cerebrolysin

the most-trialled compound on this board that still cannot say what it is. thousands of randomised patients, and no molecular weight, no sequence and no purity specification anywhere in the record.

tier D · focus · 1,070 CASTA · primary missed

verdict

does it work: mixed, and measured. early neurological scores move, 90-day disability does not, and CASTA's day-90 modified Rankin 0 to 2 ran 199 of 529 against 208 of 541 on placebo. two cochrane reviews and three independent syntheses read the whole file and rate every estimate low or very low.

on whether it actually works — mixed, and the two halves are worth keeping apart. what moves: NIHSS at day 14 pools at a mean difference of 1.39 (95% CI 0.53 to 2.25) across 1,521 randomised patients, ESCAS moved an aphasia battery by 14.8 points, CEREHETIS cut symptomatic haemorrhagic transformation, and Cochrane's vascular dementia review reports cognition at SMD 0.36 (0.13 to 0.58) and global function response at RR 2.69 (1.82 to 3.98). what does not move: 90-day disability. CASTA at n=1,070 put day-90 modified Rankin 0 to 2 at 199 of 529 on the preparation against 208 of 541 on placebo, and the pooled estimate across 1,885 patients sits at RR 1.31 (0.90 to 1.91). mortality is null at moderate certainty. independent groups did reproduce a small cognitive benefit and rate their own finding very low certainty. this record has been verified repeatedly. the verification is what produced the split.

on whether the human evidence is thin — it is the opposite of thin. Cochrane maintains two separate reviews on this compound, one in acute ischaemic stroke and one in vascular dementia. the stroke review pools 7 randomised trials and 1,601 participants in its compound-only version. the vascular dementia review runs 6 randomised trials and 597 people. two manufacturer trials at n=242 and n=217 have posted full results to ClinicalTrials.gov, numbers and all. what fails is the certainty attached to the volume. Cochrane rates the vascular dementia cognition estimate very low quality and writes that any benefit may be too small to be clinically meaningful.

on the positive trials, because there are real ones — there are several. CARS met its day-90 motor endpoint. ESCAS moved the Western Aphasia Battery by 14.8 points against placebo. CEREHETIS cut symptomatic haemorrhagic transformation after thrombolysis. the manufacturer's vascular dementia trial posted an ADAS-cog+ difference of -6.17 (-8.22 to -4.13). the pattern across all of them is the problem: the endpoints measured soonest move, and the 90-day disability endpoint keeps not moving. CASTA, the largest trial ever run at n=1,070, missed its confirmatory endpoint. the manufacturer's own three-arm comparison against donepezil came back null at p=0.6348, and the repeat attempt was withdrawn before a single patient enrolled.

on what a vial sold under this name contains — the clinical literature never answered that question, so a vendor certificate cannot answer it either. every trial in this entry dosed one Austrian process output, measured in millilitres of preparation, given by intravenous infusion in a hospital. no registry record in the 42-study set states a molecular weight, an amino acid sequence or a purity specification, and the Chinese process patent in the source set defines the article by manufacturing route plus a fingerprint similarity of at least 0.90 against a reference sample. there is no single moiety for a purity percentage to describe. the one compositional paper in the literature reports that products sharing the cerebroprotein hydrolysate name were not comparable to each other by HPLC fingerprint.

based on published trials and primary registry records. research framing only. not medical advice.

