Self-Heal: Rosmarinic Acid, Safety and Unknowns

Safety pages on herb sites tend to have a tone problem. Either everything is "gentle and safe" or everything comes wrapped in so many warnings that the reader cannot tell a real hazard from a legal reflex. Neither is honest, and for Prunella vulgaris the second failure mode is the tempting one, because it is easy to sound rigorous by inventing cautions.

We are not going to do that. The honest answer about self-heal's safety is mostly reassuring, and saying so plainly is part of accuracy. This is a plant people eat. Its leaves go into salads and soups across several continents; its dried flower spike has been brewed as a beverage in China for centuries; and no pattern of poisoning, organ injury or serious adverse events has emerged from any of it. Refusing to state that clearly, in order to seem careful, would be its own kind of dishonesty.

What this page does instead is separate three things that usually get mashed together: what is genuinely reassuring, what is genuinely unknown, and which of the cautions floating around the internet survive contact with the actual chemistry. Two of the most commonly repeated ones do not survive, and we say why.

Table of Contents

  1. The Reassuring Part, Stated Plainly
  2. The Food and Tea Record
  3. Hazards Self-Heal Does Not Have
  4. Being a Mint-Family Plant Is Not Itself Reassuring
  5. What Is Not Known: A Numbered Findings List
  6. The Anticoagulant Question, at Its Real Strength
  7. Pregnancy and Breastfeeding: Absent Data
  8. Thyroid, Blood Pressure and the Potassium Arithmetic
  9. The Immune-Modulation Caution
  10. Wild Harvest, Lawns and Misidentification
  11. Rosmarinic Acid: What Happens to It in the Body
  12. Numbers We Refuse to Give
  13. Practical Summary
  14. Key Research Papers
  15. Connections

The Reassuring Part, Stated Plainly

Four statements we are comfortable making without hedging:

  1. The plant is a food. Young Prunella vulgaris leaves are eaten raw and cooked, and appear in wild-food and ethnobotanical surveys from Europe, Turkey, Korea, Japan and China. This is not a niche foraging fad; it is a long-standing, geographically widespread food use.
  2. The dried spike is a beverage. Xia ku cao tea is drunk in China not only medicinally but as a cooling summer drink, including in commercial bottled herbal-tea products. Very large numbers of people have consumed it casually.
  3. No toxicity signal has emerged. There is no recognised syndrome of self-heal poisoning, no cluster of liver or kidney injury cases, no recall history, and no regulatory action of the kind that has attached to comfrey, germander, kava or aristolochia.
  4. It is not on any of the lists that matter. The plant does not appear in the standard hepatotoxic-herb literature, the pyrrolizidine-alkaloid literature, or the nephrotoxic-herb literature.

Be clear about what kind of evidence that is. It is absence of a signal from uncontrolled use, which is not the same as a completed safety programme. Rare harms, slow harms and interaction-dependent harms all hide from this kind of observation. But absence of a signal across centuries of widespread food use is real information, and it is considerably more than most medicinal herbs can offer. Self-heal has a better safety case than the great majority of plants sold in capsules.

The Food and Tea Record

Worth setting out in a little more detail, because it is doing most of the work in the section above.

The plant is close to cosmopolitan — native across Europe, temperate Asia and North Africa, and present throughout North America — and it grows in exactly the places people historically gathered greens: meadows, grazed pasture, lawns, path edges. It is small, mild, faintly bitter and not unpleasant, which is why it ended up eaten rather than merely tolerated.

Two features of that record make it stronger than it first appears. First, it is independent across cultures: European foragers and East Asian cooks arrived at the same conclusion without consulting each other. Second, it includes casual, non-medicinal consumption — a summer cooling drink taken for pleasure, not a remedy taken briefly for a complaint. Casual use produces far more exposure, at higher cumulative doses over longer periods, than medicinal use ever does. If there were an obvious problem, that is where it would have surfaced.

What the food record does not cover is worth naming too: concentrated extracts, standardised capsules, tinctures taken daily for months, and anything at doses above traditional practice. Centuries of tea-drinking say nothing about a 10:1 extract, and it is a mistake to lend one the reassurance of the other.

Hazards Self-Heal Does Not Have

This section exists because readers arrive here from other herb pages carrying specific worries, and it is as much a part of accuracy to say what a plant is not guilty of as to say what it is. Each item below is a real hazard of a real herb, checked against self-heal and cleared.

