Choline sources compared

Acetylcholine turnover consumes choline, and when a compound or sustained mental effort drives that turnover, a shortfall can show up as a dull frontal headache, fog, or flat mood. Topping up the supply is the standard response. It is worth being precise about the status of that claim: the mechanism is plausible and widely repeated, but there is no controlled trial demonstrating that a racetam causes headache by depleting choline, and none demonstrating that a choline source prevents it. Treat it as a reasonable hypothesis rather than an established fact.

What is established is that most people are short. Analyses of population intake put roughly ninety percent of Americans below the Adequate Intake, including most pregnant and lactating women, and the food patterns recommended by the Dietary Guidelines are themselves insufficient to reach it in most age and sex groups ^[6]. Because choline is concentrated in animal foods, vegetarians and vegans sit at greater risk.

Food concentrations are lopsided rather than uniform. Measured by liquid chromatography and mass spectrometry across 145 common foods, total choline per 100 grams runs to about 418 mg in beef liver, 290 mg in chicken liver, 251 mg in eggs, 152 mg in wheat germ, 125 mg in bacon, 116 mg in dried soybeans and 103 mg in pork ^[5]. A person who eats eggs and meat is usually fine. A person who does not is usually not, and no supplement decision on this page matters as much as that one.

The requirement also varies genetically, which is unusually well characterised for a nutrient. In a controlled depletion study, 78% of carriers of one common promoter variant in the PEMT gene developed organ dysfunction on a low-choline diet, with an odds ratio of 25, while two variants in choline dehydrogenase moved susceptibility in opposite directions ^[4]. PEMT is the enzyme that lets the body synthesise some of its own phosphatidylcholine, so a variant that lowers its expression makes dietary choline non-optional. This is why blanket dosing advice performs badly here.

Two numbers decide whether a source is any good, and only the first one appears on labels.

The first is choline by molecular weight, which is simple arithmetic: the choline cation weighs 104 g/mol, so its share of any salt or ester is fixed. Choline chloride is about 75% choline; choline bitartrate about 41%; alpha-glycerylphosphorylcholine about 40%; citicoline about 21% as the free base; phosphatidylcholine around 13%, depending on which fatty acids it carries. A gram of bitartrate and a gram of Alpha-GPC therefore contain almost identical amounts of choline, which is not what most product marketing implies.

The second number is the one that matters more, and it is a transport constant. The blood-brain barrier runs its own choline transporter, distinct from the high-affinity transporter on cholinergic nerve terminals. Measured by in situ rat brain perfusion, it behaves as a single saturable system with a Km of 39 to 42 micromolar and a Vmax of 2.4 to 3.1 nmol/min/g, does not require sodium, and is inhibited by hemicholinium-3 with a Ki around 57 micromolar. Its affinity is five to tenfold better than earlier in vivo estimates suggested, but still far weaker than the neuronal high-affinity system, which operates at 1 to 5 micromolar ^[7].

Saturable is the key word. Once that transporter is working near capacity, adding more free choline to the blood buys progressively less brain choline, which is the ceiling every plain choline salt runs into. It is also why the useful comparison between sources is not how much choline they contain but whether they arrive in a form that bypasses or supplements that route.

Age moves the ceiling down. When young and older adults were both given a choline load of 50 mg/kg, plasma choline rose by a similar proportion in each, around 76% and 80%. Brain cytosolic choline compounds measured by proton magnetic resonance spectroscopy rose by 60% in the young group and only 16% in the older one ^[8]. Identical delivery to the blood, very different delivery to the brain. That single result is the strongest argument for using an efficient form in an older reader, and the strongest argument that a young person with a good diet may be getting very little from any of this.

Alpha-GPC is L-alpha-glycerylphosphorylcholine, a natural intermediate in phosphatidylcholine metabolism. It is the source with the clearest positive cholinergic trial and the one most often chosen when the goal is acetylcholine specifically.

CDP-choline, also called citicoline, is cytidine 5'-diphosphocholine. It is hydrolysed before absorption and reassembled in tissue, which is why it delivers two things rather than one: choline, and cytidine, which humans convert onward to uridine. Its clinical literature is built around stroke, traumatic brain injury and vascular cognitive impairment rather than healthy cognition ^[2].

Choline bitartrate is choline paired with tartaric acid. It is cheap, it is a perfectly good way to correct a dietary shortfall, and it is the form most likely to run straight into the saturable barrier transporter described above. The review literature on cholinergic precursors in dementia is blunt about the plain forms: well-controlled trials did not confirm clinical utility for choline or lecithin, while the phospholipid forms did show modest benefit ^[10][11].

Phosphatidylcholine, sold as lecithin, carries the least choline per gram and has the weakest record; it is also the form most implicated in the trimethylamine pathway discussed below ^[17].

Two things commonly shelved next to these are not choline sources at all. Betaine is a choline metabolite, useful as a methyl donor, and it cannot be converted back into choline. DMAE, dimethylaminoethanol, is a structural near-relative of choline that is not efficiently converted to it; whatever DMAE does, treating it as a choline supplement misdescribes it.

Finally, an honest note on centrophenoxine, which appears on most lists of this kind. It is a DMAE ester rather than a choline ester, its antioxidant claims come mainly from old lipofuscin work, and it has no modern controlled trial in healthy adults.

Alpha-GPC has the single best trial in this whole category, and it is not in healthy people. Two hundred and sixty-one patients with mild to moderate Alzheimer's dementia were randomised to 400 mg of choline alfoscerate three times daily or placebo for 180 days. The ADAS-Cog score fell by 3.20 points on drug and rose by 2.90 on placebo, with the Mini-Mental State Examination, Global Deterioration Scale and Clinical Global Impression all separating in the same direction ^[10]. The ASCOMALVA trial then tested it as an add-on: donepezil plus choline alphoscerate against donepezil plus placebo in Alzheimer's disease with documented ischaemic damage. At twelve months, in an interim analysis of 91 of a planned 210 patients, the combination did better on the MMSE, ADAS-Cog, instrumental activities of daily living and caregiver distress ^[14].

