Lemon Balm

Lemon Balm

Lemon balm is a Lamiaceae leaf sold as calm, sleep, and a milder mind. The chemistry is a phenolic-acid marker plus a thin, unstable oil. The clinical file is small, often subjective, and frequently a combination product. This article keeps Melissa officinalis alone apart from valerian/lemon-balm and other multi-herb mixtures, and it does not let a GABA-T assay stand in for an anxiolytic trial.

Compiled by South Beach Longevity · 20 August 2026 Copyright 2026 Series SBL-41 / SP-LEMON-BALM · Register A scientific article Sources peer-reviewed human trials, meta-analyses, labelled animal and in-vitro work, and regulator opinions · verified NCBI records Constraint This document describes published research. It is not medical advice. No human use, dose, route or schedule is recommended anywhere in this document.

How to read this document Every finding is labelled, in the sentence that reports it, by the kind of study that produced it. A GABA-transaminase IC50 is not a treatment for anxiety. A valerian/lemon-balm restlessness study is not a lemon-balm trial. An acute laboratory attention factor is not everyday cognition and is not dementia. Amounts and durations appear only as reported experimental parameters, always with the population attached. Nothing here is a recommendation.

Findings are graded in place as established, strongly supported, supported, emerging, plausible, or speculative, and where a claim fails on its own evidence, as refuted. Two further labels mark careful absences rather than verdicts: not established, where the evidence is too thin to place a claim on the ladder at all — untested or insufficient, an absence of proof rather than disproof; and not supported, where the weight of evidence leans against a claim but stops short of a formal refutation. They grade the reviewed record in hand, not the plant’s commercial future. Lemon-balm-alone rows and combination-product rows are never collapsed.

01 What lemon balm is — and what it is not

Lemon balm is Melissa officinalis L., a perennial mint whose crushed leaf smells of lemon because a small essential-oil fraction carries citral and related monoterpenes. The part used in European herbal articles is the leaf — Melissae folium — not the oil, not isolated rosmarinic acid, and not a multi-herb night tablet (European Medicines Agency HMPC, 2013; Shakeri, Sahebkar, and Javadi, 2016).

Four things follow immediately and will be enforced for the rest of this document.

First, lemon balm is not one exposure. Dried leaf, infusion, hydroalcoholic extract, a Cyracos-class extract standardised to hydroxycinnamic acids, a phospholipid “phytosome,” inhaled or topically applied essential oil, and isolated rosmarinic acid are six different objects. Kennedy’s 2002 Pharmaton extract and his 2003 cholinergically screened dried leaf already produced different cognitive profiles in the same laboratory (Kennedy et al., 2002; Kennedy et al., 2003). That is a product-chemistry fact, not a footnote.

Second, lemon balm is not valerian. The two plants are sold together more often than they are tested apart. Müller and Klement’s 918-child open study, Gromball’s school-age combination study, Taavoni’s menopausal sleep study, and Kennedy’s 2006 laboratory-stress combination study are combination files (Müller and Klement, 2006; Gromball et al., 2014; Taavoni, Nazem Ekbatani, and Haghani, 2013; Kennedy et al., 2006). They do not grade Melissa alone.

Third, lemon balm is not a proven GABA drug. Rosmarinic acid and related triterpenes inhibit GABA transaminase in rat-brain homogenates (Awad et al., 2007; Awad et al., 2009). The essential oil displaces a channel-site ligand at GABA-A and also inhibits GABA-evoked currents in cortical neurons (Abuhamdah et al., 2008). Those are laboratory hypotheses. They are not a demonstrated human mechanism.

Fourth, this document is not medical advice and not a sleep protocol. Reported milligrams appear only as study parameters attached to named populations. They are not instructions for any reader.

02 Botanical identity and pharmacognosy

Melissa officinalis is a member of the Lamiaceae. The English name records the lemon scent; the Latin name records the older association with honey-bees. The leaf is ovate, crenate, and glandular. European traditional-use products are defined on the leaf, comminuted or powdered, or as ethanol–water extracts and tinctures whose solvent range the HMPC set at 45–53% V/V ethanol (European Medicines Agency HMPC, 2013). That is a pharmacognostic boundary. It is not a clinical licence.

Shakeri, Sahebkar, and Javadi’s 2016 ethnopharmacology review remains the reviewed record’s best single map of traditional uses, phytochemistry, and pharmacology (Shakeri, Sahebkar, and Javadi, 2016). Petrisor and colleagues later restated composition and derived products (Petrisor et al., 2022). Neither review converts traditional use into a modern RCT. The HMPC assessment is explicit: the European article rests on traditional use for mild mental stress and as a sleep aid, and for mild gastrointestinal complaints including bloating and flatulence — not on well-established clinical efficacy (European Medicines Agency HMPC, 2013).