Nine specific hazards, checked and cleared. A reader who has been told that "herbs are unregulated and dangerous" deserves to know that this particular plant is not carrying any of the things that actually make herbs dangerous.

Being a Mint-Family Plant Is Not Itself Reassuring

Having cleared self-heal, we should immediately spike the bad argument that most often gets used to clear it.

You will read that self-heal is safe because it is in the mint family, alongside basil, oregano, thyme, sage, rosemary and peppermint — a family of kitchen herbs. That reasoning does not work, and the counterexample is decisive: germander (Teucrium chamaedrys) is a Lamiaceae, and it caused well-documented, sometimes fatal liver injury when it was sold as a slimming aid in France, leading to its withdrawal. Germander is also the species most often implicated in the historic adulteration of skullcap, another mint-family herb whose reputation was damaged by something that was not actually skullcap.

Family membership predicts a great deal about a plant's aroma chemistry and almost nothing about its toxicology. Perilla ketone is a mint-family compound. Pulegone, the hepatotoxin in pennyroyal, is a mint-family compound. Thujone is a mint-family compound.

So the reassurance about self-heal has to come from its own record — its food history, the absence of any signal, the absence of any of the nine specific hazards above — and not from its relatives. Fortunately its own record is good. But the argument matters, because "it is just a mint" is a bad habit of reasoning that will eventually clear something it should not.

What Is Not Known: A Numbered Findings List

These are stated as findings rather than left as gaps, because "we do not know" is a result and readers deserve to see it as one. Every item below is a real hole in the record for Prunella vulgaris.

  1. No verified formal toxicology programme. We looked for published acute and repeated-dose toxicity studies of a defined self-heal preparation and could not verify a coherent set. If such work exists it is not prominent, and we are not going to cite a study we cannot confirm.
  2. No human pharmacokinetics. Nobody has published what happens to any self-heal constituent after a person drinks an infusion: no plasma concentrations, no time course, no metabolites, no elimination data. Every statement about what the herb "does in the body" is therefore inference.
  3. No dose-finding. The traditional Chinese range of roughly 9 to 15 grams of dried spike a day is a convention inherited from materia medica texts, not a dose derived from any pharmacological study. There is no established minimum effective dose and no established maximum.
  4. No drug-interaction studies. None. Not for cytochrome P450 enzymes, not for P-glycoprotein, not for platelet function, not for anything. Note carefully that this means "no known interactions" is a statement about absent research, not a clean bill of health. The phrase means nobody looked.
  5. No pregnancy or lactation data. No human data, and no animal reproductive toxicology we could verify.
  6. No paediatric data.
  7. No long-term use data. Traditional use is typically short courses for a complaint. Daily use for years, which is how supplements are actually taken, is unstudied.
  8. No product standardisation. There is no agreed marker compound, no assay requirement, and no reason to expect two commercial self-heal products to contain comparable amounts of anything. This makes every dose statement about the herb unreliable in practice even where one exists in principle.
  9. No compositional certainty for the material you buy. Rosmarinic acid content in Lamiaceae varies with genetics, growing conditions, harvest timing, drying method and storage — often several-fold. Reported figures for this plant are not consistent enough to quote.
  10. No case-report surveillance. Because the herb is used casually and is not a suspect in any known syndrome, nobody is looking for adverse events attributable to it. The absence of case reports partly reflects the absence of anyone looking, which is a weaker form of reassurance than it appears.

The Anticoagulant Question, at Its Real Strength

Self-heal is frequently listed as a herb to avoid with blood thinners, and this is the caution most in need of an honest audit — because part of the usual reasoning is simply wrong.

The coumarin argument does not work

You will see it argued that self-heal contains "coumarin-type" compounds and therefore thins the blood. Two errors are stacked here.

First, the chemistry. Rosmarinic acid is not a coumarin. It is an ester of caffeic acid and 3,4-dihydroxyphenyllactic acid — a different structural class entirely. Self-heal's characteristic phenolics are caffeic-acid derivatives and flavonoids, not coumarins.

Second, and more importantly, coumarin is not an anticoagulant. This is one of the most persistent confusions in herbal writing. Plain coumarin — the sweet-hay compound in woodruff, tonka bean and cassia cinnamon — has no meaningful anticoagulant effect in humans. The anticoagulant is dicoumarol, a 4-hydroxycoumarin dimer formed by fungal action in spoiled sweet clover hay, which is where warfarin's development began. Rotten hay chemistry is not fresh plant chemistry. So even if self-heal did contain coumarins, the inference to blood thinning would still not follow.

