Phytotherapy and evidence: what is established for herbal medicines

You are standing in a pharmacy. In front of you are two boxes, both showing the same plant. One says “Zul.-Nr.” in small print, the German marketing authorization number. The other says “Reg.-Nr.”, a registration number, together with a sentence about long-standing use. In the drugstore around the corner there is a capsule with the same name, labeled as a food supplement.

Herbal sounds gentle. Safe. Like something you can simply try. Many people know this pattern, and many are surprised that a herb from the pharmacy shelf can contribute to a transplant rejection.

I want to look with you at what is established for herbal medicines. Where guidelines say yes, where maybe and where no. And how you can tell whether a pack stands for a tested benefit, a long tradition or a food.

Animal and cell data play hardly any role here, because this is about people, preparations and guidelines. In the examples, a label shows the level on which a statement stands: Supported by meta-analyses and guidelines Mixed or disputed Mechanistically plausible, human studies thin Clinical tradition without a strong study base What I observe clinically

What phytotherapy is, and what the number on the pack tells you

“Isn't that just herbal tea and homeopathic globules?” The question is understandable. In everyday life, much of what the law separates neatly becomes blurred.

According to Section 4 (29) of the German Medicines Act (AMG), herbal medicinal products are medicines that contain “as active substances exclusively one or more herbal substances or one or more herbal preparations”. Phytopharmaceuticals is simply another word for them. In pharmacy, a herbal drug is not a narcotic but the dried plant part. An extract is made when constituents are dissolved out of this herbal drug with a solvent. The term rational phytotherapy is used when such medicines are measured against the same scientific standards as other medicines.

A distinction in one sentence: In Germany, homeopathic medicinal products can be registered without submitting information on effects and indications (§ 38 (2) AMG), and anthroposophic medicinal products are developed according to the anthroposophic understanding of the human being and nature (§ 4 (33) AMG); both are separate therapeutic approaches that this article is not about. My view of anthroposophy is set out in anthroposophic medicine put into context.

An image for it: the plant is the raw material, the extract is the medicine. As with grapes and wine, processing decides what ends up in the bottle.

Three routes onto the shelf

“Both packs say St John's wort. Where is the difference?” It lies in what a manufacturer had to show before the product could be sold.

The mandatory statement on traditional medicines is remarkably honest. According to Section 10 (4a) AMG, it must state that the product is “registered exclusively on the basis of long-standing use for the indication”. For food supplements, the German Federal Office of Consumer Protection and Food Safety (BVL) states that it carries out “no assessment, review or approval”. Companies are themselves responsible for safety, and checks are carried out afterwards by the federal states. When food supplements can make sense and when they tend to get in the way, I describe in When food supplements make sense.

Commission E and the HMPC

Commission E was founded in 1978 as an expert panel and published around 380 monographs on herbal drugs until its legal mandate to produce monographs ended after 1994. It assessed the preparations mainly on the basis of bibliographic data, as a paper by Keller already described in 1991. A positive Commission E monograph is therefore a valuable historical document, but not proof of efficacy in the sense of today's randomized trials.

Today the Committee on Herbal Medicinal Products, the HMPC at the European Medicines Agency, sets the standards. It was established in 2004 and had adopted around 120 EU monographs by early 2014. These visibly separate “well-established use” from “traditional use”, perhaps the most useful distinction of all.

Herbal does not mean non-prescription, and it is rarely covered by insurance

The German Ordinance on Prescription-Only Medicines (AMVV), for example, lists “St John's wort, for the treatment of moderate depression” and kava-kava as prescription-only. And statutory health insurance may cover non-prescription medicines only by way of exception, when they are considered the standard of care for serious diseases. This restriction does not apply to children up to their 12th birthday or to adolescents with developmental disorders up to their 18th birthday (§ 34(1) SGB V). The exception list of the Federal Joint Committee (G-BA) names, among others, a Ginkgo biloba leaf extract for this, “standardized 240 mg daily dose”, “only for the treatment of dementia”, which the S3 guideline on dementia confirms. Voluntary extra benefits offered by individual insurers are not a general rule.

And now you know why two green boxes with the same name can legally be two different things.

Why extracts are not interchangeable

“St John's wort is St John's wort, isn't it?” No. And that may be the most important idea in this article.

Imagine two cups of coffee from the same bean. One is a strong espresso, the other a thin filter coffee. Same name, different drink. With plant extracts it is similar, except that you cannot taste the difference.

Interactions also depend on the preparation. St John's wort can activate a receptor called PXR, which can ramp up breakdown enzymes such as CYP3A4 and the transport pump P-glycoprotein, and the extent is significantly related to the hyperforin content (Nicolussi 2020). The German National Disease Management Guideline describes over-the-counter St John's wort preparations as “very variable” in composition.

