Mexidol (Emoxypine Succinate)

Identity, and three salts that share a name

The name Mexidol resolves to a salt rather than a molecule, and the registries say so in the way they write it out. PubChem holds a single record: CID 122298, CAS 127464-43-1, molecular formula C12H17NO5, molecular weight 255.27 g/mol, InChIKey IKMNOGHPKNFPTK-UHFFFAOYSA-N. The IUPAC name on that record, butanedioic acid;2-ethyl-6-methylpyridin-3-ol, uses a semicolon to mark a 1:1 salt rather than a single covalent molecule. The FDA substance registry files the same entity under a different preferred name, EMOXYPINE SUCCINATE, with UNII 2R985002CT, the formula split as C8H11NO.C4H6O4, mass 255.2676, and the same CAS number. Its code list carries MeSH C070020, EVMPD SUB35088, NCI Thesaurus C221980 and EPA DTXSID30155577, plus the World Health Organization drug dictionary name methylethylpiridinol succinate.

Three related substances circulate under overlapping names. The free base, emoxypine, is CID 114681: CAS 2364-75-2, C8H11NO, 137.18 g/mol, UNII V247P5H4E1. The hydrochloride is CID 128852, C8H12ClNO, 173.64 g/mol, UNII SN1FWZ77AE, the substance behind the ophthalmic preparation Emoxipin. The succinate is what the neurology trials administered. PubChem's synonym list for the hydrochloride nonetheless carries the string Mexidol hydrochloride, which does not describe the product those trials used. The pyridinol accounts for 137.18 of 255.27 daltons, 53.7 per cent by mass, so a milligram figure quoted for the salt and the same figure quoted for the base are not the same quantity of the same thing.

Literal-string lookup is unsafe here. Asking PubChem for ethylmethylhydroxypyridine succinate, the name used in the trial titles and on the WHO entry attached to the FDA record, returns CID 129761399: C12H13NO5, 251.23 g/mol, 10-propyl-4,8-dihydro-3H-[1,4]dioxocino[2,3-c]pyridine-2,5,9-trione. That is an unrelated molecule whose mass sits four daltons from the right answer, and nothing in the returned record flags the mismatch. The strings mexidol, emoxypine succinate and 2-ethyl-6-methyl-3-hydroxypyridine succinate all resolve correctly to CID 122298.

Secondary sources describe the molecule as a vitamin B6 derivative. Pyridoxine is CID 1054, C8H11NO3, 4,5-bis(hydroxymethyl)-2-methylpyridin-3-ol. Emoxypine is C8H11NO. The two formulas differ by exactly two oxygen atoms and both structures are 3-hydroxypyridines, but the ring substituents differ at three positions and the hydroxymethyl groups that pyridoxine is phosphorylated on are absent from emoxypine entirely, and no synthetic route from pyridoxine was located in PubMed. The derivative claim, and the search that failed to support it, are recorded in the unsourced array below.

Four placebo-controlled trials, one journal, one sponsor

EPICA randomised 150 patients aged 40 to 79 with hemispheric ischaemic stroke, 75 per arm, at Russian centres. The trial gave 500 mg daily by intravenous infusion for ten days followed by 375 mg daily orally for eight weeks, with placebo on the same schedule. Its registered primary endpoint was the mean modified Rankin Scale score at the final visit, days 67 to 71. Posted registry results give 1.098 (SD 0.768) against 1.46 (SD 0.947), a separation of 0.36 points on a six-point scale. Stakhovskaya and colleagues published it in 2017 and reported p=0.04.

MEMO ran at 15 centres in Russia and Uzbekistan and randomised 318 patients aged 40 to 90 with what the investigators termed chronic brain ischaemia. The primary endpoint was change in Montreal Cognitive Assessment score at day 75. Registry results give a mean rise of 4.22 points (SD 2.59) against 2.17 (SD 2.20); the publication reports the lower bound of the 95 per cent confidence interval for the between-group difference at 1.51 points. Six of the 318 did not complete. Fedin and colleagues published it in 2021.

