No trial has tested dihexa under its own name. Its injectable prodrug reached phase 2/3 and missed.

Why this note exists

Dihexa is now on FDA's compounding agenda. FDA's early announcement, current as of April 15, 2026, lists cathelicidin (LL-37), GHK-Cu, dihexa acetate, Melanotan II and pegylated mechano growth factor for a Pharmacy Compounding Advisory Committee meeting "before the end of February 2027," and says the time and location will be scheduled "in the coming months."[1] No date or briefing document is public on that page as of October 9, 2026.

Under its own name, dihexa's published record is animal and cell work.[22] That leaves out the one place its molecule was studied in people. This note sets out the regulatory record, the chemistry that links dihexa to fosgonimeton, what the fosgonimeton trials found, and what has happened to the papers dihexa's reputation rests on. A note on molecules: every human result below is a result for fosgonimeton, the prodrug, and is labelled that way.

The regulatory record

Not approved, in no 503A category, and on FDA's withdrawn-safety list

An openFDA query of Drugs@FDA data on October 9, 2026 returned no application with dihexa or fosgonimeton as an active ingredient.[24] FDA's list of substances nominated for 503A compounding, updated May 14, 2026, sorts nominations into category 1 (under evaluation), category 2 (significant safety risks) and category 3 (nominated without adequate support). Dihexa appears in none of the three.[3]

It appears on FDA's page of bulk drug substances that may present significant safety risks, current as of April 22, 2026, under "Bulk drug substances nominated but withdrawn," which FDA describes as substances "previously in category 2 of the interim policies" whose nominations "were withdrawn by the nominators."[2] The entry reads in full: "FDA has not identified any human exposure data on drug products containing dihexa acetate administered via any route of administration. FDA lacks important information regarding any safety issues raised by dihexa acetate, including whether it would cause harm if administered to humans."[2]

An advisory meeting is not a decision. FDA says "[a]dvisory committees make nonbinding recommendations to the FDA, which generally follows the recommendations but is not legally bound to do so."[4] A 503A bulks-list discussion is also about whether pharmacies may compound with a nominated substance, not about whether a drug has been shown to be safe and effective.[3]

The chemistry

Fosgonimeton is dihexa with a phosphate attached

Dihexa was reported in 2013 by a Washington State University group as "N-hexanoic-Tyr-Ile-(6) aminohexanoic amide," a metabolically stabilized analog built from the first three residues of an angiotensin IV analog, which the authors called "orally active" and "blood-barrier permeant," with antidementia activity in rat models.[7] Several authors of the group's 2014 paper also listed an affiliation with M3 Biotechnology, Inc.[9]

FDA's Global Substance Registration System lists dihexa under UNII 9WYX65A5C2, records "ATH-1001" among its names, and links it to fosgonimeton with the relationship "PRODRUG->METABOLITE ACTIVE."[5] The fosgonimeton record (UNII H91OA9858J, also listed as ATH-1017 and NDX-1017) gives the same hexanoyl-tyrosyl-isoleucyl-aminohexanamide chain with a phosphono group on the tyrosine.[6] In other words, the registered structure of fosgonimeton is dihexa carrying a phosphate. The company's phase 1 paper describes fosgonimeton as "a prodrug" that "is rapidly converted into the active metabolite ATH-1001 in plasma."[11]

Why build a prodrug at all? The developer's 2023 preclinical paper, which calls the metabolite "fosgo-AM," gives the reason. It reports only 7.8% of the metabolite remaining after one hour in simulated gastric fluid and 0.1% in simulated intestinal fluid, "poor oral bioavailability with high variability in exposure," and concludes it "is likely not suitable as an oral drug."[10] Its aqueous solubility was below 50 µg/mL, animal studies needed a pure DMSO vehicle, and the phosphate prodrug was made to dissolve at more than 70 mg/mL for subcutaneous injection.[10] Those are laboratory and rodent findings from the company, not human data. They do sit uneasily with the 2013 description of dihexa as orally active, and pepmg found no human study of oral dihexa to settle it.[7][20]

The human evidence

Six registered fosgonimeton studies, one large randomized trial

A ClinicalTrials.gov search on October 9, 2026 for dihexa returned no studies.[20] A PubMed search the same day returned 19 records for dihexa; none reports giving it to people, and several match only in cell or chemistry work.[22] A ClinicalTrials.gov intervention search for fosgonimeton, ATH-1017 or NDX-1017 returned six studies, all sponsored by the developer.[21]

Enrollment as listed on ClinicalTrials.gov on October 9, 2026. The extension enrolled people who finished a parent trial, so the counts overlap and are not summed.[13][15][16][17][18][11]

The phase 1 trial gave subcutaneous fosgonimeton to 88 people: 48 healthy young men in single doses from 2 to 90 mg, 29 healthy older adults at 20 to 80 mg daily for nine days, and 11 people with Alzheimer's at 40 mg daily for nine days.[11] The active metabolite peaked in plasma at about half an hour and had a half-life of about 1.5 hours. The authors reported an effect on the P300 brain-wave latency in the Alzheimer's group (p = 0.027), but that comparison was 7 people on drug against 4 on placebo.[11]