why D-tier

D on the conflicting clause rather than the thin one. The human data comes first, because the usual gray-market complaint does not apply: two Cochrane reviews, 1,601 randomised participants in the compound-only stroke review, 597 in vascular dementia, 185 pooled in the randomised traumatic brain injury meta-analysis, 42 registered trials, and two manufacturer trials with full results posted to the registry. On volume alone this belongs high on the board. The certainty is what fails, and it fails from the independent side. Cochrane rates the vascular dementia cognition estimate very low quality and writes that the effects may be too small to be clinically meaningful. The stroke all-cause death estimate is moderate certainty for little to no difference. Neither review contains a pooled disability estimate: the stroke review records that no included trial reported death or dependence at the end of follow-up, early death, quality of life or time to restoration of capacity for work in a poolable form, and the vascular dementia review records that none reported quality of life or caregiver burden. Where the individual trials did measure it, the answer arrived: CASTA's day-90 modified Rankin 0 to 2 ran 199 of 529 against 208 of 541 on placebo. The independent Glasgow synthesis puts short-term cognition at standardised mean difference -0.16 (95% CI -0.30 to -0.03) with very low certainty and substantial funnel-plot asymmetry, and reports the effect gone by three to seven months at -0.07 (p=0.32). Its own conclusion is that the effects were modest and probably less than would be considered clinically relevant. Not C. C fits a compound whose preclinical work points one way and has not reached people yet. Thousands of randomised, blinded humans have been studied here, several of the trials are quadruple-masked, and two have posted their numbers in full. Not F. F requires the compound to be fake, a scam or actively harmful beyond reasonable doubt, and cerebrolysin is none of those. Four randomised trials met a pre-specified primary endpoint. The manufacturer's vascular dementia trial posted ADAS-cog+ -6.17 (-8.22 to -4.13). The rodent layer includes a properly randomised, blinded, placebo-controlled dose-response study. The non-fatal serious adverse event signal, RR 2.15 (1.01 to 4.55) in 2020 and RR 2.39 (1.10 to 5.23) in 2023, is a flag rather than a demonstration of harm, since mortality and total serious events come out null in every synthesis including the manufacturer-independent ones. Not B, and this is the real contest. B needs the words good evidence. In the 2025 vascular cognitive impairment review, cerebrolysin's Cohen's d of 0.28 (95% CI 0.03 to 0.52) sits level with galantamine at 0.28 and just below donepezil at 0.29, which is a fair point in the compound's favour. It fails on what sits behind the number: GRADE Low, only 173 participants pooled for this compound in a review covering 22,347, no functional endpoint poolable at all, follow-up beyond four weeks analysed in not one trial, and the review's own text and table printing different point estimates. Every synthesis in this file carries a low or very low rating from the people who read the trials. Voluminous, low certainty, sponsor dominated and internally contradictory is exactly what D describes. If C-RETURN reads out at 440 patients on a 90-day Glasgow Outcome Scale Extended endpoint and it separates, the tier moves. Until an adequately powered trial outside the manufacturer's orbit reports a durable endpoint, D is the ceiling.

the core tension

the endpoints measured soonest keep moving and the endpoints that matter most keep not moving, and the trials large enough to settle it mostly belong to the company that makes the product. CASTA missed its confirmatory endpoint at n=1,070 and has posted nothing to the registry in roughly 15 years. CEREHETIS cut symptomatic haemorrhagic transformation and day-14 NIHSS, then found day-90 modified Rankin about the same in both arms. ESCAS moved the aphasia battery by 14.8 points and got only a trend on modified Rankin. the Alzheimer's meta-analysis shows cognition significant at 4 weeks and gone at 6 months. the Glasgow synthesis shows the same decay at three to seven months and reports funnel-plot asymmetry suggesting publication bias. CARS is the standout exception, and its own discussion volunteers that a poorer outcome than typically expected of the control group can explain the superiority of the treatment arm. neither Cochrane review contains a disability endpoint at all, because across thousands of randomised patients no included trial collected one.

what it is

Cerebrolysin is a brain-tissue protein hydrolysate, manufactured in Austria by EVER Neuro Pharma, sold as an aqueous preparation and dosed by volume rather than by mass of any defined molecule. The Chinese process patent in the source set names pig brain as the source tissue and describes successive pepsin and trypsin hydrolysis followed by ultrafiltration. It has been in clinical use for decades and holds national marketing authorisations rather than a single central one. No registry record among its 42 registered trials states a molecular weight, an amino acid sequence or a purity specification. Every trial dose in the literature is expressed in millilitres of preparation: 20 mL, 30 mL, 40 mL, 50 mL.

what it does

The claim made for it is neurotrophic and neuroprotective activity after acute brain injury and in cognitive decline. Randomised trials have run it in acute ischaemic stroke, vascular dementia, Alzheimer's disease, traumatic brain injury, post-stroke aphasia and subarachnoid haemorrhage. The proposed mechanism rests on review articles asserting neurotrophic factor content, and on rodent work showing subventricular-zone progenitor proliferation and CREB-pathway suppression of neuroinflammation. No assay located in this sweep quantifies BDNF, GDNF, NGF or CNTF per millilitre of preparation. The compositional paper that does exist measured a neurofilament reporter and an HPLC fingerprint, and states that it did not measure neurotrophin content.

