Alpha-GPC and Stroke Risk: What the Korean Data Show

A retrospective cohort study built on South Korea’s National Health Insurance Service database — more than 12 million adults aged 50 and over, followed from 2009 to 2018 — found that the 108,877 people who used alpha-GPC had an adjusted hazard ratio of 1.43 for total stroke compared with non-users, with a dose-response gradient. That is roughly a 43% higher relative rate over ten years, not a proven cause: people prescribed alpha-GPC in Korea are by definition people with cognitive complaints, and that group already carries higher vascular risk. No randomized trial has been run that could settle the question, and none is planned. If you have had a stroke or TIA, have cerebrovascular disease or uncontrolled hypertension, or take an anticoagulant, this is a finding to put in front of the person who writes your prescriptions before you take alpha-GPC at all.

Written by the Fog & Clarity editorial team · Published 24 July 2026 · Last updated 24 July 2026

What the Korean cohort actually found

Alpha-GPC (L-alpha-glycerylphosphorylcholine, also sold as choline alfoscerate) is a prescription drug in Italy and South Korea and a dietary supplement in the United States. That regulatory split is why the safety data exist at all: because Korean doctors prescribe it, Korean insurance records track who took it and what happened to them afterward.²

The analysis compared 108,877 alpha-GPC users against roughly 11.9 million non-users aged 50 and over, over a ten-year follow-up window. The adjusted hazard ratios were 1.43 for stroke of any type, 1.34 for ischemic stroke and 1.37 for hemorrhagic stroke. Higher cumulative exposure was associated with higher risk — the dose-response relationship is the part that makes this hard to dismiss.²

Study	Population	Exposure	Duration	Outcome
Korean NHIS retrospective cohort	>12 million adults aged 50+; 108,877 alpha-GPC users vs ~11.9 million non-users	Prescribed alpha-GPC, cumulative dose measured	10 years (2009–2018)	Adjusted HR 1.43 total stroke, 1.34 ischemic, 1.37 hemorrhagic; dose-response present
Second Korean nationwide analysis	Korean adults using alpha-GPC	Prescribed alpha-GPC	Not stated here	Delayed conversion to dementia in users — a finding in the opposite direction
Sagaro et al. 2023 meta-analysis (J Alzheimers Dis)	8 studies, 861 participants with dementia (Russia, Italy, Mexico)	1,200 mg/day, usually 400 mg three times daily	Varies by trial	Monotherapy vs placebo: MMSE mean difference 3.50 (95% CI 0.36–6.63). Added to donepezil: MMSE MD 1.72, ADAS-Cog MD −5.76

Two things are worth reading off that table. The efficacy evidence and the safety signal come from completely different populations and study designs — a set of small dementia trials on one side, a national insurance database on the other. And the efficacy trials used 1,200 mg per day, four to eight times what a consumer nootropic blend typically contains. Sports and acute-cognition studies use 300–600 mg.¹

The mechanism, labeled as what it is

Nobody has demonstrated a mechanism by which alpha-GPC causes strokes. There is a hypothesis, and it should be read as a hypothesis: dietary choline is metabolized by gut bacteria into trimethylamine, which the liver converts to TMAO (trimethylamine N-oxide), and TMAO has been proposed as a mediator of atherosclerosis. Alpha-GPC is a concentrated choline donor. That chain of reasoning is plausible and completely unproven as an explanation for the cohort finding.³

The intended mechanism is cholinergic: alpha-GPC raises acetylcholine availability, which is the same target donepezil and the other cholinesterase inhibitors work on from a different angle. That is why the dementia trials pair the two.¹ It is also why stacking alpha-GPC with a prescribed cholinesterase inhibitor is a bad idea without a clinician in the loop — the effects are additive, and so are the side effects.

⚠️ Talk to your prescriber first

If you take donepezil, rivastigmine or galantamine, do not add alpha-GPC, citicoline or huperzine A on your own. These act on the same cholinergic system and the effects stack. Huperzine A is pharmacologically a reversible acetylcholinesterase inhibitor — the same drug class as your prescription — and Cochrane’s assessment is that the evidence is inadequate to recommend it for Alzheimer’s disease in the first place.⁵

What has been documented, and what has not

Keeping these three categories separate is the whole job on a page like this.

Observed in a very large cohort: a 43% higher adjusted rate of stroke over ten years in alpha-GPC users, with a dose-response gradient.² Observational. Confounding by indication is a live concern, not a technicality.
Observed in randomized trials: cognitive benefit at 1,200 mg/day in people with mild-to-moderate Alzheimer’s or vascular dementia, most convincingly as an add-on to donepezil.¹ Those trials were not powered or long enough to detect a stroke signal.
Theoretical only: the TMAO pathway, and any claim about what alpha-GPC does in a healthy 35-year-old. There is no convincing healthy-adult cognitive data for this compound at all — the consumer case for it is an extrapolation from dementia patients.³

The literature is also genuinely conflicted. A separate Korean nationwide analysis has suggested delayed conversion to dementia among alpha-GPC users.⁴ Any honest account cites both results together. Citing the stroke cohort alone, or the dementia-delay analysis alone, produces a distorted picture in opposite directions.