Subspecies and chemotypes exist. Heitz, Carnat and colleagues isolated luteolin 3′-glucuronide as the major flavonoid of M. officinalis subsp. officinalis (Heitz et al., 2000). Oil yield and citral fraction move with cultivar, drying, and storage. A tea made from a year-old bag is not the GC-MS chromatogram of a fresh-leaf oil.

03 Phenolic chemistry and rosmarinic acid

Rosmarinic acid is an ester of caffeic acid and 3,4-dihydroxyphenyllactic acid. Petersen and Simmonds mapped its occurrence in Lamiaceae and Boraginaceae, its eight-enzyme biosynthesis from phenylalanine and tyrosine, and its constitutive defence role in the plant (Petersen and Simmonds, 2003). PubChem records the molecule as C18H16O8, molecular weight 360.3, CID 5281792, InChIKey DOUMFZQKYFQNTF-WUTVXBCWSA-N. Those identifiers are established. They describe a marker, not a medicine.

In lemon-balm leaf, rosmarinic acid is the dominant hydroxycinnamic-acid marker. Awad’s bioassay-guided fractionation of a methanol extract identified rosmarinic acid as the major GABA-T-active principle, representing about 1.5% of dry leaf mass in that accession, with ursolic and oleanolic acids as additional actives and a possible synergistic contribution (Awad et al., 2009). Lamaison’s older Lamiaceae survey already treated rosmarinic acid as a shared antioxidant marker across the family (Lamaison, Petitjean-Freytet, and Carnat, 1991). Shared markers do not make Melissa, Rosmarinus, and Salvia interchangeable clinical objects.

Cyracos-class extracts are defined by the manufacturer as a 30% ethanolic leaf extract standardised to more than 7% rosmarinic acid and more than 15% hydroxycinnamic-acid derivatives (Ibarra et al., 2010; Cases et al., 2011). That is a brand specification. It is not a pharmacopoeial identity for every capsule on a shelf. Noguchi-Shinohara’s later Japanese programme used a Melissa extract defined by a 500 mg rosmarinic-acid daily parameter in pharmacokinetic and cognition protocols (Noguchi-Shinohara et al., 2015; Noguchi-Shinohara et al., 2020; Noguchi-Shinohara et al., 2023). That is a third object again: a high-RA extract, not tea, not oil.

Human plasma kinetics of rosmarinic acid after such an extract are measurable. In eleven healthy adults, serum total rosmarinic acid peaked at one hour in the fasted state after an extract containing 500 mg rosmarinic acid, with a reported Cmax of 162.20 nmol/L and an AUC of 832.13 nmol·h/L; food increased AUC (Noguchi-Shinohara et al., 2015). Nanomolar circulating RA is supported. Brain occupancy at a GABA-T or amyloid site is not.

04 Volatile constituents

The essential oil of Melissa is typically a small fraction of dry leaf mass — often well below 0.5% — and is chemically unstable. The character compounds are geranial and neral (together, citral), citronellal, and lesser monoterpenes and sesquiterpenes. PubChem records geranial as (2E)-3,7-dimethylocta-2,6-dienal (CID 638011, C10H16O, 152.23), neral as the 2Z isomer (CID 643779), and citronellal as 3,7-dimethyloct-6-enal (CID 7794, C10H18O, 154.25). Those structures are established.

What is not established is that a cup of tea, or a hydroalcoholic capsule standardised to rosmarinic acid, delivers a citral dose comparable to a distilled oil. The oil is the object in Ballard’s and Burns’s aromatherapy trials and in Abuhamdah’s channel work (Ballard et al., 2002; Burns et al., 2011; Abuhamdah et al., 2008). The leaf extract is the object in Kennedy, Cases, Alijaniha, Haybar, and the Noguchi-Shinohara programme. Crossing those files is a category error.

Stojanović and colleagues showed that the oil and citronellal can modulate biological systems in experimental models (Stojanović et al., 2022). Experimental modulation is not a human sleep trial. Abdellatif’s later composition paper is a chemistry note (Abdellatif et al., 2021). Composition notes do not grade agitation.

05 Extracts, teas, oils, and commercial forms

The HMPC lists comminuted leaf as tea, powdered leaf in solid oral forms, liquid extract DER 1:1 in 45–53% V/V ethanol, tincture 1:5 in the same solvent range, and dried extracts corresponding to those preparations (European Medicines Agency HMPC, 2013). Traditional-use amounts in that assessment are historical European pharmacy parameters. They are reproduced here only as regulator language, not as instructions.

