Sleep Ingredients — Evidence Review
Sleep ingredients:
what the human research actually shows
A survey of the compounds sold for sleep — the real trials, the doses, the outcomes, and the limitations behind each. Ranked by evidence, not by popularity.
At a glance
Sorted by strength of human sleep evidence. Tap a name to jump to its studies.
| Ingredient | Evidence | Best-supported effect | Primary sleep role |
|---|---|---|---|
| Ashwagandha (Shoden®) | WASO ↓, efficiency ↑, cortisol ↓ | Maintenance | |
| Glycine | onset ↓, slow-wave, next-day vigilance | Onset / architecture | |
| Saffron | sleep quality, mood on waking | Serotonergic calm | |
| L-tryptophan | WASO ↓ at ≥1 g | Serotonin precursor | |
| L-theanine | onset, calm, some efficiency | Alpha-wave calm | |
| Magnesium | onset ↓ ~17 min | Cofactor | |
| Lavender (Silexan) | sleep quality via anxiety ↓ | Anxiolytic | |
| Valerian | onset, subjective quality | Sedative (contested) | |
| Tart cherry | WASO ↓ (older adults) | Anti-inflammatory | |
| Lemon balm | ISI ↓, slow-wave ↑ (1 RCT) | GABAergic calm | |
| Apigenin | mechanism only; RCT negative | GABA-A adjunct | |
| Passionflower | mild; data insufficient | GABAergic calm | |
| GABA (oral) | combination-only signal | Inhibitory (poor uptake) | |
| Melatonin | circadian timing (not general) | Circadian signal |
Strongest human evidence
Replicated randomised trials, ideally with objective (PSG or actigraphy) measures. Still mostly small — this is the supplement field, not pharma — but these clear the bar.
Ashwagandha (Shoden®) 120 mg
The best-evidenced natural ingredient for sleep maintenance specifically — and one of the few that lowers cortisol rather than sedating.
Systematic review & meta-analysis · 5 RCTs · n=400
A small but significant improvement in overall sleep (SMD −0.59, 95% CI −0.75 to −0.42), with benefit on quality, onset latency, and mental alertness on rising in the insomnia subgroup. Larger at ≥600 mg/day and ≥8 weeks.
RCT, double-blind, placebo-controlled · n=150 · Shoden® 120 mg · 6 wk · actigraphy
72% self-reported sleep-quality improvement vs 29% placebo, with actigraphy-confirmed gains in efficiency, total sleep time, onset latency and wake-after-sleep-onset.
The honest counterweight. Effective for maintenance and, unusually, associated with better morning alertness rather than grogginess — but the evidence base is small, geographically concentrated, and partly sponsored, and the wider literature carries emerging liver-injury and thyroid safety signals worth monitoring.
Moderate evidence — real, but small, mixed, or indirect
Genuine human signals, but the data is thinner: subjective endpoints, small samples, single research groups, or benefit that runs through anxiety rather than sleep directly.
Glycine 3,000 mg
Well-motivated for onset, slow-wave sleep and next-day clarity — but its fame outruns its trial base.
Randomised crossover · polysomnography · n≈11 · 3 g
3 g shortened latency to sleep onset and slow-wave sleep (p=0.01 / 0.019) and improved efficiency without altering architecture; a representative hypnogram showed fewer awakenings.
Placebo-controlled · 75% sleep restriction · n=10 (7 analysed) · 3 g
Improved next-day psychomotor vigilance (p=0.003 / 0.05) and day-1 fatigue (p=0.022).
Context. All positive glycine sleep data traces to a small cluster of Japanese trials from one company; a systematic review rated the body high risk of bias. Real for onset and thermoregulation; the maintenance signal is suggestive only. Note: standard magnesium bisglycinate delivers ~half the studied 3 g.
Saffron 30 mg
Replicated across brands for sleep quality, with an unusual mood-and-alertness-on-waking benefit.
