Amino acids and ADHD: why the studies give no answer

Dopamine is built from tyrosine, serotonin from tryptophan. From this follows an obvious question: are these building blocks missing in ADHD, and if so, which ones? Online there are very clear answers. Amino acid profiles are offered as diagnostics, tyrosine and taurine are recommended as supplements, and single percentages travel from one page to the next.

I looked at the work behind this, every single study in the original. What emerges has less to do with your values than with the way those values were measured. In this article we look at what was actually studied, why most findings vanish when you recalculate, and why the existing studies probably cannot answer your question at all.

Are amino acids missing in ADHD?

A consistent amino acid deficiency typical of ADHD is not established. The largest and methodologically cleanest study found normal values in the blood, in the urine and in protein intake of 83 children with ADHD. The studies that report abnormalities are smaller and calculated differently.

The narrative goes back to a single piece of work. In 1990 Bornstein and colleagues examined the blood of 28 children with attention disorder against 20 comparison children and found five amino acids lowered: phenylalanine, tyrosine, tryptophan, histidine and isoleucine. Their own interpretation in the abstract: a general deficiency in the uptake or transport of amino acids (Bornstein et al., 1990).

Two things about this before the finding travels on. Apart from the case numbers, the abstract contains not a single figure, no p-value, no effect size and no statement of which amino acids were unchanged. And the second paper from the same group, often cited as confirmation, examined an overlapping sample (Baker et al., 1991). Bergwerff and colleagues explicitly describe the two publications as two papers that describe the same sample (Bergwerff et al., 2016). It is one finding that was published twice.

Against this stand three newer studies:

Bergwerff 2016 examined 155 children between 6 and 13 years old, 83 of them with ADHD. Measurement was in dried blood, additionally in 18-hour urine, and protein intake was recorded with a food diary. The result: normal levels of the aromatic amino acids in blood and urine and a normal protein intake compared with the control group. The odds ratios for below-average values all went in the same direction, none reached significance: phenylalanine 2.22 (p = 0.07), tryptophan 2.09 (p = 0.11), tyrosine 1.83 (p = 0.22).

Hung 2025 compared 70 children who had never received an ADHD medication with 70 children matched for sex, age and IQ. 156 metabolites were measured. Four of them stood out: cholic acid, homoveratric acid, inosine and nicotinuric acid. Not a single amino acid.

Johansson 2011 examined transport in connective-tissue cells from 14 boys with ADHD. For tryptophan the transport capacity was reduced, the binding affinity by contrast increased, and for alanine the capacity was greater than in the controls. Tyrosine did not differ.

So a finding from 1990 stands against three later studies, among them the one with the best comparison method and the only one that corrected statistically for the number of measured values.

Why the studies report many hits and still show nothing

Here lies the real reason the field looks so contradictory, and it has nothing to do with ADHD.

A plasma amino acid profile measures 20 to 22 values at once. At the usual significance threshold of 5 per cent, about one abnormal result per study is expected by chance alone, even when there is no difference between the groups. Anyone who tosses 22 coins inevitably gets a few striking runs. The only thing that helps against this is to adjust the threshold to the number of comparisons.

Study	values measured	hits reported	p-values	correction
Bornstein 1990	full profile	5	not reported	none detectable
Skalny 2021	22	4 to 5	0.018 to 0.045	only within the ANCOVA, not across the 22 values
Johansson 2011	several	3	0.010 to 0.039	none detectable
Bergwerff 2016	3 aromatic	0	all above 0.05	yes, threshold 0.004

With Skalny’s 22 values, the adjusted threshold would be p = 0.0023. The lowest value reported there is 0.018, about eight times above it.

Of all the profile studies, exactly one corrected across all measured amino acids. It is at the same time the only one with a null result. There is a simple explanation for this: once you calculate correctly, nothing remains.

What happens when you look in adults

And here comes the point that is probably the most important for you. The entire profile literature on ADHD is a children’s literature. A complete plasma amino acid profile in adults with ADHD does not exist. Bornstein, Bergwerff, Skalny, Hung, Johansson: all children between 6 and 14 years old.

That does not mean nothing exists on adults. Individual metabolic pathways have certainly been measured there. What is missing is the measurement that would answer your question, namely the complete profile compared with a control group.