Commercial forms in the trial file include: a Pharmaton standardised extract (Kennedy et al., 2002); encapsulated dried leaf selected for cholinergic binding (Kennedy et al., 2003); Cyracos (Ibarra et al., 2010; Cases et al., 2011); lyophilised aqueous leaf extract (Alijaniha et al., 2015); hydroalcoholic extracts in Iranian clinic trials (Haybar et al., 2018; Safari et al., 2023; Araj-Khodaei et al., 2020); a phospholipid carrier / “phytosome” (Di Pierro et al., 2024); foods and beverages carrying a standardised extract (Scholey et al., 2014); and topical dried-extract creams for herpes labialis (Wölbling and Leonhardt, 1994; Koytchev, Alken, and Dundarov, 1999). The last pair is a dermatological object. It does not grade anxiety.

Product quality is a live threat. Oil content collapses on drying. RA content moves with accession and extraction. Combination tablets hide a second plant. A label that says “lemon balm” does not specify which row of Figure 1 was filled.

06 Human exposure is a product question

There is no single human pharmacokinetic file for “lemon balm.” What exists is a rosmarinic-acid plasma curve after a high-RA extract (Noguchi-Shinohara et al., 2015), in-vitro cholinergic binding of selected extracts (Wake et al., 2000; Kennedy et al., 2002; Kennedy et al., 2003), and the tacit assumption that an oral leaf product reaches a CNS target. Di Pierro and colleagues stated the bioavailability problem directly: hydroxycinnamic-acid pharmacology has not always been reproducible in humans, “perhaps due to the poor bioavailability of these compounds,” which is the rationale they offered for a phytosome (Di Pierro et al., 2024). That sentence is a formulation hypothesis. It is not proof that a phytosome repairs the clinical file.

Scholey’s food-matrix studies showed that a lemon-balm extract can be administered in a drink or yoghurt and that rosmarinic acid is absorbed from a water-based drink in a pilot (Scholey et al., 2014). Absorption of a marker is supported. Delivery of a GABA-T-inhibitory concentration to human cortex is not.

07 GABA-related hypotheses

The GABA story has three layers. They are not one finding.

Layer 1 — enzyme assays. Awad, Levac and colleagues screened aqueous and ethanolic extracts of anxiolytic botanicals against rat-brain GAD and GABA-T. Aqueous Melissa was among the extracts that inhibited GABA-T at concentrations that the authors treated as potentially relevant; most other botanicals did not (Awad et al., 2007). Bioassay-guided fractionation then assigned the activity chiefly to rosmarinic acid, with ursolic and oleanolic acids as additional principles (Awad et al., 2009). This is strongly supported as an in-vitro enzymology result. It is not a demonstration that oral lemon balm raises human synaptic GABA.

Layer 2 — animal behaviour. Ibarra administered Cyracos orally to mice for 15 days. Anxiety-like reactivity on the elevated plus-maze fell dose-dependently; the open-field task did not move; exploratory and circadian activity were not altered at the anxiolytic-like doses (Ibarra et al., 2010). Ghazizadeh’s later hydro-alcoholic extract work in rodents is the same class: animal anxiety/depression models, not patients (Ghazizadeh et al., 2020). Animal plus-maze data are emerging as a mechanistic hint. They are not a human grade.

Layer 3 — the essential oil at GABA-A. Abuhamdah’s dual binding-and-electrophysiology study is the paper most often flattened into “lemon balm is GABAergic.” The oil inhibited [35S]TBPS binding to the rat forebrain GABA-A channel (apparent IC50 0.040 mg/mL) and had no effect on NMDA, AMPA, or nicotinic receptors. In cultured cortical neurons it reversibly inhibited GABA-induced currents (0.01–1 mg/mL) (Abuhamdah et al., 2008). Inhibition of GABA-evoked current is not the same sentence as benzodiazepine-like potentiation. The oil is also not the leaf extract. Savage’s later systematic review of GABA-modulating phytomedicines correctly keeps this file in the preclinical column (Savage et al., 2018).

The red-team challenge is therefore already live: mechanistic GABA assumptions do not explain the human outcomes, and they do not license a sedative claim. The HMPC itself treated Awad and Abuhamdah as non-clinical pharmacology supporting a traditional-use narrative, not as proof of well-established efficacy (European Medicines Agency HMPC, 2013).

08 Cholinergic hypotheses

The cholinergic story is older than the GABA-T story and was built for a different claim: failing memory.

Wake and colleagues screened ethanolic extracts of European plants reputed to restore cognition. One Melissa accession was among the most potent displacers of [3H]nicotine and [3H]scopolamine from human cortical membranes, with IC50 values below 1 mg/mL (Wake et al., 2000). Kennedy’s 2002 Pharmaton extract, by contrast, showed low nicotinic and muscarinic binding relative to earlier accessions — and that extract still moved mood and some CDR factors (Kennedy et al., 2002). Kennedy’s 2003 programme therefore screened dried-leaf samples for cholinergic activity first, then tested the most active leaf at 600, 1000, and 1600 mg. In-vitro IC50 values were 0.18 mg/mL (nicotine site) and 3.47 mg/mL (scopolamine site). No cholinesterase inhibition was detected in that chosen extract (Kennedy et al., 2003).