In healthy adults, the Alpha-GPC evidence is much thinner. Forty-eight college-aged men took 250 mg or 500 mg of Alpha-GPC, 200 mg of caffeine or placebo daily for a week. Serum free choline rose sharply. There was no difference in isometric strength or in psychomotor vigilance, and the only separation was on countermovement jump velocity and mechanical power, best in the 250 mg group ^[15]. One small trial, mostly null, with a non-monotonic dose response.

Citicoline has the broadest literature and the most mixed verdict. The Cochrane review pooled fourteen randomised placebo-controlled trials in older adults with chronic cerebral disorders and found evidence of benefit on memory and behaviour, stronger evidence on global impression, and no effect on attention, with heterogeneity in dose, duration and population large enough that the reviewers declined to say much more ^[12]. Mechanistically it does something measurable: 500 mg daily for six weeks raised brain phosphodiesters by 7.3% on phosphorus magnetic resonance spectroscopy in healthy older subjects, with the rise correlating with improvement on a verbal learning test; a second six weeks added nothing further ^[9].

And it has the largest outright failure in the category. ICTUS randomised 2,298 patients with moderate to severe acute ischaemic stroke to citicoline or placebo and was stopped for futility, with an odds ratio for global recovery of 1.03 ^[13]. That result matters here because a great deal of citicoline's reputation was built on earlier pooled stroke analyses that ICTUS did not confirm.

The reason citicoline is credited with more than its choline content is the cytidine half. Feeding gerbils uridine plus choline plus docosahexaenoic acid raised brain phosphatidylcholine by 45%, other phosphatides by 39 to 74%, and synaptic proteins including synapsin-1 and PSD-95 by around 40%, substantially more than any single precursor alone ^[18]. That is a real and specific mechanism for membrane synthesis; it is also an animal study, and it is the argument for pairing citicoline with uridine monophosphate rather than evidence that the pairing works in people.

Start with food. If eggs, liver or meat are absent from the diet, a plain choline supplement is correcting a real nutritional gap and any of these forms will do it ^[5][6]. If they are present in reasonable quantity, the marginal value of everything below drops sharply.

If the goal is a clear cholinergic lift, Alpha-GPC is the defensible choice, because it is the only source with a large positive randomised trial on a cognitive endpoint ^[10] and the only one shown to raise serum free choline by a specified amount at a specified dose ^[15]. Note carefully that the strong trial was in dementia, where there is a deficit to correct, and read the risk section below before committing to daily use.

If the goal is membranes, recovery after injury, or a gentler all-round option, citicoline has the larger and more varied literature, and the extra cytidine gives it a second mechanism that plain choline does not have ^[2][18]. Its record is best in damaged brains and unconvincing in acute stroke ^[13].

Skip plain bitartrate for cognition and keep it for nutrition. It costs almost nothing, it fixes a dietary shortfall, and the controlled record on plain choline and lecithin for cognitive outcomes is negative ^[10][11].

Age tilts the decision. Brain uptake of a choline load falls markedly with age ^[8], so the argument for an efficient form gets stronger the older the reader is, and correspondingly weaker for a young person already eating eggs.

On pairing with a racetam: the practice is near-universal and the evidence for it is absent, as noted above. If a racetam produces a headache, adding a choline source is a cheap and low-risk thing to try, and it is a trial of one rather than a proven fix. See nootropic stacking for the general problem with untested combinations.

Cholinergic overload is real and unpleasant. Stacking a choline source with an acetylcholinesterase inhibitor such as huperzine A or donepezil increases supply and blocks breakdown at the same time, and the result can be headache, nausea, sweating, cramps, bradycardia and a distinctly heavy head. The cholinergic system rewards balance rather than maximisation.

The trimethylamine problem applies to all of them. Gut bacteria metabolise dietary choline and phosphatidylcholine to trimethylamine, which the liver oxidises to trimethylamine N-oxide. In a large clinical cohort, plasma levels of choline, TMAO and betaine each predicted cardiovascular risk; in mice, dietary choline or TMAO promoted atherosclerosis, germ-free animals did not generate TMAO, and suppressing gut flora abolished the effect ^[17]. This is also the chemistry behind the fishy body odour that appears at very high choline intakes, which is the classic sign of overshooting.

The Alpha-GPC stroke signal deserves to be stated plainly, and then qualified just as plainly. A retrospective cohort drawn from South Korea's national health insurance data covered 12,008,977 people aged 50 and over. The 108,877 who had been prescribed Alpha-GPC had an adjusted hazard ratio of 1.43 for total stroke over the following ten years after matching on all measured covariates, 1.34 for ischaemic and 1.37 for haemorrhagic stroke, with a dose-response by duration of prescription ^[16]. The qualification matters: this is observational data in a population prescribed a drug for cognitive complaints, and confounding by indication is the obvious alternative explanation, since early cognitive decline and cerebrovascular disease travel together. Matching reduces that concern without eliminating it. What it is not is a reason to ignore the finding; it is a large signal in a very large study, and it belongs in front of anyone taking a gram a day for years.

Dose sensibly and know what the trials used. The positive Alzheimer's trial used 1,200 mg of Alpha-GPC daily ^[10]; the healthy-volunteer study used 250 to 500 mg ^[15]; the citicoline brain-imaging study used 500 mg ^[9]. Nothing in this literature supports the multi-gram doses that circulate informally. None of this is medical advice.