Dihexa: The Ten Million Times Figure Is From a Press Release, Not a Paper
Neither founding paper mentions BDNF, three of the four were retracted in April 2025, and the same molecule went into a 554 person trial under another name and did not beat a dummy injection.
What it is
Dihexa came out of angiotensin IV, a fragment of the blood pressure hormone system that improves rat maze performance but is destroyed by enzymes. Joseph Harding's laboratory at Washington State University cut it to a three amino acid core and capped both ends, so enzymes could not recognise it and it could reach the brain. (PMID 23055539) The university spun it out as M3 Biotechnology, later Athira Pharma, now LeonaBio.
What the trials found
Search the United States government's trial registry for dihexa and nothing comes back; the European register answers "Query did not match any clinical trials."
It has been in people under another name. Fosgonimeton is dihexa with a phosphate group on the tyrosine, the trick that makes a greasy molecule soluble enough to inject, and it comes off again in the blood. Athira's own Phase 1 paper calls it a prodrug of ATH-1001, and ATH-1001 is dihexa.
Hua and colleagues published it in 2022 (PMC9108585). Of 88 people, 65 received fosgonimeton under the skin and 23 a dummy injection, at single doses of 2 to 90 mg and daily doses of 20 to 80 mg for nine days.
Then LIFT-AD, NCT04488419. 554 people with Alzheimer's disease, treated 26 weeks with a daily injection under the skin at 40 mg or 70 mg or with a dummy. The result needs its population: it was assessed only in the 40 mg arm, and only in people not on an acetylcholinesterase inhibitor, the older class of Alzheimer's drug, 287 of the 554. On a combined score of thinking and daily function, lower being better, the drug came out at -0.208 against -0.126 on the dummy. That is a gap of about eight hundredths of a unit in the drug's favour, which the company called not statistically significant.
The animal record is mostly one laboratory. McCoy and colleagues 2013 gave dihexa to male Sprague-Dawley rats whose memory had been chemically blocked with scopolamine, a drug that interferes with the signalling nerve cells use to lay down what they have learned, 8 to 10 a group, and the top dose of each route, 1 nmol into a brain fluid space, 0.50 mg/kg into the belly cavity, 2.0 mg/kg by mouth, brought them level with rats that had never been drugged. (PMID 23055539) Sun and colleagues 2021, unconnected to that laboratory, repeated it in mice bred to develop Alzheimer's changes, 1.44 and 2.88 mg/kg for three months, 6 to 12 a group, by a route their methods state two ways. (PMID 34827486) Wells and colleagues 2024 gave it to 40 male Wistar rats in a Huntington's model for five weeks, at a dose and by a route their abstract does not state and whose full text I could not reach, and it did nothing. It is easy to read that as independent replication and it is not: its senior author co-authored angiotensin IV work with Harding for about twenty years. (PMID 38489193)
Ten million times what, exactly
The number traces to a Washington State University news release dated 11 October 2012, which found dihexa "to be seven orders of magnitude more powerful than BDNF", and put the retail version in the same breath: it would take 10 million times as much BDNF to get as much new synapse formation.
It is not in the paper. I searched the full texts of McCoy 2013 and the 2014 follow up for BDNF and for brain-derived neurotrophic factor and found neither in either. (PMID 25187433) The Alzheimer's Drug Discovery Foundation got there in 2021, writing that the two "were not directly compared in the publication". A Europe PMC search for both words returns eleven papers, none a head to head experiment.
What it was scaled from is a dish: rat hippocampal neurons in culture, dihexa in the medium at ten to the minus twelve molar for five days, and dendritic spines, the knobs where a nerve cell takes contacts, at about 41 per fifty micrometres against roughly 15 with vehicle.
The comparison is the wrong shape. Dihexa is not supposed to switch anything on by itself. The story, and it comes from the 2014 paper that has since been retracted, is that it binds hepatocyte growth factor, a repair signal the body already makes, and sharpens it, while BDNF does the whole job. Comparing them by concentration is like comparing how far you turn a volume knob against how much electricity the speakers draw.