Dastmalchi later fractionated a Melissa extract for acetylcholinesterase inhibition and reported time- and dose-dependent activity, expressed as physostigmine equivalents (Dastmalchi et al., 2009). That is a second extract and a second assay. It does not retroactively give Kennedy 2003 an AChE mechanism.

The cholinergic hypothesis is therefore supported as accession-dependent receptor binding and not supported as a single, transferable clinical mechanism. The 2003 paper’s last sentence — that high doses “may therefore be a valuable adjunct in the treatment of Alzheimer’s disease” — is an authors’ inference from an acute healthy-adult crossover. It is not an AD trial.

09 Other mechanistic claims

Antioxidant, antiviral, and anti-inflammatory activities are repeatedly reported in reviews (Shakeri, Sahebkar, and Javadi, 2016; Behzadi et al., 2023; Mathews et al., 2024). They are real laboratory properties of phenolics and of some oils. They do not grade anxiety, sleep, or cognition.

Rosmarinic acid’s amyloid and tau laboratory file is a separate research programme. Hase, Yamamoto, and the Noguchi-Shinohara group have published rodent and cell data on RA and amyloid/tau (Hase et al., 2019; Yamamoto et al., 2021; Noguchi-Shinohara et al., 2020). Those models motivated the Japanese high-RA extract trials. They do not convert a null ADAS-cog into a disease-modifying claim.

10 Anxiety

The isolated-plant anxiety file is smaller than the aisle implies.

Healthy-adult laboratory mood. Kennedy 2002 found Bond-Lader “calmness” elevated at the earliest time points after 300 mg of a Pharmaton extract, while 900 mg reduced “alertness” at all post-dose times (Kennedy et al., 2002). Kennedy 2003, using cholinergically active dried leaf, found increased “calmness” at all post-dose times after 1600 mg (Kennedy et al., 2003). Kennedy 2004, in 18 healthy volunteers on the Defined Intensity Stressor Simulation, found that 600 mg increased self-rated calmness and reduced alertness during the stressor (Kennedy, Little, and Scholey, 2004). These are acute, small, crossover, often industry-adjacent sessions. Grade: EMERGING as a laboratory mood description. They are not a GAD trial.

Open-label Cyracos. Cases, Ibarra and colleagues gave Cyracos for 15 days to 20 volunteers with mild-to-moderate anxiety and sleep disturbance. Clinician-rated anxiety manifestations fell 18%, associated symptoms 15%, insomnia 42%; 19 of 20 were counted as responders (Cases et al., 2011). There was no placebo arm. Expectancy, regression to the mean, and the investigators’ prior mouse Cyracos paper (Ibarra et al., 2010) are live threats. Grade: NOT SUPPORTED as confirmatory evidence; EMERGING as a signal-generating pilot.

Clinic populations, lemon balm as the named isolated product. Alijaniha’s 14-day lyophilised aqueous extract (500 mg twice daily as studied) in 55 completers with benign palpitations reduced episode frequency and the number of anxious patients versus placebo (Alijaniha et al., 2015). Haybar’s eight-week trial in 80 patients with chronic stable angina (3 g daily as studied) reported reductions in DASS-21 depression, anxiety, and stress scores and in Pittsburgh sleep disturbance versus placebo (Haybar et al., 2018). Safari’s 12-week hydroalcoholic extract (700 mg/day as studied) in type 2 diabetes with depressive symptoms (44 completers of 60 randomised) moved BDI-II and BAI versus toasted-flour placebo but not PSQI, fasting glucose, or hs-CRP (Safari et al., 2023). Araj-Khodaei compared Melissa and lavender with fluoxetine for mild-to-moderate depression in an eight-week pilot (Araj-Khodaei et al., 2020). These are small, mostly single-site, subjective-primary trials. Grade: EMERGING. They do not establish lemon balm as an anxiolytic medicine.

Pooled estimate. Ghazizadeh’s 2021 systematic review and meta-analysis reported standardised mean differences of −0.98 for anxiety and −0.47 for depression versus placebo (Ghazizadeh et al., 2021). A pooled SMD is not a licence to ignore open-label pilots, combination contamination, and small-n subjectivity. The meta-analysis is EMERGING as a quantitative summary of a heterogeneous harvest. It is not STRONGLY SUPPORTED.

