal fluid. Its own conclusion was that the molecule is likely not suitable as an oral drug, which is why a phosphate prodrug was built instead.

Is Dihexa dangerous?

Nobody knows, and that is the problem. The signalling pathway it is meant to strengthen runs through MET, a gene that drives several cancers when it is overactive. No carcinogenicity study, no long-term toxicology and no genetic toxicity data exist. The FDA writes that it lacks important information on whether the substance would cause harm in people.

Is Dihexa banned in sport?

Yes, at all times, in and out of competition. It is not named on the 2026 Prohibited List, but two catch-all clauses cover it: section S2.3, which names hepatocyte growth factor and extends to substances with a similar biological effect, and section S0 for substances no health authority has approved. Which of the two applies changes the penalty range, not the ban.

Is Dihexa a peptide like BPC-157 or TB-500?

Not really. It carries only two genuine amino acids and was deliberately designed to be fat-soluble, with a calculated octanol-water partition coefficient of 177.8. Its maximum solubility in water is under 50 micrograms per millilitre, so it does not dissolve the way research peptides do. Laboratory work with it used pure dimethyl sulfoxide as the carrier.

Why does this page list Dihexa doses?

Because the studies cannot be read without them, and because the figures themselves carry a warning. Every dose on record comes from animals or cells, mostly from the one paper that is under an unresolved integrity question. The only independent, blinded, sample-size-planned animal study used exactly that dose and found no effect at all.

Cite this page

The facts on this page were checked on 15 September 2026, and every register search behind the status table was run on that date. Nobody is developing this molecule any more, so the retraction record and the anti-doping reading are the parts most likely to change.

myPeptides Research & Editing. (2026). Dihexa: what the studies show, status and safety. Version 1.0, 15 September 2026. myPeptides Peptide Register. Retrieved from https://mypep.app/peptides/dihexa

How pages in this register are compiled and graded is described under methodology; the full register is at peptides.

Last verified: 15 September 2026. Next review: on any resolution of the 2021 Notice of Concern, on any change to the compounding or anti-doping entries, or on any publication that tests this molecule in people.