Does L-Theanine Add Anything to Caffeine? Part Three: The Side-Effect Claim

Abstract

The most attractive claim about the L-theanine–caffeine combination is not about performance at all. It is that theanine removes what people dislike about caffeine — the jitters, the racing heart, the edge of anxiety — while leaving the alertness intact.

This part covers the only trial in the series that measured those side effects: twelve elite Iranian wrestlers, tested under four conditions, with anxiety inventories and symptom reports alongside physical and cognitive tasks.

It reports the most striking numbers in this series and is the least equipped to support them. It also closes the series — including a correction to an earlier version of this part, prompted by a fourth trial we should have covered from the start.

1.The Trial

Rouzbeh Razazan, Mohammad Hemmatinafar and colleagues at Shiraz University published this randomised, double-blind, placebo-controlled crossover trial in the Journal of the International Society of Sports Nutrition in September 2025.

Twelve elite male wrestlers from the Iranian National Wrestling League completed four conditions separated by two-week washouts: placebo, 3 mg/kg caffeine, 3 mg/kg L-theanine, and both together at 3 mg/kg each. Testing began sixty minutes after ingestion.

The battery was broad: a wall-squat hold, a backward overhead medicine ball throw, a vertical jump, handgrip dynamometry, and the Specific Wrestling Fitness Test — three thirty-second bouts of repeated throws with a weighted dummy. A Stroop task was run before and after that test, so cognition was measured both fresh and exhausted. State and trait anxiety were assessed with the State-Trait Anxiety Inventory, and side effects were recorded.

One inclusion detail governs everything that follows. All participants were low habitual caffeine consumers, taking 70 mg a day or less. A caffeine-naive population is where stimulant side effects are most likely to appear, which makes this a sensitive setting for the question — and an unrepresentative one for most people who would buy such a product.

2.The Side-Effect Result

Caffeine alone raised state anxiety significantly above placebo, above theanine alone, and above the combination. Tachycardia — a racing heart — was reported by 92 per cent of participants after caffeine alone.

Under the combination, anxiety symptoms were reported by 8 per cent, which is below the 33 per cent reported on placebo, and tachycardia by 17 per cent.

One performance finding gives that result its interest. Caffeine alone did not beat placebo on the wrestling-specific test, while the combination did. The authors' reading is that at 3 mg/kg in caffeine-naive athletes, the side effects cancelled out the ergogenic benefit — impairing coordination, focus and willingness to push to exhaustion — and that adding theanine let the benefit through.

That is a genuinely interesting hypothesis. It reframes theanine not as a cognitive enhancer but as something that removes an obstacle, which would explain why theanine alone does so little in the meta-analytic literature while the combination keeps outperforming its parts. If it were established it would matter.

3.Why It Is Not Established

The side effects were self-reported. No heart rate was recorded at any point. "Tachycardia" in this paper means a participant said their heart was racing, and the authors name this among their own limitations — noting that they used self-reported perceptual measures rather than psychophysiological tools.

Twelve participants, all male, all caffeine-naive, all from one training camp. The authors are candid that this cannot generalise to women or to habitual caffeine users, which between them account for most of the people the finding would be marketed to.

The abstract and the results section disagree with each other. The abstract states that trait anxiety was lower under the combination than under caffeine, at p = 0.018. The results section states that no significant difference was observed between those two conditions, at p = 0.513. These are the same comparison, and only one of the two can be right.

There is a deeper problem with that outcome. Trait anxiety is defined as a stable disposition to perceive situations as threatening — the thing that distinguishes it from state anxiety is precisely that it does not move with circumstances. A capsule swallowed sixty minutes earlier should not change it. The paper reports it changing in both caffeine conditions, which is a reason to doubt the measurement rather than a discovery about theanine.

Finally, the degrees of freedom. With twelve participants across four conditions, a one-way repeated-measures analysis of variance has 3 and 33 degrees of freedom. The five physical outcomes, all drawn from that one design, report five different pairs — including two whose error degrees of freedom exceed eighty. Twelve participants cannot supply eighty. Whatever these analyses were, they were not the analysis described in the methods.

4.Where the Errors Meet

The sample size calculation deserves its own note, because it links this trial to the previous one.

The authors justify twelve participants by citing a standardised effect size — a Cohen's d, they write — of 0.653, taken from the curling trial covered in Part Two. In that paper, 0.653 is not a Cohen's d. It is a partial eta squared, reported for reaction time on the neutral Stroop condition.

Those are different quantities on different scales. Treating that particular partial eta squared as though it were a d assumes a smaller effect than the source implied, so the error did not by itself shrink the sample — it points the conservative way. The objection is not the direction but the anchor: a power calculation is only as sound as the effect size it borrows, and this one borrows a number that is not what it was taken to be, from a paper whose statistics do not reconcile internally.