Three of the four papers behind the compound were retracted in April 2025, that 2014 paper among them. It is also where the binding figure still quoted comes from, 65 picomolar, a number that says dihexa sticks to that growth factor very tightly. (PMID 25187433, retracted by PMID 40312093) The retraction notice says a Washington State University investigation found falsified and fabricated data in specified figures and named Leen Kawas, who had been chief executive of the company developing the compound until she resigned in October 2021, and Joseph Harding as solely responsible. In January 2025 the company agreed to pay just over four million dollars to settle allegations that it had not disclosed the misconduct suspicions to the National Institutes of Health, which resolves them without a finding of liability. The fourth paper, McCoy 2013, was not retracted: it has sat under an expression of concern since September 2021, and it holds every rat dose, spine count and half life anybody quotes. (PMID 34551989)
How it compares
Nearly every entry in this reference fails the same way: no human data, so nothing to argue with. This one fails the other way. Add a phosphate to dihexa and you get fosgonimeton, which went into a 554 person trial and missed its primary endpoint. The molecule people buy has no trial. The molecule one atom group away has a negative one, and the vendor pages quote neither.
Where the dose came from
No human oral dose of dihexa has ever been established in anything I could find. The figures on vendor pages and community dosing guides disagree: across the pages I opened they ran from about 5 mg a day to about 50 mg a day, and not one showed its working. Dihexa does not dissolve in water, so the rat work dosed it into a vein and into the abdominal cavity in 75 percent DMSO, an industrial solvent that carries things through skin, and the material sold to individuals is commonly supplied the same way.
Here is arithmetic that lands near those figures, and it is mine, not a derivation I found anybody publishing. The rat oral dose in the founding paper is 2 mg/kg, and one standard body surface area conversion to a human dose divides by 6.2, giving 0.32 mg/kg, or about 22.6 mg for a seventy kilogram adult. (PMID 27057123)
Two things break it. The method is written for the highest animal dose producing no adverse effect in toxicology, not for one that worked in a maze. And it divides the result by a further ten before anybody is dosed. Run that step and the figure is 2.26 mg.
What could go wrong
The Alzheimer's Drug Discovery Foundation put the safety record in one line in 2021: "There are no publications documenting the long-term safety of dihexa in humans or animals." I found nothing published since that changes it.
The hazard people raise is cancer, because the system dihexa is meant to amplify is a well documented driver of tumour growth and spread. The same review calls that theoretical and untested, and a Europe PMC search I ran across dihexa with cancer, tumour and carcinogen terms returned 43 hits with no study testing it against tumour growth in any animal. The dihexa evidence for cells letting go of each other and moving sat in the retracted paper.
The numbers that exist belong to the injected version. Across 549 people through thirty weeks there were no deaths in any group, and serious events ran 15 of 218 on the dummy injection, 11 of 224 at 40 mg and 3 of 107 at 70 mg. Injection site reactions dominated, 14 percent of the dummy arm against 57 percent at 40 mg and 73 percent at 70 mg, and Alzforum records the higher dose arm was stopped mid-trial. Two people came off the Phase 1 study, one for a fall in neutrophils, a type of white blood cell, at 40 mg and one for an allergic skin reaction at 80 mg.
The timing does not line up either. In rats dihexa lasted 12.68 days after 10 mg/kg into a vein and 8.83 days after 20 mg/kg into the belly cavity, that second figure on four animals. In people the active molecule lasted about an hour and a half. Days against hours, and I found nothing published explaining the gap.
What we do not know
Whether the fourth paper survives. I found no statement from the journal about whether it is being kept or is still under review.
What is in the vials. The one chemical survey I found analysed 159 samples sold as smart drugs across nine European countries and Australia, found dihexa in a single dietary supplement, and reported neither purity nor content against label. (PMID 40558871)
What this adds up to
Human evidence exists, it is negative, and almost nobody selling this mentions it. Every vendor page I have seen says dihexa has never been tested in humans. True of the exact molecule. Not of the molecule with a phosphate on it, which went into a 554 person trial, was given to 331 of them, and did not separate from the dummy injection.
The retractions do not prove dihexa does nothing. They remove the binding measurement and the mechanism paper while leaving the rat doses and the spine counts in the one paper I could find no ruling on. The numbers survive. The reason to think they mean anything does not.
If you have seen dihexa for sale in the past year, I want to know whether the page mentioned fosgonimeton or the 554 person trial anywhere.
Frequently asked
Is dihexa really seven orders of magnitude more potent than BDNF?
That figure comes from a Washington State University press release dated 11 October 2012, not from a study, and it refers to counts of new connections in rat nerve cells grown in a dish. The paper the release was announcing does not contain the word BDNF anywhere, nor does its 2014 follow up, and the Alzheimer's Drug Discovery Foundation reported in 2021 that the effects of BDNF and dihexa were not directly compared in the publication. I searched for any study putting the two in the same experiment and found none.
