Aniracetam (Ro 13-5057)

Chemical identity, and the bond that does not survive the liver

Aniracetam's identity is unambiguous. One PubChem record covers it — CID 2196, CAS 72432-10-1, formula C12H13NO3, molecular weight 219.24, InChIKey ZXNRTKGTQJPIJK-UHFFFAOYSA-N, IUPAC name 1-(4-methoxybenzoyl)pyrrolidin-2-one — and the FDA substance register agrees on that CAS and formula under UNII 5L16LKN964, status approved. That same register entry gathers the WHO ATC code N06BX11, INN code 4902, ChEMBL36994, MeSH C036466, ChEBI 47943 and DrugBank DB04599. The Roche development code Ro 13-5057 and the trade names Draganon, Sarpul and Ampamet all point back to the one substance. The scaffold is the one the name implies: a 2-pyrrolidinone ring bearing a 4-methoxybenzoyl group on the nitrogen, where piracetam bears an acetamide.

That benzoyl amide governs everything downstream. Cleaving it yields 2-pyrrolidinone and p-anisic acid; opening the ring instead yields N-anisoyl-gamma-aminobutyric acid, written elsewhere as 4-p-anisamidobutyric acid. Nakamura, writing from Nippon Roche in 2002, records that systemic administration converts the parent principally to 2-pyrrolidinone and p-anisic acid in rats, and to N-anisoyl-GABA in humans (CNS Drug Rev, PMID 12070527). A rat result and a human result therefore describe partly different chemical species in circulation, and nearly all of the pharmacology below was measured in rodents.

Regulatory position differs by jurisdiction and has moved. Cohen and colleagues in 2021 listed the compound among four drugs not approved for human use in the United States (PMID 34484905). A 2025 market-surveillance report by twelve official medicines control laboratories with Australia lists it in appendix Table A1 as a prescription medicine in the European Union and Australia, while the same paper's narrative says Australia does not authorise piracetam but does authorise aniracetam, and that same table lists piracetam as prescription-only in both (PMID 40558871). The two statements do not agree.

What reaches the blood, and for how long

Human exposure data comes from one crossover study. Tian and colleagues gave 400 mg orally, under fasting conditions to twenty healthy Chinese men in an open, randomised, single-blind two-period design, measuring plasma by liquid chromatography-tandem mass spectrometry (PMID 19025058). Peak plasma concentration was 8.75 plus or minus 7.82 ng/mL for the test formulation and 8.65 plus or minus 8.70 for the reference; time to peak was 0.4 plus or minus 0.1 hours; terminal half-life was 0.47 plus or minus 0.16 and 0.49 plus or minus 0.24 hours. The standard deviations approach or exceed the means on peak concentration and on both area-under-curve measures; on terminal half-life the standard deviation runs a third to a half of the mean. Both formulations still fell inside the 80 to 125 per cent bioequivalence bounds. The only intravenous human administration located here is a 1984 hypoxia-challenge study that reports no pharmacokinetic parameters at all (PMID 6434496).

Rodent disposition is better characterised than human disposition. Ogiso and colleagues reported oral bioavailability of the parent in rats as 11.4 and 8.6 per cent at 50 and 100 mg/kg, fitting a seven-compartment model with ten first-order rate constants and one Michaelis-Menten constant, and recorded that the fit failed for 4-p-anisamidobutyric acid (PMID 9572910). In rat brain the same group found the ratio of brain to plasma area under the curve was 53 to 55 per cent for 2-pyrrolidinone against 2.4 to 3.2 per cent for the parent and 3.9 to 4.2 per cent for p-anisic acid, with 4-p-anisamidobutyric acid below detection in brain tissue (PMID 10784432).

Age changes the picture. Endo and colleagues studied six hospitalised women with cerebrovascular disease, mean age 84.5 years and mean creatinine clearance 20 to 30 mL/min, after 200 mg orally; serum anisic acid and p-methoxyhippuric acid peaked at two hours and returned to baseline by six, and metabolite half-lives ran four- to sevenfold longer than in young volunteers (PMID 9062694). Six patients is the largest human pharmacokinetic dataset in older people located in PubMed for this compound.