How serious this is

Our read of the safety evidence: Avoid if you have cerebrovascular disease, a prior stroke or TIA, uncontrolled hypertension, or take an anticoagulant. For everyone else it is a genuine open question, and the honest position is caution rather than alarm.

A hazard ratio of 1.43 is a relative number. Without the baseline stroke rate in that population it does not translate into a personal risk figure, and this article will not invent one. What can be said: the study is unusually large, the follow-up long, the dose-response gradient present, and no randomized evidence points the other way on this outcome. Against that sits confounding by indication — people are prescribed alpha-GPC because they have cognitive complaints, and cognitive complaints track vascular disease.

The risk-benefit arithmetic settles it for most readers. If you have dementia and a clinician who has weighed this, the efficacy evidence at 1,200 mg/day is real. If you are a healthy adult who bought a blend because it promised focus, you are accepting an unresolved stroke signal in exchange for an effect never demonstrated in people like you.

What to do instead

If the goal is a choline-linked compound with a cleaner safety record, citicoline (CDP-choline) is the more conservative option: no serious safety issues reported, including in use up to three years, with benefits concentrated in age-related and vascular cognitive impairment rather than in healthy high performers. It also failed to beat standard care in large acute ischemic stroke trials, which the marketing never mentions.⁶ Doses in healthy-adult trials cluster at 250–500 mg/day.

If the goal is simply “the thing with the best evidence per dollar,” two options outrank alpha-GPC on evidence quality and carry no comparable safety signal:

Creatine monohydrate, 3–5 g/day. A meta-analysis of 8 RCTs in 225 healthy participants found a memory effect of SMD 0.29, concentrated almost entirely in adults aged 66–76 (SMD 0.88) and essentially absent in those aged 11–31 (SMD 0.03).⁷ A larger 2024 meta-analysis of 16 RCTs found memory SMD 0.31 and no significant effect on overall cognitive function.⁸
A plain daily multivitamin-mineral, if you are over 65. COSMOS-Mind randomized 2,262 participants for three years and found a benefit on global cognition; COSMOS-Web found a memory benefit at one year sustained through three.⁹¹⁰ It is the only large replicated positive cognitive result in this field, and it came from a supermarket multivitamin.

Neither is a treatment for anything. Both have better evidence behind them than the compound this page is about, and neither has a 12-million-person cohort attached to a vascular outcome.

Signs that would mean stopping

The routine side effects reported with alpha-GPC are constipation, nervousness, heartburn, headache, insomnia, dizziness, rash and confusion.³ Those are reasons to reconsider, not emergencies.

⚠️ Stroke symptoms are a 911 call, not a supplement question

Sudden face droop, arm weakness on one side, slurred or garbled speech, sudden severe headache, sudden loss of vision or balance — call emergency services immediately. Do not wait to see whether it passes, and do not spend the time reading about supplements. This applies whether or not you take anything at all.

What would change our mind

A randomized trial of alpha-GPC with adjudicated cardiovascular and cerebrovascular endpoints, or a cohort analysis in a non-Korean health system that reproduces the dose-response gradient after adjusting properly for baseline vascular risk. A well-designed nested case-control study inside another national database would move us most, because the single biggest weakness of the current finding is confounding by indication — and a study that matches users to non-users on cognitive-complaint status would address exactly that. If two independent systems showed no signal, we would rewrite this page.

Frequently asked questions

Does the stroke risk apply at supplement doses?

Unknown. The Korean cohort captured prescription use, where the standard dose is 1,200 mg/day. Most consumer blends contain a fraction of that. A dose-response gradient was present in the data, which would imply lower exposure means lower risk — but that is an inference from the shape of the curve, not a measured result at supplement doses.

Is citicoline safer than alpha-GPC?

It has a cleaner reported safety record — no serious safety issues in use up to three years — but there is no head-to-head study, and no comparably large cohort has looked at citicoline and stroke. “Fewer reported problems” is not the same as “demonstrated to be safer.”⁶

I already took alpha-GPC for a few months. What now?

Nothing in the cohort data tells you anything about an individual who has already taken it. If you have vascular risk factors, that is worth a conversation at your next appointment along with your blood pressure numbers. It is not a reason for an urgent call.

Why do supplement blends use 150–300 mg when trials used 1,200 mg?

Cost and label space. Alpha-GPC is expensive per gram and blends compete on ingredient count, not ingredient dose. A blend containing an eighth of the trial dose is not a low-dose version of the trial — it is an untested intervention that borrows the trial’s credibility.

Sources

These statements have not been evaluated by the Food and Drug Administration. Nothing here is intended to diagnose, treat, cure or prevent any disease.

This article is information, not medical advice. Talk to a doctor or pharmacist about your own situation, especially if you take prescription medication.