And where adults have been examined, something different shows up than in children. The meta-analysis of tryptophan metabolism evaluated 8 studies with 1,177 participants. In the overall analysis, tryptophan does not differ between ADHD and controls (SMD 0.08; p = 0.68). The variation between studies, however, is enormous: I², the measure of how strongly the individual studies diverge, is 85.7 per cent, where 0 per cent means complete agreement and 100 per cent maximum inconsistency. In the subgroup of medication-free children, tryptophan is then raised (SMD 0.31; 95% CI 0.11 to 0.50; p = 0.002), and this finding is statistically clean with I² = 0 per cent (Cavaleri et al., 2024). In adults the depletion studies point the other way (Zimmermann et al., 2012).

This reversal of direction is one reason not to transfer the children’s results to adults. When the same quantity comes out in opposite directions in two age groups, the average over both says nothing about either one.

For you as an adult with ADHD this means two things. The question is legitimate, and the existing studies answer it for eight-year-olds.

Amino acid by amino acid

Tyrosine. Not reproducibly lowered, and that is remarkable, because the entire dopamine-precursor narrative rests on tyrosine. One finding from 1990 against three negative studies. Skalny did not find tyrosine among the changed values, Hung did not find it among 156 metabolites.

Tryptophan. No deficit in the meta-analysis (p = 0.68), raised in medication-free children. The causal test has also been done: 14 studies specifically depleted tryptophan. If a tryptophan deficiency were a driver of ADHD symptoms, they would have to worsen during this. Only two of these works examined attention at all, with mixed results, and not a single one measured hyperactivity (Dinu et al., 2022). Controlled studies that give tryptophan in ADHD do not exist.

Histidine. The most-cited single value in the whole field, and the most important correction of this research. Skalny reports a serum value 29 per cent lower in children with ADHD. The associated p-value stands at 0.081. The study itself defines 0.05 as the threshold in its methods section. The difference is therefore not significant there, and the 29 per cent stand in the abstract without a p-value. There is no third histidine study, no intervention studies either, and the only measurement of histamine in ADHD so far found no association with inattention or hyperactivity (Bruton et al., 2024).

Glutamine and glutamate. Contradictory, and the methodologically strongest works are negative. Apart from Skalny (p = 0.024, uncorrected across 22 tests), glutamine was not measured in any further dedicated blood study. For glutamate, the largest study with 145 against 120 participants found no difference (p = 0.059; Cheng et al., 2026); Hung does not find it among his results.

GABA. The most solid blood finding in the whole field, and it argues against supplements. In serum, GABA is raised in ADHD, with p < 0.001 and a discriminatory power of 0.87; it correlates with symptom severity and falls under methylphenidate (Cheng et al., 2026; Miniksar et al., 2022). In the brain the finding is unremarkable in meta-analysis (Schür et al., 2016). On top of this comes the blood-brain barrier: it actively transports GABA outward, and the efflux clearance is at least 16 times higher than the influx clearance (Kakee et al., 2001). A value that is raised in the blood and rises with symptom severity, plus a barrier that actively pumps it out, is no argument for GABA capsules. Anyone wanting to work on the excitation-inhibition balance finds the better-supported point at magnesium, which blocks the NMDA receptor: magnesium in ADHD.

Taurine. The biggest nothing in the field. There is not a single study that has measured taurine in people with ADHD and compared it with a control group, and none that has tested it as a treatment. Skalny measured taurine along the way, it is in the list of 22 values, and it is not among the changed ones. A large part of the “taurine improves attention” literature comes from energy-drink studies in which taurine was given together with caffeine and sugar. There the effect cannot be attributed to the taurine.

Protein intake. Two independent studies with direct dietary assessment find no difference (Bergwerff et al., 2016; Holton et al., 2019). The often-told chain from low protein intake via low tyrosine to low dopamine is already unsupported at its first step.

The one finding that holds

For all the criticism, something remains, and it is more interesting than what is usually told.

Of all the values in tryptophan metabolism, exactly one is stable across the studies: kynurenic acid is lowered (SMD −0.33; 95% CI −0.49 to −0.17; p < 0.001), and with a variation between studies of I² = 0 per cent. For comparison: for the raised kynurenine, which is cited more often, the same figure is 90.3 per cent, so the studies there contradict each other considerably (Cavaleri et al., 2024). The finding was replicated in children who had never received an ADHD medication, and there tryptophan, kynurenine and the ratio of the two were all unchanged (Ünal et al., 2026).