Metabolite exposure dwarfs parent exposure wherever both were measured together. In pigeons given 100 mg/kg orally, peak plasma concentration was 242.96 ng/mL for the parent against 4254.87 ng/mL for N-anisoyl-GABA (PMID 31002681). Cai and Wang built the first tandem mass spectrometry assay for that metabolite in human plasma and applied it after a single oral dose, validating over 0.0485 to 19.4 micrograms per millilitre, three orders of magnitude above the range used for the parent (PMID 22552003). The primary human disposition study behind all of this remains Roncari's 1993 paper in a Drug Investigation supplement, which PubMed does not index.

The receptor result, and the concentration at which it was obtained

Ito and colleagues at Chugai reported reversible allosteric potentiation of ionotropic quisqualate responses in Xenopus oocytes injected with rat brain mRNA, observed above concentrations of 0.1 mM (J Physiol 1990, PMID 1975272). Kainate, N-methyl-D-aspartate and GABA responses in the same oocytes were unaffected. Receptor affinity for agonist and channel reversal potential did not change. The same paper reported potentiation of intracellular responses in rat hippocampal CA1 and of population excitatory postsynaptic potentials at Schaffer collateral and mossy fibre synapses.

Isaacson and Nicoll established the mechanism a year later. Patch clamp recording in hippocampal slices showed enhancement of glutamate-evoked whole-cell currents and, in outside-out patches, strong reduction of glutamate receptor desensitisation, with fast synaptic currents prolonged and their peak amplitude raised (PNAS 1991, PMID 1660156). That finding is why the molecule became a template rather than a product: Arai and Kessler describe the ampakine class as structurally derived from it (PMID 17504103).

Metabolite work complicates attributing the receptor effect to the parent at all. Nishizaki and Matsumura expressed heteromeric GluR1,2, GluR1,3 and GluR1,2,3 receptors in Xenopus oocytes and compared the parent against three metabolites (PMID 11834304). 2-Pyrrolidinone potentiated kainate-evoked GluR1,2,3 currents in a bell-shaped dose-response from 1 nM to 300 microM, maximal at 100 microM, reaching about 180 per cent of basal 60 minutes after a five-minute exposure; the parent reached about 130 per cent and p-anisic acid about 103 per cent. On GluR1,3 the metabolite potentiated as it did on GluR1,2,3; on GluR1,2 it depressed currents instead, which is the basis of the authors' conclusion that the target is the calcium-permeable subtypes. KN-93 inhibited the potentiation; inhibitors of protein kinase C and of cAMP-dependent protein kinase did not.

Two numbers from this section and the one before it have never been reconciled in any paper located here. Potentiation in Ito's oocyte preparation begins above 0.1 mM, which is 100 micromolar. Mean peak plasma concentration after 400 mg orally in twenty men was 8.75 ng/mL, which at a molecular weight of 219.24 is roughly 0.04 micromolar, lower by a factor of about two and a half thousand. Neither figure is a brain concentration, and the rat brain-to-plasma ratio for the parent is 2.4 to 3.2 per cent. That gap is the reason metabolite pharmacology matters, and the secondary literature does not address it.

Impaired animals, and healthy ones

Most in vivo work comes from one laboratory using one strain. Nakamura and Shirane perfused metabolites directly into brain regions of stroke-prone spontaneously hypertensive rats and reported delayed increases in acetylcholine release from N-anisoyl-GABA at 0.1 and 1 microM and p-anisic acid at 1 microM in nucleus reticularis thalami, dorsal hippocampus and prefrontal cortex, with no effect from the parent compound (PMID 10513566). Shirane and Nakamura then blocked the prefrontal response with the group II metabotropic glutamate antagonist MCCG and with MCPG, but not with the group I antagonist AIDA or the AMPA antagonist YM90K (PMID 10699452), which places the effect away from the AMPA account.

Behavioural findings from that programme track the impairment rather than the compound. Site-specific increases in dopamine and serotonin release were reported in prefrontal cortex, basolateral amygdala and dorsal hippocampus, but not in striatum or nucleus accumbens shell, at 30 and 100 mg/kg orally in the same hypertensive strain (PMID 11282361). In the forced swim test, 10 to 100 mg/kg orally failed to shorten immobility in nine-week-old rats while 100 mg/kg shortened it in rats aged 25 to 30 months, and mecamylamine or haloperidol reversed that effect completely (PMID 11702095).