MEGA is the outlier in the set: 333 children aged 6 to 12 with ADHD diagnosed by ICD-10 and DSM-5 criteria, randomised 1:1:1 across 14 Russian centres to two oral schedules and placebo for 42 days. The primary endpoint combined the inattention and hyperactivity/impulsivity subscales of the SNAP-IV. Mean changes were minus 9.33 (95% CI minus 10.59 to minus 8.07) on the twice-daily schedule, minus 7.31 (minus 8.44 to minus 6.17) on the once-daily schedule and minus 3.93 (minus 5.11 to minus 2.75) on placebo. Zavadenko and colleagues published it in 2022.

MIR is the largest and most recent stroke trial: 304 patients randomised at 17 centres in Russia, Kazakhstan and Uzbekistan, restricted to non-reperfused first-ever hemispheric stroke with a baseline NIHSS of 9 to 15. The posted registry results give the mean change in modified Rankin Scale at day 71 as minus 2.45 against minus 2.01, with no statistical analysis attached to the primary outcome. Shamalov and colleagues, publishing the main report in 2025, give no means and no confidence intervals, and report instead a difference in median mRS change at visit 4 against baseline at p=0.003. The confidence limits also disagree: the registry posts minus 2.70 to minus 2.23 for the treated arm, while Koltsov and colleagues, publishing the post-hoc analyses in the same journal later that year, print minus 2.70 to minus 2.32 for the same estimate and record the endpoint at days 69 to 73 rather than the registry's day 71.

All four appeared in a single title, the Korsakov Journal of Neurology and Psychiatry, in Russian with English abstracts. Every one of the nine studies registered on ClinicalTrials.gov under any of the compound's names lists Pharmasoft, the manufacturer, as both lead sponsor and responsible party. No study run by an independent group is registered anywhere the search reached, and none of the four trials has an independent replication.

Nine registrations, none of them prospective

A ClinicalTrials.gov intervention search across Mexidol, emoxypine, emoxypine succinate and ethylmethylhydroxypyridine succinate returns nine studies: eight completed, one recruiting, five with posted results. They cover ischaemic stroke in four records, chronic cerebral ischaemia, vascular cognitive impairment without dementia, ADHD in children, and primary open-angle glaucoma in two. Enrolment ranges from 60 to 333. Locations are Russia, with Uzbekistan and Kazakhstan appearing in the two multinational records.

None was registered before it began. EPICA started on 6 March 2015 and was first submitted on 31 May 2016, four days after its stated completion date. MEMO started in November 2019, completed in December 2020 and was submitted in February 2025, more than four years after the last patient and more than three after the primary publication. MEGA started in July 2019, completed in July 2020 and was submitted on that same February 2025 date. MIR started in November 2019, completed in August 2023 and was submitted in May 2024. The glaucoma phase 2 study started in November 2021 and was submitted in March 2025.

Retrospective registration removes the check that registration exists to provide: a record created after the analysis is complete cannot demonstrate that the primary endpoint was fixed in advance. The MIR post-hoc paper states that the protocol was registered on ClinicalTrials.gov as NCT06437626, without stating when relative to the trial start. It also cites a Russian permits-registry identifier, PHS-APIS-004-MEX-SOL-TAB, which was not retrievable from outside that system for this record.

Where the posted results and the papers diverge

MIR's registry entry and its published abstract do not describe the NIHSS the same way. The registry gives change from baseline at day 71 as a median of minus 7 in both arms, with interquartile ranges of minus 9 to minus 6 on drug and minus 8 to minus 5 on placebo. The abstract reports a difference in median NIHSS at p<0.001. Because the abstract does not say whether it is comparing absolute scores at the final visit or changes from baseline, the two statements are not necessarily contradictory; they also cannot be reconciled from the published text.

Two MIR secondary outcomes are titled as percentages and typed as counts. The outcome named percentage of subjects with mRS 0 to 1 carries the values 85 and 57 with its unit field set to Participants; its companion, mRS above 3, carries 22 and 39. Read as percentages the pair is impossible. Read as counts against the 141 and 138 participants who completed, the odds are 1.52 against 0.70, an odds ratio of 2.16, which matches the 2.2-fold figure the post-hoc paper reports. Read as counts the values reconcile with the published odds ratio; read as percentages they do not.