Kynurenic acid is formed from kynurenine via a particular enzyme. When the raw material is present at normal levels and the end product is missing, the narrow point lies at this enzyme. That is a different statement from “tryptophan is missing”, and it cannot be captured by an amino acid profile.

One limitation belongs here. The study most often cited for a shift in metabolism in adults measured eight values at once and did not correct. There kynurenine itself was not significant (p = 0.15), and after accounting for the tryptophan level only three values remained. On top of this comes a comparison problem: 66.2 per cent of the patient group were smokers by nicotine level against 12.2 per cent of the control group, and smoking lowers tryptophan (Aarsland et al., 2015).

Would more of it even help?

Suppose a building block really were missing. Then “more of it” would still not be the answer, and this is well studied in healthy adults.

The review on tyrosine summarises the conditions like this: it works only when the transmitter system is intact and dopamine or noradrenaline are temporarily reduced (Jongkees et al., 2015). Two prerequisites at the same time, and the evidence for them comes from cold exposure, from staying awake for over 24 hours, and from a military training course.

The opposite direction is the more revealing part:

• In a double-blind study with 45 people and 2 grams of tyrosine, the weaker participants improved, the stronger ones got worse. The group average did not change, the effect showed only in the spread (Jongkees, 2020).
• In 70 people, tyrosine worsened mental flexibility under high cognitive load. The authors call this themselves the opposite of what they had expected (Robson et al., 2020).
• In older adults, working memory worsened in a dose-dependent way, most clearly in those with the highest rise in the blood (van de Rest et al., 2017).

Tyrosine therefore shifts the position on a curve with an optimum. Anyone already well placed slides past it.

The only tyrosine study in ADHD shows the same pattern over time. Twelve adults received L-tyrosine for eight weeks; eight of them improved after two weeks, and after six weeks the effect had disappeared. The authors write that L-tyrosine is without benefit in attention disorder (Reimherr et al., 1987). In supplement marketing this regularly becomes “8 of 12 responded”. A randomised controlled trial of L-tyrosine in ADHD does not exist to this day.

One note belongs here. The most frequently cited work for amino acid precursors in ADHD, reporting 77 per cent improvement, was retracted in 2020. It was a retrospective chart review, and the first author himself sells corresponding supplements (Hinz et al., 2011, retracted 2020).

Taurine, glycine, tyrosine: what is proven for the supplements

Up to here it has been about status, that is, the question of whether something is missing. The second question is a different one: does one of these amino acids work as a supplement, regardless of whether a deficiency is present? Does taurine calm, does tyrosine make you briefly more alert, does glycine relax? This question is better studied than the first, and the answers are more concrete.

One pattern runs through all three. Taking it away is proven in humans, adding it is not. For tyrosine there is a study that measured in the brain scanner what a tyrosine depletion does: the amphetamine-triggered dopamine release in the striatum falls in eight men, in a linear relationship with the drop in the blood (r = −0.82; Leyton et al., 2004). The same shows via the prolactin level (Harmer et al., 2001). A study that measures a rise in dopamine release in humans after giving tyrosine does not exist.

From “taking it away lowers” it does not follow, for this enzyme, that “adding it raises”. Tyrosine hydroxylase is braked by its own end product; dopamine binds with high affinity and blocks it until the neuron fires and the brake is released (Gordon et al., 2008). The specialists from whom the whole precursor narrative originates limit their own finding accordingly: more tyrosine increases catecholamine production only in nerve cells that are currently active (Fernstrom & Fernstrom, 2007).

On top of this comes a curve that explains the rest. In animal experiments, tyrosine levels and synthesis rate in the brain rise as the dietary protein content increases from 2 to 10 per cent. Between 10 and 20 per cent, nothing more happens (Fernstrom & Fernstrom, 1995). The synthesis rate rose only in the retina and hypothalamus, not in the prefrontal cortex, that is, precisely not in the region that matters in ADHD. Anyone eating normally is already sitting on this plateau.

Taurine: the blood-brain barrier is the point, not the study record.

