What the Choline Research Actually Measured, and In Whom

Two numbers circulate in the marketing for alpha-GPC. One is around 300 mg. The other is 1,200 mg, usually introduced as "the clinically studied dose". Both are real and both come from published trials, and they are not on the same scale — because they come from trials run in two different kinds of people.

That distinction is the entire subject of this piece. It is the difference between a figure describing what a compound did in people with a diagnosis and a figure describing what it did in people without one — the most common error in this category, made by companies that are otherwise careful, and made by this desk in its own earlier drafts.

• 250–630 mgThe doses given in the small alpha-GPC trials that recruited healthy adults, mostly as a single dose
• 1,200 mgThe daily dose in the trials most often quoted — all in diagnosed populations
• 48Young men in a randomised comparison of 250 mg against 500 mg

Alpha-GPC — L-alpha-glycerylphosphorylcholine, also written choline alfoscerate or alphoscerate — has been studied for decades. In much of Europe it is a prescription medicine rather than a supplement, and that history explains the shape of the evidence base: most of the money went into trials in patients.

The reviews make this visible at a glance. A 2025 systematic review comparing choline alphoscerate with citicoline pooled three randomised trials totalling 358 participants — all of them patients with dementia disorders (Sagaro & Amenta, Frontiers in Neurology, 2025; n=358). An earlier review of adult-onset cognitive dysfunction analysed seven randomised trials and one cohort study, again in dementia and cerebrovascular injury populations (Sagaro, Traini & Amenta, Journal of Alzheimer's Disease, 2023).

The individual trials behind the 1,200 mg figure carry the same signature. A randomised trial of 1,200 mg a day over 12 months was run in 60 patients with mild cognitive impairment and associated vascular damage (Carotenuto et al., Medicina, 2024; n=60). A study running the same daily total as 400 mg three times a day, across four three-month courses over five years, recruited 122 first-degree relatives of patients, all with subjective cognitive impairment (Selezneva et al., Zhurnal Nevrologii i Psikhiatrii, 2026; n=122).

Nothing about that literature is disreputable. It is a substantial body of clinical research on a compound used as a medicine. The problem begins where a number lifted out of it is printed on a bottle sold to someone with no diagnosis at all.

Here is the part supplement copy almost never states.

The best-known randomised dose comparison in healthy adults gave 250 mg or 500 mg of alpha-GPC to 48 college-aged men and measured physical and psychomotor performance; the 250 mg group showed the greatest improvements on the jump measures (Marcus et al., Journal of the International Society of Sports Nutrition, 2017; n=48). An earlier single-dose trial compared alpha-GPC, caffeine and placebo on markers of mood, cognitive function, power, speed and agility (Parker et al., same journal, 2015). It found no significant benefit for alpha-GPC over placebo on any of them — the negative result in this literature, and the one least often quoted.

Three features stand out:

1. It is small. Dozens of participants, not hundreds.
2. It is short. Single doses and days, against six and twelve months in patients.
3. It is mostly sports nutrition. Jump height, reaction time, power and agility — not word retrieval, not working memory over weeks.

The 250–630 mg range that appears in careful copy describes what those trials dosed; it is not evidence that any dose within it improves cognition in people without a diagnosis.

The most interesting detail in that literature is the one most likely to be left out: in the only randomised dose comparison, the lower dose produced the larger effect on the jump measures. More was not better. Any argument for a bigger number that cites this trial is citing a trial that points the other way.

The intuitive objection is that a dose is a dose: if 1,200 mg was safe and effective in a trial, why is quoting it to a general buyer misleading rather than merely optimistic? Three reasons, none of them a technicality.

The baseline is different, and that changes what "improvement" means

Trials in impaired populations measure movement on scales built to detect impairment, so a change describes partial recovery of a deficit. There is no equivalent construct without the deficit — a normal score cannot become more normal in the same direction. An effect size does not transfer between the two settings even in principle, and the trials were never designed to let it.

The endpoints are different

Clinical trials use instruments designed for diagnosis and staging. The healthy-adult trials measure reaction time, jump height and agility. These are not the same measurement at different sensitivities; they are different measurements, and a result on one says nothing about the other.

The safety context is different

A trial dose is administered under supervision, with exclusion criteria filtering out people for whom it would be inappropriate. The same number on a pack is taken unsupervised by whoever bought it. The number transfers; the protections around it do not — which is why supplement labels carry a warning to consult a physician before use for anyone pregnant or nursing, taking medication, or managing a medical condition.

Publishing the study that did not work is the test of whether a page is a review or an advertisement, so here is the relevant one.

A pilot trial gave 20 postmenopausal women a single 1,650 mg oral dose of choline bitartrate three hours before functional MRI scanning during a working-memory task. Against placebo, the choline condition showed decreased working-memory-related brain activation in regions of the right temporal lobe (Dumas et al., Nutrients, 2026; n=20).

Now the honest reading, because a result like this is easy to stretch in both directions.

It is frequently cited as proof that choline bitartrate "does not reach the brain" or "does not work". It shows neither: the compound clearly did something measurable. What the trial does not deliver is a demonstration that bitartrate improves cognitive performance in people without a diagnosis, and that absence is fair to report.

It is also one pilot study, in 20 people, from one specific population, after one dose. Quoting it as though it settled the question about healthy adults in general would be exactly the population error this page is about — which is why the population is named every time the study appears here.

All of the above is academic unless it changes what a buyer does in a shop, so this desk went and counted.

Method. The National Institutes of Health maintains the Dietary Supplement Label Database, a public archive of marketed supplement labels. On 14 August 2026 we pulled every on-market label whose ingredient rows name choline, removed duplicate listings by brand and product name, and counted the chemical form each declares. That left 635 distinct labels; one label can name more than one form.

Ten labels out of 635. That is the gap between what the research studied and what the shelf sells, and anyone can reproduce the count against the same public database.

Two caveats belong with it. The database archives labels rather than sales, so a form on few labels may still sell in volume; and an archive lags the market. Neither changes the shape of the finding.

The form is declared inside a parenthetical on the Supplement Facts panel, and finding it takes about ten seconds: locate the ingredient row, read the bracketed name that follows it, and check that the milligram figure sits beside the nutrient itself rather than beside the raw material it was made from.

The four sentences worth carrying away

The 1,200 mg figure in supplement marketing comes from trials in people with diagnosed cognitive impairment. Trials recruiting healthy adults dosed roughly 250–630 mg, are few and small, and mostly measured physical and psychomotor endpoints. In the one randomised dose comparison, the lower dose performed better. And on the shelf, 10 of 635 labels naming choline name the form most of those trials used.

The correct response is not to buy something. It is to read the parenthetical on whichever pack is already in the cupboard, and to notice whether the number on the front came from a population that resembles the buyer.