What is left, once the bad argument is removed

The interaction that is actually more plausible runs the other way

Here is the part that gets left out. If someone on warfarin starts eating self-heal leaves as a green, the plausible interaction is not blood thinning at all — it is vitamin K. Green leaves contain phylloquinone, warfarin works by antagonising vitamin K, and the practical effect of adding a green vegetable to the diet is a tendency to reduce the INR, not raise it. That is the opposite direction from the popular claim.

We cannot quantify it: the vitamin K content of Prunella vulgaris leaf is not in the food composition databases, and the quantities anyone actually eats are small. The practical point for a warfarin patient is the standard one that applies to all leafy greens — consistency matters more than avoidance, and a big new source of greens is worth mentioning to whoever manages the INR.

Honest verdict

The anticoagulant caution for self-heal is weak and theoretical. It is worth mentioning to someone on warfarin mainly because warfarin is sensitive to anything unquantified in the diet, not because self-heal has a demonstrated effect on clotting. Presenting it as an established interaction would be manufacturing a hazard, and we are not going to do it.

Pregnancy and Breastfeeding: Absent Data

The position here is genuinely empty, and the empty position deserves precise language rather than a reflex.

There is no human pregnancy data for self-heal and no animal reproductive toxicology we could verify. There is also no traditional signal in either direction — the plant is not a classical emmenagogue or abortifacient in European herbalism, and it does not carry the pregnancy contraindications that attach to herbs with a known uterine action.

The reasonable reading:

Note the epistemic state precisely: this is absent, not negative. Nothing has shown self-heal to be harmful in pregnancy. Nothing has shown it to be safe either. Those are different, and conflating them either way is sloppy.

Thyroid, Blood Pressure and the Potassium Arithmetic

Two traditional claims generate two theoretical interaction worries, and both benefit from arithmetic.

The potassium-and-blood-pressure story does not survive the numbers

The proposed mechanism for self-heal's traditional blood-pressure use is often given as its potassium salt content, acting as a mild diuretic. Let us be deliberately generous to the herb and see whether that can work.

Potassium in dried leafy plant material typically runs somewhere in the region of one to three percent of dry weight. Take the top of that range, three percent, and assume — unrealistically — that all of it extracts into the water:

  1. A cup of tea from 2 g of dried herb: at most about 60 mg of potassium. A medium banana carries roughly 420 mg. Current intake recommendations sit in the region of 2,600 to 3,400 mg a day. So the cup contributes on the order of two percent of a day's potassium, or about a seventh of a banana.
  2. A full traditional Chinese decoction of 15 g: at most about 450 mg — roughly one banana.

Both figures are ceilings chosen to flatter the herb, since real extraction is partial. And the conclusion is unambiguous: the potassium content cannot be the mechanism. A pharmacological diuretic shifts potassium by far more than one banana's worth, and no plausible cup of tea delivers a diuretic potassium load. If self-heal affects blood pressure at all, potassium is not how.

The same arithmetic is quietly good news for safety. Someone on an ACE inhibitor, an angiotensin receptor blocker, spironolactone or with chronic kidney disease is genuinely advised to watch potassium — and one banana's worth from a decoction is not zero, but it is not the thing that will hurt them either. Proportion matters in both directions.

The thyroid question

Xia ku cao's flagship traditional indication is dissipating nodules, including goitre, and there is Chinese-language research interest in the plant and thyroid function. We could not verify a controlled human study, and we are not going to imply one exists.

What is proportionate:

The Immune-Modulation Caution

Self-heal extracts alter immune-cell behaviour in laboratory models, and this gets converted into a caution about immunosuppressant drugs and autoimmune disease. It deserves to be stated at its real strength, which is low.

Essentially every plant extract ever put in front of immune cells "modulates" them in some direction. The finding is so nearly universal that it carries very little information about any individual herb, and the direction of effect frequently reverses between models, cell types and concentrations. As the basis for a clinical caution it is thin.