The transporter that brings taurine into the brain has been measured in a human model. It runs half-saturated at about 19 micromoles per litre (Tega et al., 2025). After 4 grams of taurine, the peak level in the blood is around 690 micromoles per litre (Ghandforoush-Sattari et al., 2010). The transporter is therefore occupied far beyond its saturation point and already runs near its maximum at normal levels. In addition there is an active pump that carries taurine out of the brain (Lee & Kang, 2004), and when taurine is in excess the transporter is turned down at the level of gene expression (Kang et al., 2002). The half-life in the blood is about one hour.

There is no published evidence anywhere that oral taurine measurably raises taurine in the human brain, neither via magnetic resonance spectroscopy nor via the cerebrospinal fluid level. And there is a direct counter-indication: in one patient group, taurine was raised in the blood and unchanged in the cerebrospinal fluid (Samuelsson et al., 2013).

The study record fits this. A review of all eight randomised studies on taurine and cognition in humans concludes that acute doses produced at best small and inconsistent improvements and that most endpoints remained unchanged (Moore et al., 2026). The meta-analysis of 7 studies with 402 people finds no effect (Cao et al., 2025). On sleep there is not a single randomised study, on ADHD not a single intervention study. The review on mood counts one human study against 13 animal studies, and the one human study is an observational study with no association (Pedregosa et al., 2026).

And even in the animal model the finding does not hold reliably. One work that tested this specifically found no effect on anxiety- or depression-like behaviour and no stimulating effect (Whirley & Einat, 2008).

Where the calming claim then comes from can be traced back. It begins with a correct finding. Taurine does indeed act at the glycine and the GABA-A receptor; that has been measured since the eighties. The half-maximal effective concentration at the glycine receptor is 406 micromoles per litre (Hussy et al., 1997), at some preparations 1,250 (Jiang et al., 2004). Taurine is moreover only a partial agonist there, reaching at most 70 per cent of the maximal glycine response (Hussy et al., 1997). In the brain, the actual concentration between the cells is 25 micromoles per litre, measured in the rat striatum (Molchanova et al., 2004). The receptor experiments therefore run far above anything that ever occurs in the head.

The second step is the route of administration. The animal works that show a dampening effect gave taurine directly into the brain. And one work compared both routes in the same model. The result: injected into the brain, taurine lengthened sleep; swallowed, it shortened it (McBroom et al., 1986). The authors name the reason themselves, namely the slow uptake into the brain. At the same time, other groups were building fat-soluble taurine derivatives, expressly because it was assumed they would cross the blood-brain barrier (Ahtee et al., 1985). The problem was known in 1985.

The third finding goes beyond the mere absence of an effect. In mice, taurine injected once is anxiolytic, given continuously in the drinking water by contrast anxiogenic (El Idrissi et al., 2009). The reason lies in adaptation: under constant taurine supply, the GABA-A receptors are downregulated (L’Amoreaux et al., 2010). A regulated system evades the constant stimulus.

Two things round this off. Oral taurine has long been tested in humans at an endpoint of brain excitability: in the seventies a series of epilepsy studies ran with up to 8 grams daily. The last of them gave six patients 375 to 8,000 milligrams per day and found no improvement in seizure control (Mantovani & DeVivo, 1979). This literature is 45 years old and is cited nowhere. And with the European food authority, six health claims were submitted for taurine for assessment, among them immune system, heart, muscle and cognition. Not a single one on relaxation, anxiety or sleep (EFSA, 2011). The effect for which taurine is sold today was never submitted for assessment.

For accuracy’s sake, two transport studies contradict each other on whether the active transporter sits on both sides of the vessel wall or only on the side facing the brain (Tamai et al., 1995; Rasgado-Flores et al., 2012). This changes nothing about the overall picture.

Two things belong to accuracy. First, the only place where taurine does something calming-like in humans is the comparison with caffeine: in the two studies that tested taurine in isolation, it dampens the caffeine effect rather than stimulating on its own (Giles et al., 2012; Peacock et al., 2013). Both works attribute the entire energy-drink effect to the caffeine, and neither measured anxiety. Second, there is one positive study: 4 grams of taurine over twelve weeks improved the symptoms in first-episode psychosis (p = 0.004). That comes with 29 per cent dropouts, a second primary endpoint on cognition with no effect (p = 0.582) and a population that has nothing to do with this question (O’Donnell et al., 2016).

Glycine: the source record is the problem.