11 Acute laboratory stress

Kennedy, Little, and Scholey’s 2004 DISS experiment is the cleanest isolated-plant acute-stress paper in the reviewed record. Eighteen healthy adults; 300 mg, 600 mg, and placebo; seven-day washout; mood during a 20-minute multi-task stressor. The 600 mg condition increased calmness and reduced alertness; 300 mg increased speed of mathematical processing without loss of accuracy (Kennedy, Little, and Scholey, 2004). Scholey’s later food-matrix studies extended the same laboratory logic into a beverage and a yoghurt (Scholey et al., 2014).

What this is: a short, healthy-adult, self-report-plus-task signal under a defined laboratory stressor. What it is not: occupational stress, GAD, or a hypnotic. Alertness fell at the dose that raised calmness. That is the opposite of a clean “calm without sedation” slogan, and it matches the 900 mg alertness reduction in Kennedy 2002. Grade: EMERGING.

12 Sleep

Isolated lemon balm as a hypnotic has a thin controlled file.

Cases’s open-label Cyracos study reported a 42% fall in insomnia ratings (Cases et al., 2011). Haybar reported improved Pittsburgh scores as a secondary cluster in angina patients (Haybar et al., 2018). Safari did not move PSQI (Safari et al., 2023). Di Pierro’s phytosome crossover is a formulation experiment motivated by claimed poor hydroxycinnamic-acid bioavailability (Di Pierro et al., 2024). Shirazi’s menopausal quality-of-life trial compared a Melissa–fennel combination with citalopram — a combination row, not this section (Shirazi et al., 2021).

The HMPC traditional-use indication “to aid sleep” is a historical-pharmacy sentence, not a PSG programme (European Medicines Agency HMPC, 2013). Isolated lemon balm as a hypnotic: NOT SUPPORTED. Isolated lemon balm as a small subjective sleep-quality signal in selected clinic samples: EMERGING and entangled with anxiety scores.

The large, often-cited “lemon balm sleep” studies are combination products. They are graded in Part Four.

13 Cognition

Acute healthy-adult cognition is the Northumbria file. It is real, small, and internally inconsistent across products.

Kennedy 2002 (Pharmaton extract, low cholinergic binding): sustained improvement in Accuracy of Attention after 600 mg; time- and dose-specific reductions in Secondary Memory and Working Memory factors; calmness up at 300 mg; alertness down at 900 mg (Kennedy et al., 2002).

Kennedy 2003 (cholinergically active dried leaf, no AChE inhibition): 1600 mg improved memory performance and calmness at all post-dose times; lower doses produced decrements in speed of timed memory tasks and on rapid visual information processing (Kennedy et al., 2003). The authors themselves noted that different preparations of the same species can exhibit different properties.

The red-team challenge on acute cognition is therefore internal to the file: the same laboratory, two products, opposite memory directions, and an explicit product-dependence warning. Acute CDR factors in 20 young adults are EMERGING as psychopharmacology. They are NOT SUPPORTED as a claim about everyday cognition, occupational performance, or dementia.

Disease and ageing cognition, isolated high-RA extract. Akhondzadeh’s 2003 Iranian RCT of a Melissa extract (60 drops/day as studied) in mild-to-moderate Alzheimer’s disease is a small, positive, single-site trial that has not been reproduced at that design in the reviewed record (Akhondzadeh et al., 2003). Noguchi-Shinohara 2020 randomised 23 patients with mild AD dementia to a high-RA Melissa extract (500 mg rosmarinic acid daily as studied) or placebo for 24 weeks: safety was acceptable; cognitive measures did not differ; mean NPI-Q improved by 0.5 points in the extract group (Noguchi-Shinohara et al., 2020). Noguchi-Shinohara 2023 randomised 323 older adults with subjective or mild cognitive impairment to the same RA parameter for 96 weeks: no significant difference on the primary ADAS-cog or on other cognitive measures; a CDR-SB subgroup without hypertension was reported as statistically different and is a post-hoc sliver, not a primary win (Noguchi-Shinohara et al., 2023). Isolated lemon balm as a cognitive therapeutic in AD or MCI: NOT SUPPORTED on the adequately sized modern primary endpoint; Akhondzadeh remains EMERGING and unreplicated at that design.

14 Agitation in dementia

This is the sharpest internal contradiction in the clinical file, and it is an oil file, not a leaf-extract file.

Ballard and colleagues randomised 72 UK care-facility residents with severe dementia and clinically significant agitation to Melissa essential oil or sunflower-oil placebo, combined with a base lotion and applied to face and arms twice daily for four weeks. Seventy-one completed. A 30% CMAI reduction occurred in 21/35 (60%) on Melissa and 5/36 (14%) on placebo; mean CMAI fell 35% versus 11% (Ballard et al., 2002). Quality-of-life mapping also favoured the oil. This is a positive, modestly sized, aroma-and-touch intervention. Blinding of a lemon-scented oil is intrinsically difficult.