3-arm RCT · n=165 · 20 & 30 mg (Safr'Inside™) · 4 wk
Reduced insomnia symptoms (Athens Insomnia Scale) and improved sleep-quality and perceived-stress scores vs placebo.
RCT · n=120 · affron® 14 & 28 mg · 28 days
Improved sleep-quality ratings and mood after awakening, with changes in evening cortisol/melatonin.
RCT pilot · n=52 older adults · 30 mg · EEG tracker
Objective support: improved sleep quality (p=0.02), efficiency (p=0.04) and latency to persistent sleep (p=0.003).
Context. A serotonergic route distinct from the GABAergic botanicals, with a genuine mood-on-waking angle. One athlete trial found no sleep effect — real, but not universal.
L-tryptophan ≥1 g
One of the older, better-replicated sleep aids — but the benefit is narrow and regulatory status varies by market.
Systematic review, meta-analysis & meta-regression
Tryptophan shortened wake-after-sleep-onset, with a clear dose threshold: ≥1 g produced markedly less WASO than <1 g (28.9 vs 56.6 min, p=0.001). No effect on other sleep components.
Review of 40+ controlled studies
Doses ≥1 g reduce sleep-onset latency, especially in mild situational onset-insomnia; effect can appear on the first night.
Context. Serotonin/melatonin precursor with a real, dose-dependent WASO effect — mechanistically adjacent to saffron. Narrow benefit and market-dependent legality keep it a secondary option.
L-theanine 200 mg
A reliable onset-and-calm ingredient for the "tired but wired" mind — more onset than maintenance.
RCT, placebo-controlled · n=30 · 200 mg · 4 wk
Improved sleep quality (PSQI) and reduced latency and sleep-medication subscores, alongside lower anxiety and better cognition.
RCT, double-blind · n=98 boys with ADHD · 400 mg · actigraphy
Objective: increased sleep percentage and efficiency vs placebo.
Context. Raises alpha activity within 30–45 min and blunts sympathetic tone; well-tolerated. Calms entry to sleep more than it holds it.
Magnesium 150–250 mg elem.
Modest, mostly onset benefit — with genuinely low-quality evidence, strongest in the magnesium-deficient.
Systematic review & meta-analysis · 3 RCTs · n=151 older adults
Reduced onset latency ~17 min vs placebo (p=0.0006); total sleep time gain non-significant.
RCT, placebo-controlled · bisglycinate 250 mg elem. · 4 wk
Reduced Insomnia Severity Index (−3.9 vs −2.3, p=0.049), larger benefit at low baseline magnesium.
Context. Real GABA-A/NMDA mechanisms; bisglycinate is the best-matched form and doubles as a glycine source. Threonate is studied for cognition, not sleep. Best seen as a cofactor.
Lavender (Silexan®) 80 mg
Sleep benefit that runs mostly through anxiety reduction — with an evidence base almost entirely from one manufacturer.
RCTs · n up to 539 · Silexan 80 mg · 10 wk
80 mg improved sleep quality and reduced anxiety, comparable to low-dose lorazepam/paroxetine on anxiety, with a sustained sleep-subscore benefit.
Context. Genuinely useful where hyperarousal is anxiety-driven; not a direct sleep-maintenance agent. The near-total sponsor involvement is the main caveat — the direction of effect is consistent, the independence isn't.
Valerian 300–600 mg
The classic herbal sedative — with a famously contradictory literature and a morning-grogginess risk at higher doses.
Systematic review & meta-analysis · 18 RCTs
Valerian reduced sleep latency (SMD −0.71) and improved subjective quality (SMD −0.46) vs placebo.
Context. A real GABAergic sedative with genuine but unreliable effects. Higher doses can bleed into the morning as grogginess — a poor fit for a "wake refreshed" goal, and the reason it's often excluded from morning-alertness-focused formulas.
Tart cherry 2×240 mL / extract
A modest, inconsistent WASO benefit in older adults — and not, as often claimed, because of its melatonin content.