The two studies on which the story of sleep-promoting glycine rests are not findable in the medical databases (Yamadera et al., 2007; Inagawa et al., 2006). No abstract, no p-values, no conflict-of-interest statement. The participant numbers were 11 and 19.

All primary works come from a single company, the largest amino acid manufacturer in the world. The only fully verifiable of these studies has seven evaluated participants, is only single-blind, was approved by the company’s own ethics committee and states in the text that there is no commercial relationship (Bannai et al., 2012). The best available review summarises the sleep evidence like this: small samples with a high risk of bias, and a meta-analysis was not possible because of the inconsistency (Soh et al., 2024).

One work with real laboratory sleep measurement tested glycine-rich collagen peptides in 13 men. Sleep onset, sleep amount, sleep efficiency and core body temperature remained unchanged; the number of waking events fell. This work, too, was carried out with a manufacturer’s involvement (Thomas et al., 2024).

The finding that settles the matter, however, is a different one. A double-blind study gave glycine by vein, 7 to 14 grams, and afterwards measured the cerebrospinal fluid. The level there rose dose-dependently without producing any clear effect in the central nervous system (D’Souza et al., 2000). It arrived in the brain and did nothing measurable there. If the multiple dose is ineffective even when bypassing the digestive tract, a noticeable effect from 3 grams out of a can is hard to justify.

The outcome of the large clinical trial fits with this. In schizophrenia research, glycine was tested in high doses; small studies had reported improvements of 23 to 30 per cent. The adequately powered study with 157 participants found no difference from placebo either in symptoms or in cognition, and this under the co-authorship of the very researchers who had previously published the positive small studies (Buchanan et al., 2007).

The time axis is the part almost no one mentions.

In ADHD, amino acids were given over a longer period twice, and both times the same thing happened. Under L-tyrosine, eight of twelve adults improved after two weeks; after six weeks the effect was gone (Reimherr et al., 1987). Under DL-phenylalanine, a double-blind effect on mood appeared, and in the subsequent open continuation the benefit disappeared completely within three months (Wood et al., 1985). Two different amino acids, the same pattern: an initial effect, then regulation catches up with it.

The most recent work on this is a randomised controlled trial of 5-HTP, the direct precursor of serotonin, in 112 adults with pronounced ADHD traits. Result: no benefit on any measure of distractibility, the only significant result ran in the wrong direction, and 19.6 per cent had side effects with fatigue, nausea or vomiting (Jackson et al., 2026).

	Randomised trial with a fitting endpoint	Evidence in the human brain	Tested in ADHD
Tyrosine	only under acute stress or exhaustion	only for depletion	open trial, effect lost after 6 weeks
Taurine	none on anxiety or sleep	none	none
Glycine	three small ones, all from the manufacturer	rise without effect	none

What this does not mean. A lack of evidence is not evidence of ineffectiveness. Taurine research in humans consists almost entirely of single doses in healthy people; longer administration with psychological endpoints is practically untested. At 3 grams daily, taurine is well tolerated, and a self-experiment is defensible. What cannot be derived from these data is a promise of effect.

Does a stool test show whether I absorb protein poorly?

This question is legitimate, because the one positive amino acid study points there itself. Bornstein interpreted his values as an uptake or transport disorder. The answer consists of two parts.

First, the hypothesis has never been tested. No study in ADHD has measured pancreatic elastase in stool, protein loss through the gut, faecal nitrogen or an absorption test with labelled protein. In ADHD, the only things ever examined in stool at all are the microbiome, once the fungal flora and once calprotectin, the latter with no difference in 14 children (Soltysova et al., 2025). That is a real research gap.

Second, a stool test would be the wrong tool for this. A study in 11 patients with confirmed pancreatic insufficiency shows this directly. With labelled protein the absorption disorder could be measured and an improvement demonstrated under enzyme therapy. In the stool none of this was visible: a protein absorption disorder was largely undetectable with the usual stool tests (Airinei et al., 2011).

The reason for this is vivid. The nitrogen in stool comes mainly from bacterial mass, shed intestinal cells and digestive secretions. What the small intestine does not absorb is used up by the large-intestine bacteria before it reaches the stool.