Two laboratories have tested healthy animals, and both reported nothing. The Baylor group gave 50 mg/kg orally each day to thirty adult male C57BL/6J mice under a double-blind design, measuring Morris water maze, fear conditioning, accelerating rotarod, odour discrimination, open field, elevated plus maze and marble burying, and no measure separated the groups (Elston 2014, PMID 25099639); the same laboratory then repeated the design in twenty-four male C57BL/6J mice at 100 mg/kg orally, dosed 30 minutes before each test, with the same result (Reynolds 2017, PMID 29946420). The Otago group found no effect on delayed matching-to-sample accuracy or on response latency in pigeons at 100 and 200 mg/kg intramuscularly and at 200 mg/kg orally, with pharmacokinetic sampling placing the test times near peak concentrations (PMID 31002681). That is one laboratory twice and one other laboratory, not three independent replications.

Anxiety is where the animal record contradicts itself. Nakamura's 2002 review records that the compound had previously shown no anxiolytic-like action in modified Vogel and Geller-Seifter conflict tests in rats, crediting that in one parenthesis both to its reference 42 and to unpublished data by J. R. Martin. Reference 42 is Martin and Haefely's 46-page pharmacology review in the same 1993 Drug Investigation supplement, 5(Suppl 1):4-49, DOI 10.1007/BF03258426, which PubMed does not index and which could not be read here. Nakamura and Kurasawa had by then reported effects at 10 to 100 mg/kg across social interaction, elevated plus-maze and conditioned fear stress in mice, blocked completely by haloperidol and nearly so by mecamylamine or ketanserin (PMID 11412837). The two null elevated plus maze measurements in healthy mice come from one laboratory, not from independent replication.

The human record, and what the trial count contains

Counts first. A PubMed search on aniracetam returns 320 records as of August 2026. Eleven carry the publication type Randomized Controlled Trial, thirteen carry Clinical Trial, one carries Meta-Analysis and none carries Systematic Review. No record in the corpus carries Retracted Publication, Retraction of Publication, Expression of Concern or Published Erratum. ClinicalTrials.gov returns zero registered studies for aniracetam, Ro 13-5057, Draganon, Ampamet, Sarpul and Memodrin alike; no trial of this compound has been prospectively registered where that registry indexes.

Those eleven do not contain what the number suggests. Four are Chinese-language trials of herbal preparations in which aniracetam is the active comparator: Shenwu gelatin capsule in 166 patients (PMID 17993008), Huannao Yicong capsule in 90 (PMID 18184542), modified Huanglian Wendan decoction in 64 (PMID 20353029) and scalp acupuncture with auricular point sticking in 90 (PMID 25022106). One is the bioequivalence study of two aniracetam formulations against each other, with no placebo arm. One is Fudenberg's 1994 interpretive review of Alzheimer's disease heterogeneity, which PubMed types as both Clinical Trial and Randomized Controlled Trial and which is a review (PMID 7888107). Five remain, and that five is a filter result rather than a census: it is what the Randomized Controlled Trial publication type returns.

The five are European, span 1980 to 1991, and disagree with each other. Saletu and colleagues gave single oral doses of 250, 500 and 1000 mg against placebo and 2000 mg piracetam to ten geriatric subjects in a crossover design and reported EEG spectral changes and shortened reaction time (PMID 6927816). Sourander and colleagues randomised 44 patients with senile dementia of the Alzheimer type to 1 g daily or placebo for three months; cognitive tests improved in both arms, clinical evaluation separated neither, and treatment was interrupted for confusion in four aniracetam cases against one on placebo (PMID 3103163). Somnier and colleagues found one of nineteen neuropsychological measures favouring the drug and another favouring placebo in 44 subjects with solvent-exposure psychosyndrome (PMID 2188276). Senin and colleagues reported a difference from placebo at six months in 109 patients meeting NINCDS-ADRDA criteria (PMID 1822317), and Canonico and colleagues reported improvement at two and four months in a trial of 60 patients randomised to 1500 mg daily or placebo, arm sizes not stated in the indexed abstract, in a population described as slight to moderate brain pathology primary or secondary to vascular forms (PMID 1767242).

A sixth placebo-controlled study sits outside that filter. Saletu and Grunberger injected placebo, 10 mg and 100 mg aniracetam intravenously at weekly intervals in a latin square design to ten healthy male volunteers breathing a fixed 11.2 per cent oxygen mixture, and reported that the compound attenuated the hypoxia-induced deterioration recorded on EEG spectral analysis and on psychometric testing (PMID 6434496). PubMed types that record Clinical Trial rather than Randomized Controlled Trial, which is the whole reason the count of five excludes it. It is also the only intravenous human administration retrieved here, and therefore the only human result unaffected by whatever the oral bioavailability of this compound turns out to be.