RCT crossover pilot · n=15 older adults with insomnia · 2 wk
Reduced insomnia severity and WASO vs placebo (effect sizes ~0.5–1.0 for WASO/SE).
RCT · tart cherry powder
No significant sleep improvement — a counterweight to the positive pilots.
Context. Promising in specific older-adult populations, but the picture is mixed and the popular "natural melatonin" framing is misleading — the melatonin dose is far too small to explain any effect.
Thin evidence — strong mechanism, sparse human data
Mechanistically plausible and popular, but the human sleep trials are absent, negative, or judged insufficient by systematic reviews. Interesting; not proven.
Lemon balm 400 mg
One encouraging recent RCT — against a backdrop of reviews still calling the data insufficient.
Double-blind crossover · n=30 chronic insomnia · 400 mg · smartwatch
400 mg reduced ISI scores and increased slow-wave sleep (~15%); 87% reported improved quality vs 30% placebo.
Multiple reviews
Consistently conclude the overall data for lemon balm (and passionflower) is insufficient to evaluate efficacy for sleep disorders.
Context. Plausible GABAergic mechanism (via rosmarinic acid) and a promising new RCT, but the standalone evidence remains early. One to watch, not to lead on.
Apigenin 50 mg
A textbook case of a strong mechanism outrunning its evidence — the one relevant human trial was negative.
RCT pilot · n=34 chronic insomnia · chamomile 540 mg (~23 mg apigenin) · 28 days
The closest human test: no significant difference vs placebo on onset, total sleep time or efficiency.
RCTs · chamomile extract · anxiety endpoints
Standardised chamomile reduced generalised-anxiety symptoms — a plausible indirect route to sleep.
Context. The GABA-A mechanism is real and confirmed in animals, but human sleep evidence at the marketed dose is effectively absent — its popularity traces to a podcast recommendation. The clearest example of the "mechanism-theatre" risk in this whole list.
Passionflower tea / extract
Traditional and plausible, but the human sleep data doesn't yet support a firm conclusion.
Small RCTs / pilots · tea & extract
A low-dose passionflower tea produced a small subjective sleep-quality improvement over placebo in a short pilot.
Context. GABAergic mechanism shared with valerian and lemon balm, and a benign safety profile, but standalone sleep evidence is preliminary.
GABA (oral) 100–700 mg
The brain's main inhibitory signal — but swallowing it is undermined by poor absorption and blood-brain-barrier penetration.
Small RCTs · oral GABA, often + L-theanine
Some reduction in sleep-onset latency reported, but robust effects appear mainly in combination (e.g. GABA + theanine), not for GABA alone.
Context. Mechanistically central to sleep, pharmacologically weak as an oral supplement. The interesting question is delivery, not the molecule.
Melatonin 0.3–5 mg
Strong evidence — but for the circadian problem, not general insomnia, and high doses can backfire.
Meta-analysis · 19 studies · n=1,683
Melatonin reduced onset latency ~7 min and increased total sleep ~8 min — real but modest for primary insomnia; much stronger for circadian conditions (jet lag, delayed phase, shift work).
Context. A timing signal, not a sedative — best for circadian misalignment at low physiological doses (~0.3 mg), where its evidence is genuinely strong. The most-misused ingredient in the category: more is not better, and often worse.
What the whole picture says
Three honest patterns run through the research. Effect sizes are almost universally small — this is supplementation, not sedation, and no natural ingredient rivals a prescription hypnotic. Sponsorship and single-research-group evidence are common, so independence matters as much as significance. And the sharpest divide is between onset ingredients (glycine, theanine, magnesium, tryptophan, melatonin) and true maintenance evidence, which is rare — ashwagandha, and to a degree tryptophan and tart cherry, are the few with any signal on the back half of the night. Popularity tracks marketing, not evidence: apigenin and oral GABA are widely sold and among the least supported.