origin

Austrian, and sponsor concentration runs through the whole literature. EVER Neuro Pharma, formerly EBEWE, manufactures the product and is lead sponsor on 10 of the 42 registered trials, including CASTA, both observational registries and both dementia trials that posted results. The independent clinical literature is largely Russian, Chinese, Romanian, Polish, Egyptian and Korean. A second concentration sits inside the academic literature. A single investigator cluster in Cluj and its collaborators authors CARS, the pooled CARS meta-analysis, the CAPTAIN trial series, the CAPTAIN prospective meta-analysis, the adverse-event meta-analysis of twelve trials, and ESCAS. The two large syntheses produced entirely outside that orbit are the Cochrane reviews and the Glasgow analysis of animal-derived nootropics. Neither lands negative. Both land small and very low quality: Glasgow reports statistically significant benefit on short-term cognition, global function and activities of daily living, then writes that the supporting evidence is weak and the effects probably below clinical relevance.

why researchers are interested

It carries marketing authorisations in a long list of countries, which reads to a buyer as validation the board does not grant. It also has what most gray-market neuro compounds lack: thousands of randomised, blinded patients in indexed journals. The trials are real, several of them are positive, and the mechanism story is supported by properly blinded rodent work including a randomised placebo-controlled dose-response study in rats.

does it work

Mixed, and every part of that was measured rather than left open. On durable function the answer is no: CASTA missed its confirmatory endpoint at n=1,070 and its day-90 modified Rankin 0 to 2 came in at 199 of 529 against 208 of 541 on placebo, with placebo numerically ahead. The 2025 fourteen-trial pool puts functional independence at RR 1.31 (0.90 to 1.91) across 6 trials and 1,885 participants. The Alzheimer's meta-analysis shows cognition significant at 4 weeks and gone at 6 months. Neither Cochrane review contains a pooled death-or-dependence endpoint, because no included trial reported one in a poolable form. On early and surrogate endpoints, several trials moved convincingly and independent groups reproduced a small benefit. NIHSS pools at a mean difference of 1.39 (0.53 to 2.25) across 1,521 participants. Cochrane's vascular dementia review reports cognition at SMD 0.36 (0.13 to 0.58) and global function response at RR 2.69 (1.82 to 3.98), and writes that courses of the preparation improved cognition and general function, at very low quality. The independent Glasgow synthesis finds short-term cognition, global function and daily living all significantly in the compound's favour, at very low certainty, and gone by three to seven months. That split is what puts this at D rather than at F or B.