Burns and colleagues designed the replication the first paper lacked: three English old-age-psychiatry centres; 114 randomised (94 at week 4, 81 at week 12); Melissa aromatherapy versus donepezil versus double placebo; raters wore nose clips; primary outcome Pittsburgh Agitation Scale at four weeks. There were no significant differences between aromatherapy, donepezil, and placebo at week 4 or week 12. All three groups improved: 18% on PAS and 37% on NPI over 12 weeks. The authors attributed the placebo movement to non-specific benefits of touch and interaction (Burns et al., 2011).

Ball EL’s Cochrane review of aromatherapy for dementia is the matching systematic caution (Ball et al., 2020). Isolated Melissa oil for agitation: NOT ESTABLISHED. The 2002 signal does not survive the 2011 blinding-controlled design. Leaf-extract AD agitation is not this section; Noguchi-Shinohara 2020’s 0.5-point NPI-Q movement is not a CMAI programme.

15 Gastrointestinal symptoms

Traditional European use includes mild bloating and flatulence (European Medicines Agency HMPC, 2013). Modern isolated-plant RCTs for dyspepsia or IBS are not the spine of this review.

Savino’s colic trial used ColiMil — Matricaria recutita, Foeniculum vulgare, and Melissa officinalis — in breastfed infants (Savino et al., 2005). That is a three-herb combination. It does not grade lemon balm. Gordon’s Cochrane review of dietary modifications for infantile colic is the matching systematic context (Gordon et al., 2018). Isolated lemon balm for GI symptoms: NOT SUPPORTED as a modern RCT claim; EMERGING as a traditional-use indication without a contemporary isolated-plant disease trial in the reviewed record.

16 Other labelled uses

Topical dried Melissa extract for herpes labialis has a separate, older dermatology file: Wölbling and Leonhardt’s open and placebo-controlled studies, and Koytchev’s randomised trial of 1% Lo-701 cream in 66 patients with recurrent herpes labialis, which reported a day-2 combined symptom-score difference (verum 4.03 versus placebo 4.94) (Wölbling and Leonhardt, 1994; Koytchev, Alken, and Dundarov, 1999). The HMPC explicitly declined to transfer those product-specific topical data into the oral leaf article (European Medicines Agency HMPC, 2013). This article follows that boundary. Antiviral reviews (Behzadi et al., 2023) remain laboratory and topical context.

Palpitations (Alijaniha et al., 2015) were treated above as a subjective autonomic/anxiety-adjacent endpoint, not as a cardiology indication.

17 Valerian and lemon balm

This is the largest “lemon balm sleep” literature, and it is not a lemon-balm literature.

Kennedy, Little, Haskell, and Scholey tested a Melissa plus Valeriana combination during laboratory-induced stress — because, as they wrote, the two herbs are most often sold together (Kennedy et al., 2006). That paper is a combination anxiolytic/stress experiment. It cannot be re-analysed as isolated Melissa.

Müller and Klement evaluated a combined valerian/lemon-balm preparation in an open, multicentre study of 918 children under 12 with restlessness and nervous dyskoimesis, individually dosed by investigators. Core symptoms were rated as improved in 80.9% (dyssomnia) and 70.4% (restlessness) (Müller and Klement, 2006). There is no placebo. Expectancy, natural history, and investigator-chosen amounts are uncontrolled. Grade as combination open-label: EMERGING. Grade as isolated lemon balm: not applicable.

Gromball and colleagues later treated primary-school children with a fixed combination of valerian root and lemon-balm extracts for seven weeks and reported improvements in hyperactivity, concentration, and impulsiveness (Gromball et al., 2014). Same attribution rule.

Taavoni, Nazem Ekbatani, and Haghani studied valerian/lemon balm for sleep disorders during menopause (Taavoni, Nazem Ekbatani, and Haghani, 2013). Combination row.

Houghton’s 1999 review of valerian’s scientific basis is valerian’s file, not Melissa’s (Houghton, 1999). It is cited only to keep the plants from collapsing.

18 Other multi-herb preparations

Savino’s ColiMil colic trial is chamomile, fennel, and lemon balm (Savino et al., 2005). Shirazi’s menopausal trial is lemon balm with fennel fruit extract versus citalopram (Shirazi et al., 2021). Ranjbar’s depression/anxiety work used Melissa plus Nepeta menthoides (Ranjbar et al., 2018). Perry’s 2018 dementia pilot used a combined herbal preparation (Perry et al., 2018). Watson’s 2019 RCT included lavender and lemon balm as a pair (Watson et al., 2019). None of these rows may be entered in a lemon-balm-alone matrix.

