Cerebrolysin: what the stroke and dementia trials show once you add them up

Cerebrolysin is the most heavily trialled molecule in this section: completed double-blind randomised trials in stroke, vascular dementia, Alzheimer’s disease and traumatic brain injury, plus two independent Cochrane reviews. The evidence splits along a clean line. On early measures the trials are positive — a nine-trial meta-analysis of 1,879 people reported better neurological scores at day 21 to 30, with a number needed to treat of 7.7 for a clinically relevant change on the NIH Stroke Scale [6], and pooled analysis of six vascular dementia trials found a cognitive benefit (standardised mean difference 0.36) [8]. On the outcomes that decide whether a stroke treatment earns a place in hospital — survival and independence — the 1,070-person CASTA trial missed its pre-specified endpoint and Cochrane found cerebrolysin probably makes little to no difference to death from any cause [1,2].

What cerebrolysin is

Cerebrolysin isn’t a single peptide. It is a mixture of low-molecular-weight peptides and free amino acids prepared from purified pig brain protein, given by intravenous infusion or injection [1,8]. Because it’s a mixture, it has no single amino-acid sequence, and batch-to-batch composition is controlled by the manufacturing process rather than by a defined chemical structure. The manufacturer, EVER Neuro Pharma, is based in Austria; several of its staff are authors on the major trials [3].

The claimed mechanism is that fragments in the mixture mimic the brain’s own growth factors, protecting nerve cells after injury and helping them rewire. The Cochrane authors describe it as having “potential neuroprotective properties” [1], which is a fair summary of where the mechanism stands: the idea comes from lab and animal work, and the mixture’s active components are still unidentified.

The stroke evidence

This is where the biggest trials are, and where the evidence is deepest and most mixed.

CASTA was run across Asia with 529 people on cerebrolysin and 541 on placebo [2]. Its pre-specified main result was neutral. In a subgroup analysis done after the fact, people with severe strokes (NIH Stroke Scale above 12) showed a trend in favour of cerebrolysin, and 90-day deaths in that subgroup were 10.5% vs 20.2% [2]. Subgroup findings from a trial that missed its main endpoint are hypotheses, and the authors said as much: the observation “should be confirmed by a further clinical trial” [2].

Several smaller trials followed, with mixed results:

• CARS (2016). Cerebrolysin or placebo for 21 days, started 24 to 72 hours after stroke, alongside a standard rehabilitation programme. At 90 days, arm function on the Action Research Arm Test was clearly better with cerebrolysin, with a Mann-Whitney effect size of 0.71 [3]. The authors called the study exploratory and asked for a larger confirmatory trial [3].
• Korean trial (2016), 70 people. Cerebrolysin plus rehabilitation for 21 days made no overall difference to motor recovery. Only in the subgroup with severe motor impairment did it do better than placebo [4].
• Iranian trial (2017), 100 people. Treatment started within 18 hours and ran for 30 days. The NIH Stroke Scale improved more with cerebrolysin at day 30, with an effect size of 0.66 [5]. Its authors include the statistician and a manufacturer scientist who also worked on CARS [3,5].

A note on the “Mann-Whitney” numbers

Many cerebrolysin papers report results as a Mann-Whitney effect size rather than as points on a scale. It answers a simple question: pick one patient at random from each group, and how often does the cerebrolysin patient do better? A value of 0.50 means a coin toss, no difference. The CARS authors describe 0.71 as a large effect and 0.62 as small-to-medium [3]. So the 0.71 in CARS [3] means the cerebrolysin patient came out ahead about 71% of the time in such pairings. It’s a legitimate statistic, but it doesn’t tell you how many extra patients regained the use of an arm or walked out of hospital, which is what you’d want to know.

A 2018 meta-analysis pooled nine randomised trials (1,879 people) and concluded cerebrolysin improved early neurological scores, with a number needed to treat of 7.7 for a clinically relevant change on the NIH Stroke Scale [6]. Its authors include investigators and statisticians from the manufacturer-funded trials, and its main outcome was an early score at day 21 or 30 rather than longer-term disability or death.

The Cochrane review asks a different question with stricter methods. Its 2023 update included seven trials and 1,773 participants and found cerebrolysin or similar mixtures (one trial tested Cortexin, a comparable extract from cattle brain) “probably result in little to no difference” in death from any cause (risk ratio 0.96, 95% CI 0.65 to 1.41; moderate certainty) [1]. None of the trials reported the outcome stroke doctors care most about, death or dependence at the end of follow-up. The review also found a possible increase in non-fatal serious adverse events with cerebrolysin (risk ratio 2.39, 95% CI 1.10 to 5.23; three trials, 1,335 people) [1].

The two analyses aren’t strictly contradictory, since one looks at early symptom scores and the other at death and harms. The gap they leave is narrower than it sounds, and more specific: the trials in this set measured symptom scores and deaths, while the outcome that decides whether a treatment changes a patient’s life — death or dependence at the end of follow-up — was not reported by any of them [1]. That is the result a stroke unit would need before adding cerebrolysin to its protocols.