ACT-AD randomized 77 people to placebo, 40 mg or 70 mg daily for 26 weeks, with 60% continuing acetylcholinesterase inhibitors such as donepezil. In the sponsor's topline release, the primary endpoint, change in P300 latency, "was not met by protocoled analysis."[12] The release highlighted a better result in the patients not taking those inhibitors, describing it both as "a pre-specified subgroup analysis" and as "a subsequent post hoc analysis."[12] The registry's posted results list only baseline values for the P300 and ADAS-Cog11 measures.[13]

The decisive trial

LIFT-AD tested the subgroup that looked promising, and it missed

After ACT-AD, LIFT-AD was amended to stop enrolling people on acetylcholinesterase inhibitors and to drop the 70 mg arm, so that its primary analysis tested 40 mg against placebo in exactly the group the earlier subgroup had pointed to.[14] The peer-reviewed report, funded by the sponsor and written with company authors, covers 554 people randomized at 90 US sites, 549 dosed, and a primary analysis population of 287 (143 on fosgonimeton 40 mg, 144 on placebo), with a mean age of 73.[14]

The trial "did not achieve its primary or secondary endpoints." At week 26 the between-group difference was −0.08 (SE 0.10) on the composite Global Statistical Test (p = 0.70), −0.70 points on ADAS-Cog11 (p = 0.35) and +0.67 points on ADCS-ADL23 (p = 0.61); plasma neurofilament light, a further secondary endpoint, differed by −3.91 pg/mL (p = 0.26).[14] Of the exploratory blood biomarkers, only p-tau217 changed with nominal significance (−13.7% versus placebo, p < 0.01).[14]

The authors give a reason the trial may have struggled to show anything: the placebo group did not decline. Placebo ADAS-Cog11 scores improved slightly, by 0.39 points, over 26 weeks, against 1.09 points on fosgonimeton, and the authors write that "[t]he lack of decline in cognition and function in the placebo group was unexpected."[14] They point to subgroups with moderate dementia or an APOE ε4 allele, where placebo did decline; those differences were also not significant (ADAS-Cog11 −1.16, p = 0.39, and −1.07, p = 0.33).[14] That is a fair limitation of the design. It is not evidence of benefit.

The safety data cover all 549 dosed. Injection-site reactions occurred in 57.1% on 40 mg and 72.9% on 70 mg, against 14.2% on placebo, and adverse events led 10.7% and 21.5% to withdraw, against 4.6%.[14] Eosinophilia was reported in 7.1% and 7.5% on drug and in no one on placebo. There were no deaths; serious adverse events were 4.9% and 2.8% on drug against 6.9% on placebo, and five serious events were judged related to fosgonimeton, including angioedema once in each dose group.[14]

The rest of the programme ended with it. The open-label extension was terminated because "the Ph2/3 parent trial (LIFT-AD; NCT04488419) did not meet its primary endpoint."[16] SHAPE, in Parkinson's disease dementia and dementia with Lewy bodies, stopped after 28 people, short of the roughly 75 evaluable participants planned, for "study design limitations."[17] In January 2026 the developer, Athira Pharma, renamed itself LeonaBio; that announcement names lasofoxifene and ATH-1105 as its lead candidates and does not mention fosgonimeton.[19]

The doses above are fosgonimeton doses in the populations studied, reported as published. pepmg does not convert a prodrug dose into an amount of dihexa, and no trial tested dihexa by mouth.

The foundation papers

The mechanism paper was retracted; the founding paper is under a notice of concern

Dihexa's proposed mechanism, binding hepatocyte growth factor and boosting signalling through its receptor c-Met, was set out in a 2014 paper in the Journal of Pharmacology and Experimental Therapeutics.[8] The journal retracted it in April 2025 "at the request of the Editor" after a Washington State University investigation found that Figures 1B and 2A/C, and data in a subsequent erratum submission, contained "falsified and/or fabricated data." The notice names two authors as "solely responsible."[9]

The 2013 paper that introduced dihexa, from the same group, carries a 2021 notice of concern from the same journal.[7] A PubMed search on October 9, 2026 for retracted publications with the group's senior author returned three papers in that journal, from 2011, 2012 and 2014.[23] Other laboratories have since published rodent studies of dihexa, and the developer's own later preclinical work on the metabolite is separate from the retracted paper.[22][10] But the mechanism most often quoted for dihexa traces back to a paper that is no longer part of the record.

Three things this note is not saying

It is not saying fosgonimeton and dihexa are the same product. They are different substances; the link is FDA's registry recording dihexa as fosgonimeton's active metabolite, and the trials measured an injected prodrug.[5][6]

It is not predicting the advisory committee. No meeting date, briefing document or nominated use is public, and any recommendation would be nonbinding.[1][4]

It is not saying the HGF/MET idea is dead. The LIFT-AD authors read the consistent but non-significant differences as a sign of biological activity, and the trial's placebo group did not decline as expected.[14] What it is saying is that the only controlled human test of dihexa's molecule did not meet its endpoints, and that FDA's panel will be discussing a substance with no human data under its own name.

Questions people are asking