Rosmarinic Acid + EGCG for Sleep: One Small RCT

Millions of people sleep poorly without meeting criteria for insomnia: they take too long to drift off, wake unrefreshed, or drag through the day. That gap creates a large market for supplements. A proprietary blend containing rosmarinic acid and epigallocatechin gallate (EGCG) has one randomized trial behind it. The trial found statistically significant improvements on two self-reported sleep scores, but it missed both primary endpoints.

That is enough to justify scientific curiosity. It is nowhere near enough to call the blend a proven sleep aid.

What the trial actually found

In a 2021 randomized, double-blind, placebo-controlled trial by Tubbs et al., adults with mild sleep complaints took either a 485 mg proprietary polyphenol botanical blend capsule or placebo once daily for 30 days. Participants needed a Pittsburgh Sleep Quality Index (PSQI) score of at least 3, but people with diagnosed sleep disorders were excluded. Of 100 enrolled participants, about 89 completed the study.

The capsule contained at least 120 mg total polyphenols, including at least 65 mg of rosmarinic acid and EGCG combined, plus no more than 4.85 mg caffeine. The investigators did not disclose the separate doses of rosmarinic acid and EGCG. That makes the product difficult to reproduce and prevents any useful dose-response conclusion.

The study pre-specified two primary outcomes: sleep-onset latency from a sleep diary and the percentage of REM sleep estimated with a Fitbit Charge 2. Neither outcome was reported as statistically significant versus placebo.

The blend group did better on two pre-specified secondary questionnaire outcomes: global PSQI score improved relative to placebo with p=0.008, and Insomnia Severity Index (ISI) score improved with p=0.044. The study reported no significant adverse events.

Those results are worth separating carefully. PSQI and ISI capture how participants experience their sleep, which matters. They do not establish that the blend improved sleep latency or REM sleep, and the study did not show a statistically significant benefit on either primary measure.

Statistical significance is not the same as meaningful improvement

The reported p-values tell us that the secondary score differences met the study's threshold for statistical significance. They do not tell us whether participants improved enough to feel a meaningful difference in daily life.

For context, published minimum clinically important difference estimates are roughly 1.74 to 3.0 points for PSQI and about 6 points for ISI. The available trial reporting does not establish whether the treatment-placebo differences reached either range. Without the absolute score changes and their uncertainty, readers cannot judge the practical size of the effect.

Primary-endpoint failure also changes how much weight to give positive secondary results. When a trial tests several outcomes and misses its central hypotheses, a statistically significant secondary result can reflect a real effect, chance, or a mixture of both. The publication does not establish whether it adjusted its secondary-outcome analyses for multiple comparisons.

This does not prove the blend failed to help anyone. It means the trial generated a signal rather than a reliable estimate of benefit.

The biology is plausible, but human proof is thin

Rosmarinic acid occurs in several herbs, including lemon balm (Melissa officinalis). The trial proposed that rosmarinic acid could influence GABAergic and cholinergic signaling, while EGCG could affect the stress-responsive hypothalamic-pituitary-adrenal axis through corticosterone-related pathways. These are biologically reasonable ideas. They are not established human mechanisms for this product at this dose.

For rosmarinic acid, a 2022 mouse study by Borgonetti et al. found reduced microglial senescence and lower pro-inflammatory cytokine expression in a neuropathic-pain model. The experiment did not measure sleep, did not involve humans, and cannot tell us whether the blend alters sleep biology in people.

The broader lemon balm literature does not close that gap. A 2024 review of lemon balm and a 2025 review focused on cognition and sleep both describe a clinical evidence base dominated by small, uncontrolled, or non-randomized studies. Lemon balm contains many compounds besides rosmarinic acid, so even a positive lemon balm result would not identify rosmarinic acid as the active ingredient.

A separate 2026 placebo-controlled probiotic trial discusses HPA-axis, neuroimmune, and neurotransmitter pathways relevant to sleep. It provides context for why researchers study these pathways, but it does not test rosmarinic acid or EGCG. A pathway can be sleep-relevant without proving that a supplement moves that pathway enough to improve sleep.

Even the mechanistic comparison argues for modest expectations

The GABA hypothesis deserves a reality check. In a 2021 systematic review and meta-analysis, Nolan et al. found modest sleep benefits from micronized progesterone in randomized trials. Progesterone metabolites directly modulate GABA-A receptors; the proposed influence of rosmarinic acid is more indirect and less well characterized in humans.

That comparison does not quantify an expected effect for the polyphenol blend. It does suggest that a large sleep effect would be surprising. Mechanistic plausibility is a reason to run a better trial, not a substitute for one.

Why the evidence remains low confidence

• One short sponsored trial: The entire human randomized evidence base for this rosmarinic acid-plus-EGCG blend is a 30-day study sponsored by the product manufacturer.
• About 89 completers: Eleven of the 100 enrolled participants did not complete the study, reducing the effective sample below enrollment.
• Primary outcomes were null: Sleep-diary sleep-onset latency and Fitbit-estimated REM percentage did not improve significantly versus placebo.
• Subjective secondary outcomes drove the positive result: PSQI and ISI improved statistically, but the available reporting does not show whether the changes met clinically meaningful thresholds.
• The formula cannot be replicated: The individual rosmarinic acid and EGCG doses remain undisclosed.
• Objective sleep measurement had limits: Fitbit Charge 2 sleep staging has moderate accuracy compared with polysomnography, the laboratory standard.

The mild baseline symptom threshold matters, too. A PSQI score of 3 qualifies as a sleep complaint but sits well below the level of many clinical sleep populations. People with mild symptoms can improve with expectation, routine changes, regression toward their usual sleep, or placebo effects. A placebo-controlled design helps, but it does not erase the uncertainty created by a small, short, unreplicated trial.

What to do with this information

The blend appears to have been tolerated for 30 days in this study, with no significant adverse events reported. Safety over one month in roughly 89 completers does not establish long-term safety, nor does it prove benefit.

For people trying to improve sleep, the evidence is stronger for interventions that have survived repeated testing. A 2025 network meta-analysis of 18 randomized trials involving 1,214 adults with sleep disorders found that exercise interventions improved sleep efficiency by 2.85 percentage points (95% CI, 0.85 to 4.84) and reduced wake after sleep onset by 10.16 minutes (95% CI, -15.68 to -4.64). Those results apply to adults with sleep disorders, not to every person with occasional bad sleep, but they rest on a broader evidence base than this supplement.

Verdict: Low confidence. The rosmarinic acid-EGCG blend has a plausible hypothesis and one encouraging secondary signal, but its only randomized trial missed the primary endpoints, used a proprietary formula, and has no independent replication. Buying it means running a personal experiment, not following settled sleep science.