claims vs the data

• cerebrolysin improves recovery after stroke — partially true — Early and surrogate endpoints moved in several randomised trials, and CARS met a pre-specified day-90 motor endpoint. The durable endpoint did not follow. CASTA showed no significant difference on its combined confirmatory test at n=1,070, and its day-90 modified Rankin 0 to 2 ran 199 of 529 against 208 of 541 on placebo, with placebo numerically ahead. Pooled modified Rankin 0 to 2 sits at RR 1.31 (95% CI 0.90 to 1.91) across 6 trials and 1,885 participants, crossing no effect, and neither Cochrane review contains a pooled death-or-dependence endpoint because no included trial reported one in a form the review could pool.
• it treats Alzheimer's disease — contradicted — The manufacturer's own three-arm trial against donepezil posted null, ANCOVA p=0.6348 across treatments, with the preparation against donepezil at -0.450 (-2.719 to 1.819), p=0.6962. The repeat attempt was withdrawn at zero enrolment. The industry-linked meta-analysis shows cognition at SMD -0.40 (-0.66 to -0.13) at 4 weeks and -0.37 (-0.90 to 0.16), p=0.1710, at 6 months. The dose-ranging trial found the lowest of three doses the only one significant on cognition.
• it works for vascular dementia — weak — This is the indication where the answer comes closest to yes. Cochrane pools cognition at SMD 0.36 (0.13 to 0.58) across 3 studies and 420 people and global function response at RR 2.69 (1.82 to 3.98) across 2 studies and 379 people, writes that courses of the preparation improved cognition and general function with no suggestion of adverse effects, then rates both very low quality on heterogeneity and high risk of bias and warns any effects may be too small to be clinically meaningful. Every study in that review that disclosed funding was industry supported, and no new eligible study had appeared since the 2013 version. The single manufacturer trial with posted numbers is genuinely positive at -6.17 (-8.22 to -4.13), and it is one trial.
• thousands of patients prove it after traumatic brain injury — overreach — The circulating 8,749 figure comes from a review of 10 studies dominated by observational and historical cohorts. Inside that same review the Glasgow Coma Scale difference was not significant at 1.344 (-0.258 to 2.945) with I2 of 94.2 percent, mortality was not significant, and length of stay was not significant. The randomised, blinded pool is 185 in the CAPTAIN prospective meta-analysis, and CAPTAIN I itself missed its intention-to-treat primary and terminated for poor recruitment at n=46.
• it crosses the blood-brain barrier — weak — Asserted in reviews and vendor copy. No human blood-brain-barrier penetration study and no human pharmacokinetic study was located in this sweep. The work usually cited nearby is a cultured brain endothelial cell reporter study of GLUT1 gene expression, which is in vitro and is not a penetration measurement.
• it contains BDNF, GDNF, NGF and CNTF — weak — Asserted in review articles rather than measured. No assay located here quantifies any neurotrophic factor per millilitre of preparation. The paper most often cited for the composition claim is a rat spinal cord injury review with manufacturer authorship, not an assay paper. The one compositional study in the literature measured an HPLC fingerprint and a neurofilament reporter and states that it did not measure neurotrophin content.
• generic cerebroprotein hydrolysate is the same product — contradicted — The one compositional comparison in the literature reports profound differences by reversed-phase HPLC across 11 preparations, and reports that products sharing the cerebroprotein hydrolysate name were not comparable to each other. A separate Chinese rat study uses cerebrolysin as the positive control against cerebroprotein hydrolysate-I as the test article, which is a design that treats them as two different drugs. Several drug databases collapse them into one synonym. That caveat cuts both ways: the compositional paper is the only source of its kind, its conclusion favours the originator against every comparator, and the originator's manufacturer supplied the reference material.
• the purity percentage on the vial tells a buyer what they have — overreach — No registry record states a molecular weight, a sequence or a purity specification, and every trial dose is millilitres of preparation. The Chinese process patent defines identity as fingerprint similarity of at least 0.90 against a reference substance plus nitrogen content, which is a similarity criterion rather than a purity peak. There is no reference moiety for a purity figure to describe.
• it is dangerous — weak — Cochrane records non-fatal serious adverse events at RR 2.15 (1.01 to 4.55) in 2020 and RR 2.39 (1.10 to 5.23) in 2023, rising to RR 2.87 (1.24 to 6.69) on the 30 mL for 10 days schedule. Total serious adverse events, all-cause death and total adverse events come out null across every synthesis, including two produced independently of the manufacturer. CARS ran the other way entirely, with 3 serious events against 7 and four deaths, all on placebo. A signal that appears when fatal and non-fatal events are separated and vanishes when they are pooled is a flag to name rather than demonstrated harm.
• approval in dozens of countries validates it — partially true — National authorisations are real and one of them is documented in a primary record: Romanian authorisation 4610/2004/03, named in the CTIS registration for the C-RETURN trial. The count of roughly 40 to 50 countries that circulates traces to commercial pages in this sweep and was not verified against a register, because the regulatory connector returned a quota error rather than data. Against that, Cochrane's 2023 update records that two joint European Stroke Organisation and European Academy of Neurology guidelines advise against use, and the review authors argue against running further trials.

key facts

• molecular formula: not defined; a hydrolysate rather than a single compound
• molecular weight: none stated in any of the 42 registry records. manufacturer and review descriptions carry roughly 15 percent peptides under 10 kDa and 85 percent free amino acids; several vendor pages instead print 25 percent and 75 percent
• amino acids: not a single sequence. the Chinese process patent specifies 16 amino acids plus nitrogen content
• half-life: not characterised; no human pharmacokinetic study was located in this sweep
• type: brain-tissue protein hydrolysate, dosed by volume of preparation rather than by mass
• CAS: no CAS number for a defined active moiety appears anywhere in the trial record
• 1,070 patients in CASTA, confirmatory endpoint missed
• 3/42 registered trials with posted results
• 10/42 registered trials sponsored by the manufacturer
• 0 human pharmacokinetic studies located

frequently asked questions

What is cerebrolysin?

Cerebrolysin is a brain-tissue protein hydrolysate manufactured in Austria by EVER Neuro Pharma and supplied in ampoules, dosed by volume rather than by mass. The Chinese process patent in the source set names pig brain as the source tissue. It is not FDA approved. It holds national marketing authorisations rather than a single central European one, and one of those is documented in a primary record: Romanian authorisation number 4610/2004/03, with the substance registered only as CEREBROLYSIN CONCENTRATE under EU substance number SUB74627. No registry record among its 42 registered trials states a molecular weight, an amino acid sequence or a purity specification.

