L-Theanine and Paraxanthine in ADHD and Autism: A Well-Aimed Pilot on a Strange Finding

Abstract

A pilot study at Texas Tech is giving adults with ADHD or autism spectrum disorder a capsule containing 200 mg of L-theanine and 200 mg of paraxanthine, then measuring inhibitory control while they lie in an MRI scanner.

It is small — 24 participants — and single-masked. It is also the most conceptually precise trial in our registry file, because its outcome measures are not generic cognitive tests. They are the specific measures on which an earlier study found something strange.

This note explains what that earlier finding was, why swapping caffeine for its main metabolite is a reasonable next move, and why the evidence behind paraxanthine is thinner than its marketing suggests.

1.The Trial

Registry entry NCT07189442 carries a long and unusually informative title: combining L-theanine and paraxanthine for transient improvement of cognitive deficits among patients with attention deficit hyperactivity disorder and autism spectrum disorder — a series of translational pilot neuroimaging studies.

It is sponsored by Texas Tech University Health Sciences Center with Texas Tech University, and the listed contacts are Chathurika Dhanasekara and Chanaka Kahathuduwa. Recruitment opened on 1 October 2025.

Twenty-four adults with a physician diagnosis of ADHD or ASD, by self-report, take either a capsule of 200 mg L-theanine with 200 mg paraxanthine, or 400 mg of corn starch, in a randomised single-masked crossover.

The three primary outcomes are all measured 30 minutes after intake, and two of them are performed inside the scanner: task-related fMRI activity, stop-signal reaction time, and d-prime — sensitivity to the Go signal in a Go/NoGo task.

One administrative note. The registry gives a primary completion date of 1 June 2026 and was last updated on 5 February 2026. That date has passed without an update. Registry dates drift routinely and this means nothing on its own, but it is the kind of detail worth recording when writing about a trial before it reports.

2.Why Those Measures

Stop-signal reaction time and d-prime in a Go/NoGo task are not general cognition tests. They measure inhibitory control — the capacity to stop a prepared response — which is the deficit most central to ADHD.

They are also the exact measures on which Kahathuduwa and colleagues reported a result in 2020 that we covered in Part One of our L-theanine series, and which has stayed with us since.

In that study, five boys with ADHD received L-theanine, caffeine, the combination, or placebo. Caffeine alone worsened inhibitory control, lengthening stop-signal reaction time with a p-value of 0.031. L-theanine alone also worsened it, at 0.053. Only the combination improved things: d-prime in the Go/NoGo task at 0.033, total cognition composite at 0.041.

Two compounds that each make a measure worse, and together make it better. That is either a genuine interaction worth pursuing, or noise in five children. The present trial is the same investigator returning to the question with more participants, brain imaging, and a different second ingredient.

That is how a pilot finding is supposed to be followed up, and it is worth saying plainly given how rarely we get to say it.

3.Why Paraxanthine Instead of Caffeine

Paraxanthine is the principal metabolite of caffeine in humans — most ingested caffeine is converted to it by the liver enzyme CYP1A2.

The rationale for using it directly is genetic. Individuals differ substantially in CYP1A2 activity, so the same caffeine dose produces different amounts of the active metabolite in different people. Administering paraxanthine bypasses that variation.

This is a coherent pharmacological argument, and it happens to matter for the earlier finding. If theanine and caffeine interact through something paraxanthine does, then a trial giving caffeine is giving a variable dose of the actual agent, and the interaction would be harder to detect than it should be.

4.What Paraxanthine's Own Evidence Looks Like

There is a human literature on paraxanthine and cognition. It is smaller and more closely held than its presence in supplement marketing implies.

A dose-response trial published in Nutrients in 2021 gave 12 healthy volunteers placebo, 50 mg, 100 mg or 200 mg of a branded paraxanthine, acutely and then daily for seven days, in a double-blind counterbalanced crossover. Its headline measure — Sternberg Task reaction time — tended to differ among groups at a p-value of 0.06. Several differences reached significance on Berg-Wisconsin Card Sorting measures. The conclusion states that paraxanthine may serve as an effective nootropic agent at doses as low as 50 mg.

A 2024 trial in the Journal of the International Society of Sports Nutrition compared placebo, caffeine, paraxanthine, and both together in 12 trained runners before and after a 10-km treadmill run. Card-sorting correct responses rose with paraxanthine from pre- to post-exercise at 0.012. Error rates rose with caffeine at 0.02 and tended to be lower with paraxanthine than caffeine at 0.059. Perseverative errors after exercise were lower with paraxanthine than caffeine at 0.026.

Both trials enrolled 12 people. Both used the same branded ingredient. And both carry the same disclosure: three authors are principals of the study's sponsor and are inventors on patent applications covering paraxanthine, though they state they were not involved in data collection or analysis.

That is not a reason to dismiss the results. It is a reason to notice that the entire human case for this compound currently rests on small crossovers conducted around a commercial ingredient by people who hold the patents — and that the Texas Tech trial is, as far as we can see, the first test of it by an unconnected academic group in a clinical population.

5.What Would Make the Result Worth Reading

Does the imaging agree with the behaviour? The 2020 predecessor reported reduced task-related activity in default mode network regions associated with mind-wandering. If stop-signal reaction time improves and the scanner shows nothing, in 24 single-masked participants, that is a thin result.

Are ADHD and ASD reported separately? The registry describes two parallel crossover trials and assigns dual-diagnosis participants to the ASD group. Twenty-four participants split across two conditions is twelve per condition, and pooling them afterwards would change what the numbers mean.

And is the combination compared against its components? The interesting claim from 2020 is an interaction — each compound alone being unhelpful. A trial comparing the combination only against placebo cannot test that, and would leave the most interesting hypothesis in this file untouched.

Editorial Comment

We have a direct interest here and it should be stated before anything else. L-theanine is in Boost at 250 mg alongside 100 mg of caffeine, and in Clarity. Paraxanthine is what most of that caffeine becomes. This trial is testing a combination adjacent to our own formulation, at a theanine dose close to ours, and we would rather say so than let a reader discover it.

What we do not do is claim anything from it. ADHD and autism spectrum disorder are medical diagnoses. Neither L-theanine nor caffeine nor paraxanthine holds any authorised health claim in the European Union relating to them, or to attention, and our products are food supplements sold to healthy adults. Nothing in this note is a suggestion that anyone with either diagnosis should take a supplement.

If attention or executive difficulties are affecting daily life, the useful step is assessment rather than self-medication. ADHD in adults is treatable, the treatments that work have effect sizes no supplement in this literature approaches, and the diagnostic process also identifies the conditions that imitate it — sleep disorders, thyroid dysfunction, anxiety, iron deficiency. That is a conversation for a doctor.

We are watching this trial because it is a well-aimed follow-up to a genuinely odd finding, run by the group that produced it, with imaging attached. Twenty-four participants will not settle anything. It may establish whether there is something here worth a trial that could.