Four more examples

Lavender. The German S3 guideline on anxiety disorders discusses a specific lavender oil extract and states explicitly: “For other lavender oil extracts […] no statements can be made.”

Pelargonium. A Cochrane review found effects in acute bronchitis with the liquid preparation, but not with tablets, and refrained from pooling some results because the type of preparation created too much heterogeneity (Timmer 2013).

Saw palmetto. The Cochrane review across all extracts is negative (Franco 2023). The 2026 guideline of the European Association of Urology, by contrast, notes that only one hexane-derived extract is recognized by the HMPC as “well-established use”, points out, regarding an older Cochrane meta-analysis of 30 trials, that different brands were included in it regardless of this extract, and recommends it weakly, with the strong advice that its efficacy may only be “modest”. This rests, among other things, on a meta-analysis of this extract alone, which mixed randomized trials with observational studies and had a co-author from the manufacturer (Vela-Navarrete 2018).

Ginkgo and echinacea. Ginkgo products from the US market contained between 24 and 36 percent flavone glycosides and ginkgolic acids ranging from below 500 to about 90000 ppm, and many did not meet the German pharmacopoeial specification, so they could not be considered pharmaceutically equivalent (Kressmann 2002, US market, no statement about today's German medicines). And the Cochrane review on echinacea stresses that products sold under the name echinacea differ “appreciably in their composition” (Karsch-Völk 2014).

For some plants, absorption in the gut adds a further layer, described in detail in Curcumin and the question of absorption.

And now you know why I so often write “a specific extract” instead of just the plant name.

Herbal medicines with solid evidence, and where it ends

“Are there any herbal remedies at all that demonstrably do something?” Yes, some. There are fewer than the shelves suggest, and each has a narrow indication. The table shows nine well-known examples; I discuss the first four here and the rest in the next section.

St John's wort for mild to moderate depression

First, three sentences that matter to me. Depression is a serious illness, not a passing low mood. St John's wort does not replace psychiatric or psychotherapeutic treatment and is no reason to postpone it. And anyone taking an antidepressant does not switch to St John's wort on their own, because the combination can become dangerous and stopping abruptly carries its own risks.

A systematic review around Apaydin with 35 studies confirmed, with moderate quality of evidence, more response than with placebo, but stressed the lack of research on severe depression and that rare harms were not captured (Apaydin 2016). A meta-analysis around Ng pointed out that all comparative trials with SSRIs lasted only 4 to 12 weeks and that it remains unclear whether people with severe depression or high suicidality benefit (Ng 2017).

What the saffron studies show is covered in Saffron for low mood, and how I look at depression as a whole in Depression seen as a whole.

Peppermint oil for irritable bowel syndrome

If you have irritable bowel syndrome, you know the feeling that your gut has its own weather. Before irritable bowel syndrome is even considered, other causes need to be ruled out by a doctor, and red flags such as blood in the stool, unintended weight loss, fever or symptoms at night need prompt assessment. After that, of all the herbal remedies in this article, peppermint oil has the strongest recommendation in a German guideline.

The German S3 guideline on irritable bowel syndrome, 2021 version, states: “Peppermint oil has proven effective for treating mainly the IBS symptoms ‘pain’ and ‘bloating’ and shall be considered.” Recommendation grade A, strong consensus. Enteric-coated capsules could be used effectively in the short term; data on long-term use were lacking. The basis includes a meta-analysis around Khanna with nine trials and 726 patients: peppermint oil was superior to placebo for global improvement with a relative risk of 2.23 and for abdominal pain with 2.14; side effects were mild, most commonly heartburn (Khanna 2014).

The guideline says “shall be considered”, not “shall be given”. And irritable bowel syndrome has more than one lever, as I describe in Irritable bowel syndrome: finding causes, Applying FODMAP correctly and gut-directed hypnotherapy.

A lavender oil extract for anxiety

Anxiety can take over the whole body. Anxiety disorders are serious illnesses and belong in specialist medical and psychotherapeutic treatment. Lavender shows how differently experts can read the same data.

A meta-analysis around Dold, financially supported by the manufacturer, pooled all five placebo-controlled trials with 1213 adults, including people with subthreshold anxiety, generalized anxiety disorder and mixed anxiety and depression. The extract was superior to placebo, the ratio of responder rates was 1.34, and there was no significant difference in side effects (Dold 2023). That does not make the result wrong, but independent replications are missing.

A ginkgo extract for dementia

Hardly any plant is so automatically linked with “good for memory” as ginkgo. Here, treatment and prevention tell two different stories.