Has dihexa ever been tested in humans?
Not under that name. The United States trial registry returns no studies for dihexa and the European register returns none. The same molecule has been in people as fosgonimeton, which is dihexa with a phosphate group attached so it can be injected, and that phosphate is cleaved off in the blood. A Phase 1 study enrolled 88 people, of whom 65 received the drug by injection under the skin and 23 received a dummy injection, and a Phase 2/3 trial in Alzheimer's disease enrolled 554 people, gave the drug to 331 of them, and did not beat the dummy injection on the measure it was built to answer.
Is dihexa a peptide?
Not a six amino acid one, whatever the product pages say. Its chemical name is N-hexanoic-Tyr-Ile-(6) aminohexanoic amide, and the structure contains two amino acids, tyrosine and isoleucine, with a six carbon fatty chain capped on the front and a six carbon amide tail on the back. PubChem's record, CID 129010512, gives the formula C27H44N4O5 and an average weight of 504.7.
What dose of dihexa do people take, and where does it come from?
No human oral dose of dihexa has ever been established in anything I could find, so every figure in circulation is convention rather than evidence and none of the pages carrying one showed a derivation. Across the vendor and community pages I opened the reported amounts ran from about 5 mg a day to about 50 mg a day by mouth, with no two agreeing. Converting the rat oral dose of 2 mg/kg by body surface area gives about 22.6 mg for a seventy kilogram adult, but that is my arithmetic rather than anyone's published derivation, and the published method both starts from a toxicology dose rather than a maze dose and divides the result by a further factor of ten before any human is given anything, which would put it at 2.26 mg. Two things travel with those figures: dihexa does not dissolve in water, so the rat work dissolved it in DMSO, an industrial solvent that carries things through skin, and the material sold to individuals is commonly supplied the same way, and although cycling schedules circulate I found no study in any species comparing one dosing schedule against another.
Are the dihexa studies retracted?
Three of the four papers underpinning the compound were retracted in April 2025, including the one that established the growth factor mechanism and produced the binding figure that still circulates. The notice on that paper states that a Washington State University investigation found falsified and fabricated data in specified figures and that two named authors were solely responsible. The fourth paper, the 2013 study that introduced dihexa and carries every rat dose, spine count and half life figure people quote, has not been retracted, but it has been under an expression of concern since September 2021 that has not been resolved.
Could dihexa cause cancer, and is it banned in sport?
The Alzheimer's Drug Discovery Foundation wrote in 2021 that no study had tested dihexa's effects on tumour growth or cancer progression, and I ran a Europe PMC search across dihexa with cancer, tumour and carcinogen terms, got 43 hits, and found no study testing it against tumour growth in any animal either. The signalling system dihexa is meant to amplify is one of the best documented drivers of tumour growth and spread, and the same review calls that hazard theoretical. On sport, the word dihexa does not appear anywhere on the 2026 World Anti-Doping Agency Prohibited List, which I pulled and searched in full, though the catch-all opening section covers substances with no current approval from any governmental regulatory health authority and dihexa fits that wording. The growth factor section names hepatocyte growth factor itself, but it limits itself to effects on muscle, tendon and ligament, so whether it reaches a compound sold for memory is arguable.
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.
Get the next one in your inbox
Free. One email a week at most. Unsubscribe in one click.
Double opt-in · No spam · Unsubscribe anytime
More in Cognitive and neuro
- Cerebrolysin Is a Licensed Medicine Made of Pig Brain, and Its Own Label Says Nobody Can Measure ItAustria has prescribed it since 1996 with a dosing table by condition. Ten cerebrolysin papers are flagged retracted on PubMed, and the amount people inject at home stops exactly where the label's syringe limits stop, 5 mL into a muscle and 10 mL into a vein.
- P-21 Is Not a Cerebrolysin Derivative, and Every Study I Found Is RodentEvery number I could find attached to this compound came out of a mouse or a rat, fed in chow, put down a tube into the stomach, leaked from a pellet under the skin, or injected into the abdomen. I found no study, in any species, of the injection under the skin people buy.
- PE-22-28: The Antidepressant Claim Is Ten Mice in a Water CylinderOne laboratory in every paper I found, male mice only, injected into the abdominal cavity. The 23 hour duration figure was measured on a different molecule, and the same group later reported the effect reverses direction with dose.