L-theanine against Silexan: on the same anxiety scale, the green tea one did not beat placebo

About a 6 minute read.

Too long, didn't read

• In the 2022 Bayesian network meta-analysis, L-theanine's HAMA result is MD −0.49, credible interval −6.54 to 5.57, crossing zero
• The one that worked on the same table is Silexan (−3.84), lavender essential oil in a capsule you swallow
• The database leaves laevo-theanine's odor type, odor, strength and substantivity all four empty, and files it under special dietary and nutritional additives
• A second record, dextro,laevo-theanine, the catalogue marks information only, not used for fragrances or flavors
• Meanwhile the notes field of lavender oil france says silexan used as an anti-anxiety agent
• Smelling it does have evidence: the 2019 pooled effect for inhalation is Hedges' g = −0.73, though the authors themselves will only call it an indication of an effect of reasonable size

Start with a question the database can answer

"Green tea is relaxing" is everywhere. The step before that is what I wanted: what does L-theanine look like inside a fragrance catalogue?

There are two records.

Record	CAS	Suppliers	Catalogue category	Odor fields
laevo-theanine	3081-61-6	12	Special dietary and nutritional additives	(empty)
dextro,laevo-theanine	34271-54-0	3	Information only, not used for fragrances or flavors	(empty)

The laevo record is real enough. Twelve suppliers sell it. Its natural occurrence field is detailed: tea anther, tea cotyledon, tea leaf, tea seed, tea shoot, tea stem. The notes field reads "precursor of ethylamine; found in green tea".

But odor type, odor, odor strength and substantivity carry not one word between them.

This is not a gap in the data. The same catalogue gives lavender oil eight odor words, a medium strength rating and 12 hours of substantivity. It is capable of filling those fields. This material simply has nothing to put in them.

The second record is blunter. For the racemic form the catalogue writes the category as information only, not used for fragrances or flavors, and the occurrence field reads not found in nature. The catalogue is saying it is not a fragrance material.

What are those 12 suppliers selling?

Raw material for supplements. That category holds 187 records site-wide, 157 of them with suppliers, and the neighbours look like this: vegetable protein hydrolyzates at 51 suppliers, xylitol at 37, fumaric acid at 33, laevo-arginine at 29, laevo-glutamine at 27.

None of that block is for making perfume.

For context, 15,670 records in the database have suppliers and an empty odor field, so an empty odor field is not remarkable on its own. What is remarkable is the category sitting next to it, which places this one firmly on the food and dietary side.

What about green tea flavor? Twenty-one suppliers, category flavored products of all types, and a notes field about water-soluble, oil-soluble, emulsion, liquid and powder forms. Its odor field is empty too. Green tea enters this catalogue as a flavour, not as a material with an odour record.

Two fragrance plants, scored on one table

The 2022 paper in Pharmacol Res (PMID 35378276) did exactly what I needed: it put 12 medicinal herbs into one Bayesian network and compared endpoint scores on the Hamilton Anxiety Scale (HAMA, the standard clinician questionnaire for anxiety severity). It covered 29 randomised controlled trials published between 1987 and 2021.

Four rows from that same table:

Herb	HAMA mean difference	95% credible interval
Withania somnifera (ashwagandha)	−4.90	−9.70 to −0.17
Ginkgo biloba	−4.63	−9.01 to −0.23
Silexan (lavender oil preparation)	−3.84	−6.31 to −1.34
Kava	−2.46	−4.47 to −0.32
L-theanine	−0.49	−6.54 to 5.57

Look at the interval on the last row. From −6.54 to 5.57, crossing zero. This body of data cannot rule out that it performs the same as placebo.

The L-theanine interval is also more than twice as wide as Silexan's, 12.11 against 4.97. A wide interval means the data are few or disagree with each other, not that the effect is small. Those are two different statements: L-theanine has not been shown to be useless, these trials just cannot support a conclusion.

The 2026 paper in Mol Psychiatry (PMID 42410082) redid the exercise with 31 RCTs and 1,168 participants and reads the same way: anxiety effects were inconsistent and non-significant, with one exception in a study of psychotic anxiety, SMD 0.54 at 400 mg/day for eight weeks.

The one that won is swallowed

The name Silexan already appears in the database. The notes field of lavender oil france (#21974, 55 suppliers) reads in full:

silexan used as an anti-anxiety agent. tsca definition 2008: extractives and their physically modified derivatives. lavandula officinalis, labiatae.

The catalogue recorded it.

Silexan is lavender essential oil in an oral capsule, standard dose 80 mg/day. The 2019 meta-analysis in Phytomedicine (PMID 31655395) gives it a HAMA mean difference of −2.90 (95% CI −4.86 to −0.95, p = 0.004, 1,173 participants), on condition of at least six weeks of continuous use.

So the one entry on that table connected to a fragrance plant that did win takes the digestive route.