One further inconsistency, of a kind that is easy to check and hard to explain. The abstract describes the participants as aged 21.8 years on average. The participants section of the same paper describes them as aged 17, with eligibility open from fifteen. Whether this trial studied adults or adolescents is not a minor bibliographic detail — it bears on consent, on ethics review, and on whether a stimulant dose is being extrapolated across a developmental boundary.

5.Where the Three Recent Trials Leave It

Three randomised trials of the combination, published across three years, all asking some version of whether L-theanine adds anything to caffeine.

The best-measured of them had no caffeine-alone condition and therefore could not answer. The two designed to answer it report statistics that do not reconcile — and the second borrows its power calculation from the first's unreconciled numbers.

So the position is where the meta-analyses left it. The combination remains the most promising claim in the theanine literature, and it is still not settled. The primary trials published since have not advanced it nearly as far as their abstracts suggest.

There is also a fourth trial, older and better reported than any of them, which we should have covered from the start.

An earlier version of this part said that across the three trials the combination was never worse than its components. That was wrong, and the correction matters enough to give it its own section.

6.The Trial That Found the Opposite

Fiona Dodd, David Kennedy, Leigh Riby and Crystal Haskell-Ramsay at Northumbria University published a four-arm crossover trial in Psychopharmacology in 2015. Twelve habitual and twelve non-habitual caffeine consumers each received, on separate visits, 75 mg of caffeine, 50 mg of L-theanine, both together, and placebo — doses chosen to approximate one to two cups of tea. Cerebral blood flow was measured by near-infrared spectroscopy alongside cognition and mood.

Caffeine alone worked. It reduced choice reaction time, improved Stroop performance and raised overall mood ratings.

The combination did not. In the authors' words: there were no positive effects on behaviour of combining L-theanine and caffeine. The improvements seen after caffeine alone, in habitual and non-habitual consumers alike, were lost when theanine was added.

Their conclusion is stated without hedging: the results do not support combining caffeine with L-theanine at the doses tested.

The blood pressure result cuts the same way, against the side-effect claim that opens this part. Here it was the combination — not caffeine alone — that raised systolic pressure relative to theanine and placebo, and diastolic relative to placebo. The authors call this unexpected and note it contradicts the earlier finding that theanine attenuates caffeine's pressor effect.

The dose caveat is real and the authors raise it themselves: 75 mg with 50 mg is low, and the ratio favours caffeine, whereas the trials in Parts One to Three used equal parts or more. A null at tea-strength doses does not refute a benefit at supplement-strength doses.

But it does dispose of the claim that every trial points the same way. This is the best-reported four-arm study in the set — open access, a specialist laboratory, no arithmetic that fails to reconcile — and it found the combination worse than caffeine alone on the outcomes caffeine improved.

7.How This Error Happened

We record it because the mechanism is instructive and because a library that audits other people's methods has no standing to hide its own.

Our topic file listed the Dodd trial from the outset. When we came to write, we recorded all three original sources as closed access without testing that assumption on each one individually. Dodd and colleagues published under a Creative Commons licence and the full text was freely available the whole time.

So the omission was not a judgement about the evidence. It was a source we assumed we could not read and never checked — and it happened to be the one trial pointing the other way. That is exactly the shape of error that produces a literature review agreeing with itself.

What this literature needs is not another positive result. It is a study whose numbers a reader can check — and, on the evidence of this correction, a reviewer who verifies that the inconvenient source is genuinely unobtainable before leaving it out.

Editorial Comment

Boost pairs 250 mg of L-theanine with 100 mg of caffeine, and this series is our audit of the reason for that pairing rather than our case for it. We think the audit is worth publishing in full, including the part where it does not come out cleanly.

The side-effect claim is the one we are most careful about, precisely because it is the one a customer would most like to hear — that this is caffeine without the jitters. We are not making that claim. It rests, so far, on twelve caffeine-naive men reporting their own symptoms with no heart rate recorded and an abstract that disagrees with its own results table.

There is no authorised health claim for either L-theanine or caffeine that we could quote here, so we formulate none. Where an authorised wording does exist for an ingredient we actually use, we quote it exactly and attribute it to that ingredient; there is nothing to quote in this case.

One practical note, offered as a health point rather than a product one. If ordinary amounts of caffeine reliably give you palpitations or a racing heart, that is worth raising with a doctor rather than managing with a second supplement. Thyroid problems, anaemia and cardiac arrhythmias all present that way, they are identifiable, and several of them are treatable. Finding the cause is worth more than smoothing the symptom.