19 Attribution rules

The rule is mechanical. If a second named plant, a second essential oil, or a second extract is in the capsule, the outcome is a combination outcome. Ghazizadeh’s pooled SMD cannot be read as if every included arm were isolated Melissa until each trial is checked (Ghazizadeh et al., 2021). Traditional sedative claims that cite “lemon balm” while describing a night tablet that also contains valerian are combination advertising, not a grade. The same rule applies to a meta-analysis that mixes open Cyracos pilots with valerian combinations and then reports a single anxiety SMD. Pooling does not create isolation.

20 Safety: sedation, thyroid, interactions, pregnancy

Sedation and alertness. The isolated-plant laboratory file shows reduced alertness at higher acute amounts (900 mg Pharmaton; 600 mg under DISS) and increased calmness at selected amounts (Kennedy et al., 2002; Kennedy, Little, and Scholey, 2004). That is a mood-scale movement in healthy adults, not a driving-impairment programme and not a demonstration of harmless “relaxation without sedation.” Combination paediatric open studies report good tolerability as authors’ impressions (Müller and Klement, 2006). Those impressions do not grade isolated Melissa in adults operating machinery. Sedation risk: EMERGING as a laboratory alertness reduction; not quantified as a clinical hypnotic adverse-effect rate.

Thyroid. Freeze-dried extracts of Melissa officinalis (with Lycopus and Lithospermum) inhibit bovine TSH binding and adenylate-cyclase stimulation in human thyroid membranes (Auf’mkolk et al., 1984) and inhibit the receptor-binding and biological activity of Graves’ immunoglobulins (Auf’mkolk et al., 1985). Santini later included Melissa among substances that dose-dependently inhibited TSH-stimulated cAMP production in CHO cells expressing the recombinant human TSH receptor (Santini et al., 2003). These are in-vitro thyroid-disruptor assays. They are not a demonstrated clinical hypothyroidism signal from tea or typical oral extracts in the reviewed record. They are a theoretical interaction with the TSH receptor / Graves’-IgG axis and a reason not to treat lemon balm as thyroid-inert. Grade of the laboratory finding: SUPPORTED. Grade of clinical thyroid harm from ordinary oral use: NOT ESTABLISHED — an evidence gap, not a reassurance.

Drug interactions. No well-powered interaction programme sits in the reviewed record. Theoretical overlaps are CNS depressants (alertness reduction in acute studies) and thyroid-axis drugs (in-vitro TSH-receptor interference). Topical herpes creams are a different exposure. Combination products add valerian’s interaction ledger. Product adulteration and misidentification are quality risks, not pharmacology.

Pregnancy and lactation. The HMPC traditional-use article does not supply a modern reproductive-toxicology or pregnancy-exposure file that would close the gap (European Medicines Agency HMPC, 2013). Isolated Melissa in pregnancy: evidence gap. Traditional use is not a safety demonstration.

General toxicology. Lobach and colleagues reported that an aqueous lemon-balm extract was not genotoxic in a standard OECD battery up to 5000 μg/plate or 5000 μg/mL and completed a 90-day dietary rat study as a food-ingredient package (Lobach et al., 2024). That is food-ingredient toxicology. It is not a clinical safety licence for disease claims. Noguchi-Shinohara 2015 and 2020 reported acceptable short- and intermediate-term tolerability of a high-RA extract in small human samples (Noguchi-Shinohara et al., 2015; Noguchi-Shinohara et al., 2020). LiverTox was sought via local Firecrawl; the reviewed recorded Bookshelf record was not a lemon-balm chapter (acetaminophen record returned on the attempted identifier). No lemon-balm DILI case series was verified in the reviewed record.

Short-term tolerability in the small isolated-plant trials: SUPPORTED as “generally tolerated in those samples.” Long-term disease-outcome safety: NOT ESTABLISHED.

21 Regulation and traditional-use status

The European Union herbal article on Melissa officinalis L., folium is a traditional-use article: relief of mild symptoms of mental stress and to aid sleep; symptomatic treatment of mild gastrointestinal complaints including bloating and flatulence (European Medicines Agency HMPC, 2013). Traditional use is a regulatory pathway. It is not a GRADE assessment of RCTs.

The HMPC declined to generalise the topical herpes file to the oral leaf article (European Medicines Agency HMPC, 2013). This document follows that refusal.

Lemon balm is sold as a food, tea, and dietary supplement in other jurisdictions. Those statuses are market facts. They are not efficacy.

22 Research matrices

Anxiety — lemon balm alone

Sleep — lemon balm alone versus combinations

Cognition — acute laboratory versus disease

Formulation matrix

A row that lists “tea” and a row that lists “Cyracos” are not two doses of one medicine. A row that lists valerian plus lemon balm is not a missing lemon-balm-alone cell. The formulation matrix is the enforcement surface for Figure 1 and Figure 2. If a later paper changes a cell, it must change the object first.