Worth pausing here. Lavender oil is a real fragrance material in this database: lavender oil (#21975) has 99 suppliers, odor type floral, eight odor words (herbal, floral, spicy, aromatic, camphoreous, balsamic, aldehydic, woody), medium strength, 12 hours of substantivity, FEMA 2622, FEMA GRAS. One plant oil can serve as a perfumery material and as a medicine in a capsule, and the evidence for those two uses is kept separate.

What about smelling it

This is the section I expected to come up empty and it did not.

The 2019 meta-analysis pulled in 65 RCTs (7,993 participants) for qualitative synthesis and 37 (3,964 participants) into the quantitative pooling. For the inhalation route:

• Anxiety on any validated scale: Hedges' g = −0.73 (95% CI −1.00 to −0.46, p < 0.00001, 1,682 participants)
• State anxiety (STAI-State): mean difference −5.99 (95% CI −9.39 to −2.59, p < 0.001, 901 participants)
• Trait anxiety (STAI-Trait): mean difference −8.14 (95% CI −14.44 to −1.84, p < 0.05, 196 participants)
• Systolic blood pressure: no significant effect

The first three are significant. Here is what the authors wrote in their own discussion, which is the point of this article:

Overall, oral administration of lavender essential oil proves to be effective in the treatment of anxiety, whereas for inhalation there is only an indication of an effect of reasonable size, due to the heterogeneity of available studies.

Earlier in the same section: the majority of the included randomised controlled trials carried a high overall risk of bias.

Inhalation is not a null result. It has a significant pooled effect size, and the authors, having looked at how those studies were built, demoted it to an indication. Significant and trustworthy are separate things, and the line between them is drawn by study quality rather than by a p-value.

That "no significant effect on systolic blood pressure" is worth reading too. The subjective scales moved and the physiological marker did not. This pairing turns up often in scent research, and it is at least a reminder that participants knew they were smelling lavender.

Why the articles you find do not read like this

I ran one traditional-Chinese SERP for "L-theanine anxiety". The first page held supplement retailers' blogs, a pet health site, Wikipedia and Zhihu. The content converged on the same few points: alpha waves rise, GABA rises, absorbed in 40 minutes, 100 to 400 mg.

Each of those claims has a source behind it and I found no obvious fabrication. The problem is that not one of them mentions the 2022 table.

That is not hard to understand. Alpha waves and GABA are mechanism findings, and a mechanism finding says what a substance does inside a brain. A HAMA meta-analysis says whether people who took it scored lower on an anxiety questionnaire than people who took placebo. The first is easier to tell and easier to sell.

On the perfumery side the same gap shows up in a different shape: oral trial results propped up behind an olfactory claim about a material.

What this means for a formula

Three usable things.

One, L-theanine will not go into your formula, because it has no odour to work with. The four empty fields are the answer. Building a green tea accord runs through green tea flavor, or through stacking single materials that do carry odour records, and that is a different article.

Two, before citing "lavender relieves anxiety", ask which route. The oral Silexan evidence is at a clearly higher grade than the inhalation evidence. Neither can borrow from the other, even for the same oil.

Three, once a credible interval crosses zero, stop reasoning forward from it. The point estimate of −0.49 looks like a small effect. The right-hand end of its interval is +5.57. Point estimates invite you to keep going, intervals do not.

I got one thing wrong while writing this, so it goes in as well. On my first pass through the literature I saw only the half-sentence "for inhalation there is only..." in the discussion and recorded it as "inhalation does not work". The results section gives g = −0.73, which is significant. The discussion section and the results section of one abstract can point different ways, and reading one of them gets you a wrong impression.

→ Laevo-theanine in the database | lavender oil | lavender oil france: fields, suppliers and categories are all there to check.

What this article does not establish

• I am not ruling on whether L-theanine works. This article says the credible interval from that 2022 set of trials crosses zero, and that the 2026 review calls the anxiety effects inconsistent. That describes evidence strength, not efficacy.
• One corresponding author of the 2026 Mol Psychiatry paper founded a company selling dietary supplements, disclosed in the paper's conflict of interest statement. Its conclusions run conservative on L-theanine, so the conflict does not flatter the result, but it belongs alongside the citation.
• Three authors of the 2019 Nutrients RCT (PMID 31623400) are from Taiyo Kagaku, whose nutrition division sells L-theanine. That trial ran 30 participants at 200 mg/day for four weeks, with self-rated depression, trait anxiety and sleep quality all falling.
• The 2007 Nagoya University study (PMID 16930802) had 12 participants. It gets cited as the origin of "L-theanine reduces stress", and that sample cannot carry a general conclusion.
• The authors of the 2022 paper call their own conclusions preliminary, citing unconvincing sample sizes and the potential effectiveness of placebos. I cite their ranking, not a verdict.
• The database's category field reflects this catalogue's editorial judgement, not regulatory status. "Information only, not used for fragrances or flavors" is the catalogue talking, not an agency.
• I did not check L-theanine's standing in food flavouring regulation, or whether it holds a FEMA number. That field is empty in both records, and an empty field is not the same as an absence.