EPICA randomised 150 and posted a per-protocol population of 124, an attrition of 17 per cent split 14 and 12 across the arms. Its registry entry posts no statistical analysis for the primary endpoint at all; the p=0.04 exists only in the publication. Age-stratified sub-analyses of EPICA in 2020 (PMID 33016677) and of MEMO in 2022 (PMID 36412160) were published as separate papers in the same journal, so the same 150 and 318 patients appear in the indexed literature more than once.

The published conclusions are written as recommendations for practice. EPICA's states that mexidol is recommended for inclusion in the therapy of ischaemic stroke patients in the acute and early rehabilitation stages, and the age sub-analysis (PMID 33016677) repeats that sentence with diabetes added. The underlying measurement is a 0.36-point mean difference on the modified Rankin Scale among 124 analysed patients, in a trial registered fourteen months after it began.

What the mechanism claims rest on

The electrophysiology cited for a central mechanism is Motin and colleagues' 2012 study (PMID 22442954). Mexidol at 5 mM depressed the excitatory postsynaptic current in medial vestibular nucleus neurons by 96 plus or minus 2 per cent in slices from albino rats aged 13 to 17 days, and at 2.5 to 5 mM depressed the Schaffer-collateral response in CA1 pyramidal neurons by 94 plus or minus 3 per cent. MK-801 reduced the effect and CNQX did not. The paper's stated conclusion is that the central favourable effect can be mediated by ion mechanisms with glutamate- and GABA-ergic components, primarily by the inhibition of ion currents through NMDA receptor complex. Five millimolar of the pyridinol is roughly 690 micrograms per millilitre. The validated assay Melnik and colleagues built for the same molecule in rat brain tissue (PMID 30176507) was calibrated from 1 to 1500 nanograms per gram, which is the working range of the method rather than a measured ceiling on tissue concentration.

The GABA-benzodiazepine framing traces to an indexed paper. A 2022 review in Current Research in Pharmacology and Drug Discovery (PMID 35992374) states that Mexidol increases binding interaction at the GABA-benzodiazepine receptor complex and cites Iasnetsov 2012. The paper carrying that statement is PMID 23457968, in Aviakosmicheskaia i Ekologicheskaia Meditsina, which reports that at the neuronal level mexidol interacts with the GABA-A-benzodiazepine receptor complex in nearly 60 per cent of cells and inhibits ion currents through NMDA-receptor ion channels in nearly 80 per cent of neurons. Those are proportions of responding cells; the report gives no binding constant, and no Ki or IC50 for any GABA-A or benzodiazepine site was located. Gnezdilova and colleagues reported separately in 2010 (PMID 21254506) that the cerebrovascular response in rats vanished in the presence of bicuculline, a functional result about GABA-A rather than a measured affinity.

Antioxidant characterisation in a Western journal exists and is computational. Hoa and colleagues published a density-functional study in Royal Society Open Science in 2022 reporting moderate hydroperoxyl-radical scavenging in water and in pentyl ethanoate, and chelation of Cu(II) into complexes harder to reduce than free Cu(II). That paper models 2-ethyl-6-methyl-3-hydroxypyridine, the free base, while naming it Mexidol, and opens by describing the compound as registered for the treatment of anxiety disorders.

Whether the succinate half contributes anything has been measured twice. Shchulkin and colleagues (PMID 37338763) gave Wistar rats 100 mg/kg intravenously and quantified succinate in plasma and in the cytoplasmic and mitochondrial fractions of cerebral cortex, left-ventricular myocardium and liver by HPLC-MS/MS. Succinate rose in the cytoplasmic fraction, most in liver and less in cortex and myocardium, and rose only slightly in the mitochondrial fraction. Earlier, Iasnetsov and colleagues (PMID 23025045) measured oxygen consumption polarographically in isolated rat brain mitochondria and reported that mexidol at 0.85 mM succinate raised it by 35 to 45 per cent, an effect suppressed by malonate, the inhibitor of respiratory complex II. That enzyme, succinate dehydrogenase, is embedded in the inner mitochondrial membrane. The polarographic result was obtained in isolated mitochondria with the compound applied directly, and does not carry to the intact animal that the distribution study addressed.

