Reading the L-Theanine Evidence: Part Two: Sleep, and Why Two Reviews Disagree

Abstract

In February 2025 a team of Australian academics published a meta-analysis of L-theanine and sleep. In November 2025 a second team published a systematic review of the same question, drawing on a heavily overlapping set of trials.

The first concluded that the effectiveness of pure L-theanine for managing sleep disturbance is not well known. The second concluded that 200 to 450 mg a day appears to be a safe and effective way to support healthy sleep in adults.

Part One of this series covered cognition. This part covers sleep, and it is mostly an exercise in working out how two competent reviews of one small literature arrive at opposite headlines. The answer turns out to be methodological, and it is visible in both papers if you read past the abstract.

1.The Meta-Analysis

Amanda Bulman, Nathan D'Cunha and Nenad Naumovski at the University of Canberra, with colleagues at Deakin and elsewhere, searched five databases and one register, registered their protocol in PROSPERO before starting, and identified 19 randomised trials covering 897 participants. Eighteen went into the meta-analysis.

Their inclusion criteria were deliberately wide: L-theanine delivered alone or in combination with other compounds, in humans of any age and health status, provided the dose was stated and sleep was measured with a validated instrument. Doses ranged from 50 to 1,000 mg a day, durations from a single night to eight weeks.

Three outcomes reached significance. Subjective sleep onset latency improved with a standardised mean difference of 0.15 (CI 0.01 to 0.29). Daytime dysfunction improved by 0.33 (CI 0.16 to 0.49). Overall subjective sleep quality improved by 0.43 (CI 0.04 to 0.83).

Everything else did not. No significant difference for objective sleep onset latency, subjective or objective sleep efficiency, subjective or objective sleep duration, or subjective or objective sleep disturbance.

The pattern is worth stating plainly: the outcomes that improved were the ones participants reported. The outcomes measured by a device did not move.

2.How Fragile Were the Positive Results

Bulman's team ran sensitivity analyses, which is the practice of removing one study at a time to see whether the pooled result survives. Their findings here deserve more attention than the abstract gives them.

For subjective sleep onset latency, removal of five of the ten studies eliminated significance. For overall sleep quality, heterogeneity was very high at 82.9 per cent, and one study was the source, with an effect size of 2.58 against a pooled 0.43; removing it dropped the pooled effect to 0.21. Removing a different study eliminated significance altogether.

One result runs in the other direction and is easy to miss. For objective sleep onset latency, removing the study responsible for the heterogeneity left a pooled result that was significant in favour of the control condition.

There is also the composition problem the authors flag first among their limitations: in 11 of the 19 included studies, L-theanine was one ingredient in a product containing others — GABA, valerian, lemon balm, saffron, magnesium, lactic acid bacteria, tart cherry, lavender. Their own sentence is unambiguous: the observed outcomes cannot be solely attributed to L-theanine alone.

Their conclusion follows from all of this rather than from the three significant results, and it is appropriately hedged: the findings show potential, and a well-designed dose–response trial in pure L-theanine is what the field needs.

3.The Systematic Review

Jack Cotter, Charlotte Caddick, Jessica Harper and James Ebajemito took the opposite methodological decision on the same question: exclude every trial in which L-theanine was combined with another potentially active ingredient, and look only at the standalone compound.

That is a defensible choice — arguably a cleaner one — and it left 13 trials in 550 participants, of which 11 were randomised controlled trials and two were single-arm open-label studies with no control group at all.

They did not pool the results. Their stated reason is sound: heterogeneity in design, outcomes and reporting made meta-analysis inappropriate. What they did instead was count studies. Beneficial effects were reported in 9 of the 13 trials, including 9 of the 11 that used 200 mg a day or more.

The counting rule matters. The phrase used throughout is statistically significant and/or trend-level beneficial effects, and trend-level results are included in the tally — p-values of 0.070 and 0.073 appear in the evidence table as supporting findings. The authors justify this on the grounds that the trials were underpowered and the direction was consistent.

That reasoning is not unreasonable. It is also the reasoning that turns a set of near-misses into a positive review, and a reader should know it is operating.

4.Who Wrote Which Review

Both papers disclose their interests. The disclosures are not equivalent and both are printed in full at the foot of the respective articles.

Bulman and two co-authors declare funding from a University of Canberra Discovery Translation Grant, registered with the university's research office. The remaining authors declare no conflict.