Safety matrix

23 Adversarial resolutions

Do small subjective trials support a clinical anxiolytic or hypnotic claim? No. Resolved on Kennedy 2002–2004 (n = 18–20, Bond-Lader / DISS), Cases 2011 (open, n = 20), and the small Iranian clinic trials (Alijaniha 2015; Haybar 2018; Safari 2023). Positive directions exist. Sample sizes, self-report primaries, single sites, and at least one unblinded pilot forbid a clinical anxiolytic or hypnotic claim. Grade of the threat: STRONGLY SUPPORTED. Surviving use of the word “signal”: EMERGING.

Can combination-product effects be attributed to lemon balm? No. Resolved on Müller 2006, Gromball 2014, Taavoni 2013, Kennedy 2006, Savino 2005, Shirazi 2021, and Ranjbar 2018. Attribution of those outcomes to Melissa is NOT SUPPORTED.

Do acute laboratory cognition findings generalise to everyday or disease cognition? No. Resolved on the internal contradiction between Kennedy 2002 (memory factors down; low cholinergic binding) and Kennedy 2003 (memory up at 1600 mg; different product), and on Noguchi-Shinohara 2020/2023 (cognition null on the modern isolated high-RA programme). Acute CDR factors are EMERGING psychopharmacology. Everyday or disease cognition: NOT SUPPORTED.

Do traditional sedative claims survive controlled evidence? As HMPC traditional-use language, they survive as a regulatory historical sentence (European Medicines Agency HMPC, 2013). As a demonstrated hypnotic or general sedative in isolated-plant RCTs, they do not. Alertness falls in some acute sessions; sleep-quality primaries are sparse; the large paediatric sleep/restlessness file is a combination open study. Traditional sedative claim as efficacy: NOT SUPPORTED. Traditional-use status: established as regulator language.

Do in-vitro GABA findings explain human outcomes? No. Resolved on Awad 2007/2009 (homogenate GABA-T), Ibarra 2010 (mouse EPM, open field null), and Abuhamdah 2008 (oil; TBPS displacement and inhibition of GABA-evoked currents; not the leaf extract). No human study in the reviewed record measured GABA, GABA-T occupancy, or a mediation path from RA plasma level to a clinical scale. Mechanistic GABA assumption as explanation: NOT SUPPORTED. Mechanistic GABA assumption as a laboratory hypothesis: plausible and unproven.

24 Where the record stands

The leaf’s identity and the rosmarinic-acid marker are established. The oil’s citral/citronellal character is established and is a different exposure. GABA-T inhibition by RA and related triterpenes is strongly supported in homogenates and not established in humans. Cholinergic binding is accession-dependent. Cholinesterase inhibition is not a consistent property of the extracts that moved human mood.

Lemon balm alone has an emerging healthy-adult acute mood and laboratory-stress file, an emerging set of small clinic trials on anxiety-adjacent scores, and a negative modern primary cognition test in SCI/MCI. Melissa oil for agitation is not established: positive in 2002, null when blinding was engineered in 2011. The large sleep and restlessness numbers that circulate with the English name belong to valerian plus lemon balm and to other multi-herb products.

What would change the grades: a multi-site, placebo-controlled GAD or insomnia RCT of a chemically specified isolated extract with a pre-specified primary; a factorial Melissa × valerian trial that can attribute; a human mediation study from RA exposure to a clinical scale; a second AD/MCI trial that wins a pre-specified cognitive primary. Until those exist, the aisle is ahead of the file.

Standing constraint This document describes published research. It is not medical advice. No human use, dose, route or schedule is recommended anywhere in this document.

25 Evidence handling

Study types are labelled in the sentence that reports them. Homogenate enzymology, receptor binding, mouse plus-maze, healthy-adult laboratory tasks, open-label pilots, single-site clinic RCTs, combination open studies, aromatherapy oil trials, traditional-use articles, and meta-analyses of mixed products are not interchangeable. Conflict is presented in the same section as the positive finding. Amounts are study parameters, not recommendations.

References in the numbered list are generated from verified NCBI records. Identifiers that resolved to unrelated papers were dropped before drafting. The HMPC melissa-leaf article and the 2013 assessment report were read as regulator sources. This title is a South Beach Longevity science article, not a peptide article.

Sibling scope: Chamomile, L-Theanine, Insomnia, and Melatonin are neighbouring titles. Valerian as a plant is a neighbouring file. This article is the Melissa officinalis depth title.

Where PubMed returned a Methods-only abstract, the matching EuropePMC record supplied the Results text used for Ballard 2002, Burns 2011, Kennedy 2002–2004, and Noguchi-Shinohara 2023. Numbers in those sections are taken from those abstracts.

References