Later syntheses do not resolve the split. Perng and colleagues pooled 235 studies and 44,854 patients and placed aniracetam inside a lumped category alongside piracetam, nimodipine, flunarizine, vinpocetine, hyperbaric oxygen, oxiracetam and EGb761; no compound-specific estimate appears anywhere in that paper (PMID 29502274). Lee and Benfield's 1994 Adis review, the document most often cited for the clinical profile, describes the evidence as preliminary and calls for trials to confirm it (PMID 8199398). Malykh and Sadaie recorded in 2010 that oxiracetam and aniracetam are no longer in clinical use (PMID 20166767).

Withdrawal in Japan, market samples, and a citation that turns back on itself

One sentence in the literature explains why the human record stops in the 1990s. Nakamura, then at Nippon Roche, wrote in 2002 that the drug had been prescribed in Japan for eight years to treat emotional disturbances such as depressed mood and anxiety or agitation following cerebral infarction, rather than memory impairment, and that Draganon had been withdrawn from the Japanese market because of the unexpected failure in the latest placebo-controlled double-blind study (PMID 12070527). No citation is attached to that sentence, which was read here in the paper's own full text. Europe PMC full-text searches for the phrase return that review and nothing else. The trial that ended the approval has no publication traceable here.

Content analyses have measured what is inside products since. Cohen and colleagues bought ten cognitive-enhancement supplements labelled with omberacetam, aniracetam, phenylpiracetam or oxiracetam and analysed them by non-targeted liquid chromatography quadrupole time-of-flight mass spectrometry; aniracetam was detected at up to 502 plus or minus 0.8 mg in a recommended serving, several detected drugs were absent from the labels, and nine of twelve checkable declared quantities were inaccurate (PMID 34484905). Table A1 of the twelve-laboratory European and Australian survey records aniracetam at a detection frequency of four, across three of the laboratories, mainly presented as a substandard or falsified medicine, with legal status given as prescription medicine in the European Union and Australia (PMID 40558871).

The last link doubles back. That 2025 surveillance paper describes what aniracetam is used for and lists its adverse events, naming unrest, anxiety, uneasiness, insomnia, urinary urgency, headache, vertigo, mild stomach pain, nausea, diarrhoea and rash, under a single citation: reference 57, which is Love 2024 in the Journal of Alzheimer's Disease. Love's paper proposes a mechanism model, is written from the research department of a commercial entity, records that its funding came from that same company and from private donors while declaring no conflict of interest, and cites for that same passage its own reference 1, Lee and Benfield's 1994 secondary review (PMID 38552113). No primary adverse-event dataset appears at any point in the chain. Love's abstract also states that the compound increases brain-derived neurotrophic factor expression, where the body attributes that to two combination experiments.

What is not known

No trial of this compound has been prospectively registered: ClinicalTrials.gov returns zero studies for aniracetam, Ro 13-5057, Draganon, Ampamet, Sarpul and Memodrin alike, and PubMed indexes no systematic review. The newest placebo-controlled human study in the corpus was published in 1991, so nothing in the human record was run to modern registration, reporting or trial-design standards, and none of the six reports blinding, allocation concealment and outcome definitions in enough detail to be appraised from the retrievable text. Human pharmacokinetics rests on twenty men in one oral crossover study and six elderly women in another; the primary disposition work sits in a 1993 Drug Investigation supplement that PubMed does not index. That supplement holds at least five Crossref-registered aniracetam papers, among them a 46-page pharmacology review by Martin and Haefely, Mayersohn's animal disposition paper, Roncari's human pharmacokinetics paper and a clinical-experience report by Senin and colleagues on senile dementia of the Alzheimer type, 5(Suppl 1):96-105, DOI 10.1007/BF03258430 — a clinical source on this compound that exists as a citation and cannot be read, by the same lead author as the 1991 trial. Absolute bioavailability, clearance and volume of distribution in people therefore cannot be checked against any readable source. The in vitro concentration at which allosteric potentiation was first demonstrated is roughly two and a half thousand times the mean peak plasma concentration measured in people, and no paper located here reconciles the two. Which chemical species produces any effect in a person is unresolved: the metabolite that dominates human plasma is not the metabolite that dominates rat brain. Nothing characterises effects in people under 18, in pregnancy, beyond six months of exposure, or on discontinuation. The trial that ended the Japanese approval has never been published, and no adverse-event dataset for this compound traces to a primary source rather than to a secondary review.

Questions