Cotter and colleagues declare that the authors are employees of The Water Street Collective or British American Tobacco; that The Water Street Collective sells products that contain L-theanine; and that The Water Street Collective is a wholly owned subsidiary of British American Tobacco. The corresponding address is a bat.com email.

We are not suggesting the review is dishonest. It is transparent about its methods, it names the four trials that failed to show benefit, its quality assessment rates most included studies as merely fair, and it acknowledges that it was not prospectively registered. Those are not the moves of a paper trying to hide something.

But the judgement calls inside it — include open-label trials, count trend-level results, decline to pool — all point the same way, and they were made by people who sell the compound. The independent team, making the opposite calls on an overlapping evidence base, reached a hedged conclusion. That is information about the evidence, not merely about the authors.

A third reading exists. Payne and colleagues at the Rowett Institute, whose meta-analysis we discussed in Part One, examined sleep as one of their three outcome domains and found the evidence for tea, theanine plus caffeine, and theanine alone to be inconclusive. Their funding comes from a tea company, which makes the direction of that particular conclusion worth noting too.

5.What Both Reviews Actually Agree On

Strip out the framing and the two papers describe the same evidence base, with the same shape.

Neither found convincing evidence that L-theanine increases total sleep time. Cotter states it directly: there was little evidence that L-theanine increases total sleep time. Bulman's pooled analysis of objective sleep duration returned a mean difference of 5.91 minutes with an interval spanning −8.08 to 19.91.

Both find the effects, where present, concentrated in perceived quality rather than measured architecture — how long it felt like falling asleep took, how refreshed the morning was, how the day went afterwards.

Both report good tolerability. Across the pooled trials the commonest complaints were headache and gastrointestinal symptoms; Cotter's set included doses up to 900 mg a day for eight weeks with few adverse events. This is a compound with an unusually mild safety profile, and that is a genuine finding in its own right.

And both leave the dose unsettled in a way that is almost comic when the two are read together. Bulman's subgroup analysis pointed to under 200 mg as the more effective range. Cotter's conclusion names 200 to 450 mg. The same literature, opposite recommendations, neither of them established by a dose–response trial because no such trial exists.

6.The Subjective–Objective Gap

The most interesting question in this literature is not whether L-theanine works. It is why the questionnaires move when the actigraphs do not.

Three explanations are available and the trials cannot currently distinguish between them.

It could be placebo response, which is powerful in sleep research and shows up preferentially in self-report. It could be that the instruments are mismatched — actigraphy and even polysomnography measure architecture, not the experience of resting, and Bulman's team note that sleep-stage data were too sparse to analyse at all. Or it could be a real effect on how sleep is experienced that current objective measures are not designed to detect.

The third possibility is the one both reviews lean towards, via the stress route: L-theanine has better-established effects on stress and anxiety at 200 to 400 mg, and a calmer person may both fall asleep more easily and remember the night more kindly. Part Three examines that stress literature directly, because it is the strongest evidence this compound has and it is not about sleep at all.

One concrete finding points at a narrower and more testable claim. In a nine-person crossover study using EEG, caffeine taken at ten to eleven at night increased time awake after sleep onset by more than an hour. Adding 50 mg of L-theanine to the same caffeine dose brought that figure back to placebo levels. The sample is tiny. But the hypothesis it suggests — that theanine's main sleep-relevant action is offsetting caffeine rather than inducing sleep — fits the rest of the data better than the sleep-aid framing does.

Editorial Comment

MindHeaven® does not make a sleep product, does not claim that L-theanine improves sleep, and has no commercial position in this question. L-theanine has no authorised health claim in the European Union for sleep or for anything else.

That disinterest is why we chose to write this part carefully rather than briefly. The sleep claim is the most heavily marketed application of this compound and the weakest-supported one in its literature — subjective measures moving while objective measures do not, the strongest positive review written by people who sell it, and no dose–response trial anywhere in the field.

Our own use of L-theanine is in a daytime formulation, paired with caffeine, on the attention evidence described in Part One. Nothing in the sleep literature bears on that, and we would not cite it as though it did.

If you are not sleeping and it has become a pattern rather than a bad week, that is a conversation for a doctor. Persistent insomnia has causes worth identifying, the international guidelines cited by Bulman's team do not recommend supplements of this kind for treating it, and the honest reading of nineteen trials is that this one changes how the night is remembered more reliably than it changes the night.

Part Three turns to stress, where the evidence is stronger, the trials are better, and a different problem appears: almost nobody tests the compound on its own.