What does cerebrolysin do?

It is claimed to be neurotrophic and neuroprotective, and it has been tested in acute ischaemic stroke, vascular dementia, Alzheimer's disease, traumatic brain injury, post-stroke aphasia and subarachnoid haemorrhage. The published record splits by how soon the endpoint is measured. Early and surrogate measures moved in several randomised trials: CARS met its day-90 arm-function endpoint, ESCAS moved the Western Aphasia Battery by 14.805 points (95% CI 9.521 to 20.089), CEREHETIS cut symptomatic haemorrhagic transformation to odds ratio 0.248 (0.072 to 0.851). Durable measures mostly did not. CASTA missed its confirmatory endpoint at n=1,070 and put day-90 modified Rankin 0 to 2 at 199 of 529 against 208 of 541 on placebo, CEREHETIS found day-90 modified Rankin about the same in both arms, and the independent Glasgow synthesis reports cognition at standardised mean difference -0.16 (-0.30 to -0.03) short term, where negative favours the compound in that paper's coding, decaying to -0.07 (p=0.32) at three to seven months.

How is cerebrolysin dosed in the published trials?

Always in millilitres of preparation, never in milligrams of an active substance, and always by intravenous infusion in a clinical setting. The common stroke schedule is 30 mL daily for 10 days, used in CASTA and in the Cochrane subgroup analyses. CARS used 30 mL daily for 21 days on top of a standardised rehabilitation programme. The vascular dementia trial with posted results used 20 mL. CAPTAIN used 50 mL daily for 10 days followed by two further cycles of 10 mL. The Alzheimer's dose-ranging trial compared 10 mL, 30 mL and 60 mL and found the lowest dose the only one significant on cognition at 24 weeks. There is no approved dosing protocol outside the countries holding national authorisations, and no FDA label exists to quote.

What are the side effects of cerebrolysin?

There is no FDA label, so there is no adverse-reactions table or contraindications section to quote, and the picture has to be built from the trials. Total serious adverse events come out null across every synthesis: Cochrane 2020 puts them at RR 1.15 (95% CI 0.81 to 1.65) and all-cause death at RR 0.90 (0.61 to 1.32), the twelve-trial adverse-event meta-analysis reports serious events at RR 0.99 (0.74 to 1.32), and the 2025 fourteen-trial pool reports RR 1.08 (0.84 to 1.40). The signal sits in one subset. Cochrane 2020 reports non-fatal serious adverse events at RR 2.15 (1.01 to 4.55), p=0.047, across 4 trials and 1,435 participants, rising to RR 2.86 (1.23 to 6.66) on the 30 mL for 10 days schedule, and the 2023 update puts the same estimate at RR 2.39 (1.10 to 5.23). Fatal and non-fatal events move in opposite directions and cancel when pooled, so quoting the null on total serious events as an all-clear misses where the signal is. Mild agitation is described in the CEREHETIS report as a known effect. Nothing here characterises long-term or non-clinical exposure, because no trial studied it.

Is cerebrolysin FDA approved?

No. There is no FDA approval, no NDA or BLA, and no label. That negative should be read with its sourcing stated: the regulatory register connector assigned to this sweep was blocked by an account quota and returned zero records, so the absence of an FDA application was not confirmed against a primary register here. What is documented in primary records runs the other way. The product holds national authorisations in Europe rather than a central one, and the Romanian authorisation 4610/2004/03 appears in the CTIS registration for the C-RETURN trial. Cochrane's 2023 update records that its 2020 version informed two joint European Stroke Organisation and European Academy of Neurology guidelines on post-stroke cognitive impairment, and that both advise against use.

How much does cerebrolysin cost?

No verified price appears in the retrieved record. Where it is authorised it is a prescription product supplied in ampoules and administered by infusion, so any price is set market by market and no US clinical price exists at all. What circulates on the research market is priced by unrelated suppliers, and identity, concentration and contamination status are unverified unless a lot-specific certificate of analysis says otherwise. Upstream of the price sits the harder problem for any certificate, which is that no defined molecule exists for a purity percentage to describe.

related peptides

• semax — the other eastern-european neuro peptide with a domestic-literature-heavy file, graded higher
• selank — same regulatory geography, same problem with independent replication
• dihexa — neurotrophic claim with the opposite evidence profile: rodents only, no human trials

reptides grades the research record and cites the literature behind every call. research reference only; not medical advice.