On the permeability values that such profiles come with, a note of its own belongs. The best known of them has been checked several times, and according to two independent studies the test kits used detect proteins other than what is stated on the report (Scheffler et al., 2018; Ajamian et al., 2019). The two larger and better-adjusted studies in ADHD find no difference there anyway (Aydoğan Avşar et al., 2021; Lee et al., 2023).

And what if IBS is also present?

Irritable bowel syndrome is indeed more common in ADHD. An analysis across 11 studies and around 3.85 million people found no increased value for intestinal disorders overall (OR 1.25; 95% CI 0.75 to 2.07), but did for irritable bowel syndrome (OR 1.63; 1.45 to 1.83; Ng et al., 2025). A large cohort confirms this and adds two things: inflammatory bowel disease and coeliac disease are not increased in ADHD, and methylphenidate itself increases digestive complaints and constipation (Kedem et al., 2020). This fits the mechanism: as a stimulant it dampens appetite and digestive activity, the gut becomes sluggish, and often less is drunk as well.

The obvious conclusion would now be that in IBS absorption suffers and from this an amino acid deficiency arises. This chain breaks in the middle. An absorption disorder is an exclusion criterion in IBS: it is the diagnosis you give after malabsorption has been ruled out. Empirically this fits. In an analysis of over 12,000 people, those with IBS actually consumed more calories than healthy people, with no signs of nutrient deficiencies (Hujoel et al., 2020). Where deficiencies are found, the author groups attribute them to reduced intake, that is, to avoidance behaviour and elimination diets.

In two places, however, the objection is right after all:

Missed coeliac disease. In an analysis of 29 studies with 7,209 people with an IBS diagnosis, 2 per cent had biopsy-confirmed coeliac disease, 6 per cent had abnormal antibodies, and 15 per cent of the abnormal ones were never investigated further (Shiha et al., 2025). Coeliac disease damages the small-intestine surface and causes a genuine protein malabsorption.

Missed pancreatic insufficiency. In the diarrhoea form of IBS, stool elastase is below normal in 6.1 per cent of those affected, against 0 per cent in two control groups (Leeds et al., 2010). The pancreas delivers the protein-splitting enzymes; it is the only one of the mentioned mechanisms that acts directly on protein digestion.

In both cases the chain holds completely. But then there is a missed diagnosis and not IBS. From this comes a usable rule. If someone with ADHD and an IBS diagnosis has measurable nutrient deficiencies, the next two questions are: what does this person actually eat, and has coeliac serology ever been done?

What you can do instead

1. An amino acid profile does not answer the ADHD question. Two questions are often mixed here. One is: in people with ADHD, is a particular amino acid typically lowered? A profile gives no answer to this, because for adults with ADHD there is no published comparison group. The other is: in you specifically, is something below the reference range? A profile does measure that, and that is what I use it for, together with the protein supply and the ratio of glutamine to glutamate.

A single abnormal value out of 22 measured, however, is not yet a finding. The way to tell them apart is the second measurement: what is low again on remeasuring was no coincidence.

2. Check the values for which there is good evidence. Ferritin, vitamin D, the omega-3 index, whole-blood magnesium and zinc are studied in ADHD, and for some of them there are intervention studies. Which values come in which order I have put together in ADHD and blood values.

3. Look at protein intake, entirely without a lab. On average it is unremarkable in ADHD, but individual people take in too little, especially in the morning. Anyone who notes it for a few days sees it quickly, and it costs nothing.

4. With persistent gut complaints, have it clarified specifically. If complaints persist, coeliac serology and, with diarrhoea, stool elastase belong in the picture, both through the doctor. In coaching we can, before and after, put into context which investigation is due and what the results mean for your everyday life.

5. Do not supplement without a reason. For tyrosine in ADHD there is no controlled study, for taurine not a single measurement, and for GABA the available data argue against it.

Where to go from here

The question of the building blocks is legitimate. Research’s answer at present is: measurement was done in children, the calculation was mostly wrong, and for adults the decisive study is missing. That is an unsatisfying state, and it is nonetheless usable, because it saves you expensive diagnostics and directs the view to the values where there is actually something to gain.

In my ADHD coaching for adults I look at your values together with your sleep, your diet and your everyday life, and sort out what comes first for you. You get your plan in writing, always within what coaching can do and what belongs in medical hands.

If you would first like to find out whether this fits you, book a free initial consultation. It is non-binding and takes place online.