The dementia evidence

The single biggest dementia trial is from Russia. In 242 people with vascular dementia, adding cerebrolysin 20 mL IV to aspirin over two treatment cycles improved the ADAS-cog+ cognitive score by 10.6 points at 24 weeks, against 4.4 points on placebo [7]. That is a large difference for a dementia trial.

When Cochrane reviewers pooled six vascular dementia trials (597 people), they found a benefit on cognition (standardised mean difference 0.36) and on global function, but rated the evidence “very low” quality [8]. Five trials were in China, Russia or Romania; every trial that reported funding was industry-supported; and the reviewers judged the included papers at high risk of bias. Their conclusion: if there are benefits, “the effects may be too small to be clinically meaningful” [8].

For Alzheimer’s disease, a 2015 meta-analysis of six trials found better cognition at 4 weeks but not at 6 months, and better global clinical impression at both time points [9]. Two of its six authors, Johannes Vester and Edith Doppler, are listed on the CARS paper with the statistics contractor and the manufacturer respectively [3,9].

Brain injury

A 2023 meta-analysis of traumatic brain injury pooled ten studies with 8,749 patients, mixing retrospective records with prospective trials [10]. It found better scores on the Glasgow Coma and Glasgow Outcome scales but no change in deaths or length of hospital stay [10]. Retrospective data are especially prone to bias, because sicker or healthier patients may have been more or less likely to receive the drug.

Safety and side effects reported

Individual trials, most of them manufacturer-supported, describe cerebrolysin as safe and similar to placebo [3,7]. The Cochrane stroke review is the more careful read: total adverse events were similar, but non-fatal serious adverse events were more common with cerebrolysin, particularly on the 30 mL-for-10-days schedule [1]. It is also a product from animal brain, given intravenously, which raises general questions about purity and immune reactions that the published trials weren’t designed to answer.

Legal and regulatory status

According to the Cochrane authors, cerebrolysin is widely used for stroke in Russia, Eastern Europe, China and other Asian and post-Soviet countries, where it is sold as a prescription medicine [1]. It is not approved as a medicine in the UK or by the US FDA, and it isn’t part of standard stroke or dementia care in either country. Products sold online outside that regulated supply are unlicensed and of unknown origin.

What is still open

• Whether cerebrolysin reduces long-term disability after stroke, since no trial has shown it on death or dependence.
• Whether the benefit in severe strokes, seen after the fact in CASTA, holds up in a trial designed to test it.
• Whether the rise in non-fatal serious adverse events in the Cochrane review is real.
• Whether the dementia benefits survive a large, independent, low-bias trial.
• Which components of the mixture, if any, do the work.
• Whether it adds anything for people who also receive modern stroke treatments such as clot-busting drugs or thrombectomy, since the trials above were not designed around them.
• Whether effects on early symptom scores, measured at 3 or 4 weeks, last beyond the first few months.

What this means

Our position is that cerebrolysin has earned a properly powered, independent confirmatory trial, and that it has not earned a place in stroke or dementia care on the strength of the trials that exist. Both halves matter. The early-score improvements repeat across several trials and two pooled analyses [6,8] — an unusually deep randomised record for a peptide — and the CASTA subgroup signal in severe strokes is a real hypothesis worth testing [2]. None of that yet reaches the outcome a patient or a stroke unit actually cares about.

So when we are asked whether cerebrolysin is worth taking, our answer is no on the claims it is usually sold on — recovery after stroke, protection in dementia — and firmly no through the unlicensed online supply that sits outside the regulated market in Russia, Eastern Europe and Asia [1]. The animal-derived, intravenous route and the Cochrane signal of more non-fatal serious adverse events on the 30 mL protocol [1] make unsupervised self-administration a bad trade against a benefit that has not been shown on disability.

We also think the honest reading is narrower than either camp usually offers. It is tempting to write off a drug whose largest trial missed its endpoint and whose main trials are manufacturer-linked [3,6,9]; it is equally tempting to point at the positive small trials and the 0.71 effect size in CARS [3] and treat the case as settled in cerebrolysin’s favour. We would rather hold both facts at once. The trial record here is unusually deep and unusually dependent on the manufacturer, and the field’s next useful step is a trial with death or dependence as its endpoint, run by investigators with no stake in the result.

That is a position, and a reader can disagree with it. If you think the early-score and subgroup signals are strong enough to act on now, you should be able to name the trial result you are relying on and the outcome it measured. If you share our view, the practical consequence is simple: cerebrolysin is a research question with a promising shape, it belongs with the rest of this section as interesting and unresolved, and it is not something to source from a grey-market vial.

Educational content only — not medical advice. Many peptides discussed on HPR are not approved for human use. Talk to a qualified clinician before making any decision about your health.