One behavioural experiment ran the salt against its own base. Volchegorskii and colleagues tested emoxypine, mexidol and reamberin in the elevated plus maze in rats and reported that the anxiolytic effect and the stimulation of risk behaviour were most pronounced with the free base. Their reading was that combining the pyridinol cation with the succinate anion attenuated both effects. That is the opposite direction from the account in which the succinate moiety is what raises the salt's activity.

Pharmacokinetics, regulatory position and sport

No human pharmacokinetic study reporting a half-life for this molecule was located. Searching the compound names against half-life, t1/2, AUC and Cmax returns two PubMed records: a 1986 Russian-language paper on a water-soluble 3-hydroxypyridine antioxidant, and a 2014 comparison of two marketed tablet brands conducted in 14 male chinchilla rabbits. The human work that exists concerns metabolism rather than exposure, chiefly Sariev and colleagues' papers from 1999 and 2001 on urinary excretion of the glucuronide conjugate in patients with organic central nervous system lesions.

The most recent human measurement came out of an anti-doping laboratory and was not about this drug. Postnikov and colleagues (PMID 41611212) had three volunteers drink 900 mL of unpasteurised cow's milk collected on the last day of a course of Emidonol, a veterinary antihypoxant that dissociates into meldonium and emoxypine. Estimated urinary emoxypine peaked at 1360 plus or minus 240 ng/mL between seven and eight and a half hours, clearing over roughly 50 to 54 hours; meldonium peaked at 322 plus or minus 68 ng/mL between 5.8 and 9.0 hours, with elimination over roughly 35 to 45 hours. The emoxypine curve is a by-product of a paper about inadvertent meldonium findings.

Regulatory position outside the former Soviet sphere is empty. A Drugs@FDA query on mexidol, emoxypine and ethylmethylhydroxypyridine returned no matches, and the openFDA labelling endpoint returned none either. The 2022 review states that approval was not obtained in the United States or Europe and that the drug sits on the Russian essential medicines list. Jedrejko and colleagues (PMID 38403950) record registration in Russia and in Ukraine. Trial conduct in Uzbekistan and Kazakhstan is a fact about where the studies ran and is not evidence of marketing approval in those countries. Nothing in the record indicates availability or legality for a person outside the jurisdictions where it is registered.

Two reviews in the anti-doping literature treat the compound as an open question. Jedrejko and colleagues place it beside meldonium and trimetazidine, both in category S4.4 of the World Anti-Doping Agency prohibited list, note that it is neither monitored nor included in testing protocols, and argue it could warrant consideration. Transporter work is in vitro: Shchulkin and colleagues reported in 2023 (PMID 38004395) that the molecule is not a substrate of ABCB1 or SLCO1B1 in Caco-2 and HEK293 systems but inhibits both, more weakly than verapamil and rifampicin respectively, with intestinal ABCB1 inhibition left untested in vivo.

What is not known

The four trials share a structure that limits what can be read out of them. All are sponsored by the manufacturer, all appeared in one journal, all were registered after they began, and none has been replicated by an independent group. No trial of this compound conducted outside Russia, Uzbekistan and Kazakhstan was located, and no non-Russian regulator has assessed a dossier for it: Drugs@FDA and the openFDA labelling endpoint return no matches on any of its names. Full texts of the trial reports are in Russian and were not retrieved for this record, so the English abstracts and the registry postings are the whole basis of what is written above, and the three divergences noted between them cannot be resolved without the full papers. Human pharmacokinetics are effectively unpublished in the indexed literature: no half-life, clearance, volume of distribution or oral bioavailability figure traceable to a primary human study was found, and the only human exposure measurement located came from an anti-doping laboratory studying contaminated milk in three volunteers. No measured brain tissue concentration in a living animal was located either, so the millimolar concentrations used in the slice work cannot be placed against achievable exposure. Nothing characterises long-term administration, reproductive or developmental effects, carcinogenicity, or behaviour in populations outside the trial inclusion criteria. No study in healthy people measured cognition, mood or physical performance. Whether the succinate counter-ion contributes anything in the intact animal remains open: the isolated-mitochondria result and the distribution study address different questions, and the one experiment that compared the salt against its own free base found the base more active.

Questions