The reason to mention it anyway is asymmetry of consequence, not strength of evidence. For a transplant recipient, or someone on a biologic for rheumatoid arthritis or inflammatory bowel disease, the downside of an unexpected immune effect is graft rejection or a disease flare — severe, and hard to attribute afterwards. When the potential harm is that serious and the product is unstandardised, caution is reasonable even on weak evidence. That is a defensible argument, and it is a different argument from claiming the herb is known to interfere with immunosuppressants. It is not.

Wild Harvest, Lawns and Misidentification

This is, in our view, the single most defensible practical caution about self-heal — and it has nothing to do with the plant's chemistry.

Self-heal thrives in mown grass. That is its ecological niche: it is a low creeping perennial that outlasts taller competitors precisely because it survives being cut. Which means the places it grows most abundantly are lawns, playing fields, park grass, verges and path edges — and those are exactly the places that receive amenity herbicides, dog urine, road runoff and foot traffic. Gathering a wild green from a mown lawn is a contamination question before it is ever a botany question.

Then the botany. In a lawn, before or after flowering, self-heal can be confused with several unrelated plants that grow in the same conditions — bugle (Ajuga reptans), ground ivy (Glechoma hederacea), the Stachys woundworts, and various small square-stemmed mints. In flower, the blunt club-shaped violet spike is distinctive and confusion is unlikely. Out of flower, a low rosette of opposite leaves is not distinctive at all.

And then the commercial version of the same problem. As the name page sets out, "woundwort", "heal-all" and "self-heal" have all been applied to several unrelated species over the centuries. A reassuring safety record belongs to Prunella vulgaris. It does not automatically belong to a bag of loose dried herb sold under a folk name, and if you cannot establish which species you have, the hazard is worse than the species-specific record suggests, not better.

Practical implications, and these are the only ones on this page we would state firmly:

Rosmarinic Acid: What Happens to It in the Body

Since rosmarinic acid is the compound most often invoked to explain what self-heal does, its fate in the body matters — and here at least there is real human data, borrowed from work on other plants.

The picture, consistently reported across studies, is that rosmarinic acid has low oral bioavailability. It is extensively metabolised — hydrolysed toward caffeic acid and conjugated to glucuronides and sulfates — both in the gut wall and the liver, and cleared quickly. Plasma concentrations of the intact compound after a realistic dietary or herbal dose are low, typically far below the concentrations at which the interesting in-vitro effects are demonstrated.

Three consequences follow, and they run through this whole Benefits section:

  1. A cell-culture concentration is not a plasma concentration. When a paper reports an effect at, say, 50 micromolar rosmarinic acid, the relevant question is whether swallowing a herb can produce anything remotely like that in blood. Generally it cannot.
  2. The topical and mucosal routes escape the problem. A gargle or a wash puts the compound on the tissue at brewed concentration without first-pass metabolism. This is a large part of why the traditional topical uses have a more coherent rationale than the modern oral supplement.
  3. And it is borrowed evidence regardless. Rosmarinic acid pharmacokinetics is rosmarinic acid pharmacokinetics. It is not self-heal pharmacokinetics, because the tea is a mixture and nobody has measured what the mixture does. Rosemary, lemon balm, perilla, sage and spearmint all carry the same compound — the point made on the Perilla allergy page and repeated here for consistency.

Numbers We Refuse to Give

Stated openly, because a confident figure invented to look authoritative is worse than an admitted gap — and because a reader is entitled to know which numbers we declined to make up.

  1. Milligrams of rosmarinic acid per cup. Content varies severalfold with provenance, harvest, drying and steeping. Published figures for this species were not consistent enough to quote.
  2. Any lethal or no-adverse-effect dose. We could not verify a toxicology programme, so any such figure would be fabricated.
  3. A safe maximum daily dose. None has been established for any preparation.
  4. A duration limit. There is no mechanistic reason for one and no data to set one. Inventing "no more than six weeks" to sound careful would be pretending to knowledge.
  5. The actual potassium content of the dried spike. The figures used above were deliberately generous ceilings from general leaf-tissue ranges, labelled as such, and used only to show that a mechanism cannot work — which is a legitimate use of a ceiling.
  6. Any pregnancy safety statement. No data exists in either direction.
  7. The magnitude of any INR shift. Never measured. Anyone quoting one is guessing.
  8. An equivalence between a tea and a concentrated extract. Without a standardised marker compound there is no honest conversion.

Practical Summary

Ranked by how confident we are, which is not the same as how alarming each item sounds:

  1. Firmest statement on the page: know what plant you have and where it grew. Identification and contamination are the real risks with this herb, and they are entirely under your control.
  2. Very likely fine: eating the young leaves as a green, and drinking the tea, in the amounts people have used for centuries. Long, widespread, cross-cultural food record with no signal.
  3. Reasonable and low-risk: a cooled infusion as a gargle, mouth rinse, or wash on intact or minimally broken skin. Make it fresh; a home infusion left standing is a growth medium.
  4. Unknown, so treated conservatively: concentrated extracts, tinctures and capsules taken daily for long periods. Not implicated in anything — simply unstudied, and not covered by the food record.
  5. Worth mentioning to a clinician, at genuinely low strength: use alongside warfarin or other anticoagulants (mainly a consistency issue, and the vitamin K direction is the more plausible one), alongside immunosuppressants or biologics (weak evidence, serious downside), and if you have thyroid disease — where the sensible move is simply to have your thyroid function checked.
  6. Avoid: medicinal doses in pregnancy and breastfeeding, on absent data rather than any known harm.
  7. The one real hazard of this herb: using it instead of getting assessed. A neck lump, a severe or one-sided sore throat, a spreading skin infection, or anything in an immunocompromised person needs care, not a gargle. The plant will not hurt you. The delay can.

Key Research Papers

All references are live PubMed searches by title or topic. Where we could not verify specific metadata we describe the finding and link the topic instead, and where we could not verify a study at all we say so rather than citing it.

  1. Bai Y, Xia B, Xie W, et al. "Phytochemistry and pharmacological activities of the genus Prunella." Food Chemistry, 2016. The most complete compositional overview, and the reference against which we checked the "hazards it does not have" list. — PubMed search
  2. Wang SJ, Wang XH, Dai YY, et al. "Prunella vulgaris: a comprehensive review of chemical constituents, pharmacological effects and clinical applications." Current Pharmaceutical Design, 2019. Note how little of it concerns safety data — that absence is itself the finding. — PubMed search
  3. Petersen M, Simmonds MSJ. "Rosmarinic acid." Phytochemistry, 2003. Establishes rosmarinic acid as a widely distributed plant compound, which is why any rosmarinic acid finding is borrowed evidence for this plant. — PubMed search
  4. Rosmarinic acid pharmacokinetics, metabolism and bioavailability in humans. The low-bioavailability and rapid-conjugation picture described above. — PubMed search
  5. Rosmarinic acid and platelet aggregation in vitro. The real basis for the antiplatelet mechanism, and worth reading for the concentrations used. — PubMed search
  6. Coumarin toxicology, and the distinction from 4-hydroxycoumarin anticoagulants. The literature that makes the "contains coumarins therefore thins blood" argument untenable. — PubMed search
  7. Vitamin K content of leafy greens and warfarin management. The interaction that is actually plausible, and it runs opposite to the popular claim. — PubMed search
  8. Germander (Teucrium chamaedrys) hepatotoxicity. The Lamiaceae counterexample that kills the "it is just a mint" argument. — PubMed search
  9. Pyrrolizidine alkaloid distribution in plant families. The basis for clearing self-heal of the comfrey and coltsfoot concern. — PubMed search
  10. Perilla ketone and pulmonary toxicity. A genuine mint-family hazard, and one self-heal does not carry. — PubMed search
  11. Herbal product adulteration and species substitution, including DNA barcoding surveys. Why a folk name is not an identification and why the woundwort problem is practical rather than academic. — PubMed search
  12. Heavy metals and pesticide residues in wild-harvested and roadside plants. The contamination question that a lawn-dwelling plant raises before any pharmacology does. — PubMed search
  13. Prunella vulgaris in ethnobotanical wild-food surveys. The evidential basis for the food-history argument on this page. — PubMed search
  14. Fang X, Chang RCC, Yuen WH, Zee SY. "Immune modulatory effects of Prunella vulgaris L." International Journal of Molecular Medicine, 2005. The source of the immune-modulation caution, and useful for judging how much weight it can carry. — PubMed search
  15. Live search for toxicology and safety studies of Prunella vulgaris. Run this to check our claim that the toxicology record is thin. If that changes, this link will show it before we do. — PubMed search

For drug-interaction and supplement-safety background more generally, the US National Center for Complementary and Integrative Health maintains accessible summaries at nccih.nih.gov, and the National Library of Medicine's liver-injury resource is at LiverTox.

Connections

Back to Table of Contents