The German S3 guideline on dementia, 2023 version, puts it weakly: “We suggest” using a ginkgo extract named in the guideline “at a dose of 240 mg daily to treat cognition and everyday functioning in mild to moderate Alzheimer's dementia or vascular dementia with non-psychotic behavioral symptoms”. The dose is a guideline statement, not a recommendation from me. The 2026 Cochrane review around Wieland with 82 studies sees possibly small to moderate benefits in dementia after six months, with low certainty and very high heterogeneity. In mild cognitive impairment, ginkgo probably has little or no effect.

And now you know why “herbal and proven” almost never applies to a whole plant, but to one preparation, one diagnosis and one group of people.

Mixed or negative: pelargonium, echinacea, valerian, saw palmetto, black cohosh

“And what about the classics for colds, sleep problems and menopause?” These are exactly the remedies found in many households. And it is precisely for them that the study situation is often more sobering than the advertising sounds.

Colds and cough: pelargonium and echinacea

The German general practice guideline from DEGAM on cough, version 3.2, describes mostly positive results for individual phytotherapeutics, “which, however, are of questionable clinical relevance”, and considers certain phytopharmaceuticals “worth considering” when the patient wishes treatment. On pelargonium, it notes that a warning about possible liver injury has been in the product information since 2014, that the benefit-harm ratio “cannot currently be conclusively” assessed, and that an increase in the blood-thinning effect of vitamin K antagonists cannot be ruled out.

For echinacea, the picture is more reserved. A Cochrane review around Karsch-Völk with 24 double-blind trials found no significant difference in any of the twelve prevention comparisons, and of seven treatment trials on the duration of colds, only one showed a significant effect. Conclusion: no demonstrated benefit for treatment; for prevention, possibly a weak benefit of individual products of questionable clinical relevance.

Sleep: valerian and other phytopharmaceuticals

The German S3 guideline on insomnia, 2025 update, states: “Treatment of insomnia with phytopharmaceuticals should not be recommended”, recommendation grade B, and an umbrella review found no proof of efficacy for valerian in insomnia, at most signs of subjectively better sleep quality without confirmation by objective measurements (Valente 2024). How phytotherapy experts criticize this assessment is described further below, and you will find more in Valerian and passionflower, a study check and herbal sleep aids put into context. If you take sleeping pills: the note on stopping from the box at the beginning applies here just the same.

Prostate: saw palmetto

The Cochrane review around Franco with 27 studies and 4656 participants concludes with high certainty that saw palmetto alone provides little or no benefit for symptoms caused by benign prostatic enlargement, after three to six months as well as after 12 to 17 months. Ten studies were industry-funded.

Prostate symptoms belong in urological assessment, immediately if there is blood in the urine or you can no longer pass urine, and an existing treatment is not swapped for a herbal preparation on your own.

Menopause: black cohosh

The German S3 guideline on peri- and postmenopause, 2020 version: “Cognitive behavioral therapy (CBT), isoflavones and Cimicifuga preparations can be used for vasomotor symptoms.” Recommendation grade 0, together with the note in the background text to inform women that there are many preparations with unclear safety and that interactions have been reported. A Cochrane review around Leach with 16 studies and 2027 women found no significant difference from placebo in the frequency of hot flashes, which could be analyzed in three studies with 393 women, and saw “insufficient evidence”. The not yet adopted draft of the 2026 revision keeps a weak “can” recommendation with low to very low certainty and states for mood swings and depression that Cimicifuga “should not be recommended because of the limited evidence”, grade B. More in Menopause put into context and Chasteberry put into context.

Incidentally, the same S3 guideline that strongly recommends peppermint oil for irritable bowel syndrome also reports negative findings: turmeric with no significant difference from placebo, ginger alone without positive effects, St John's wort without a positive effect and, in diarrhea, even worse than placebo. Where turmeric can be interesting is covered in Turmeric and inflammation.

And now you know why “herbal” is not a class of effect: the same plant can be well studied for one complaint and perform worse than placebo for another.

Herbal does not mean harmless: interactions, pregnancy, liver

“What could be dangerous about tea and tablets made from plants?” Herbal remedies often go unmentioned in conversations with doctors, probably because many people do not consider them medicines. That is exactly where a risk can arise.

The EU monograph lists concomitant use of the extracts with well-established use and of traditional preparations with more than 1 mg hyperforin per day as contraindicated with, among others, coumarin-type anticoagulants, ciclosporin, everolimus, sirolimus and systemic tacrolimus, HIV protease inhibitors and nucleoside reverse transcriptase inhibitors, as well as irinotecan, imatinib and other cytostatics broken down via these routes. How serious this is was shown as early as 2000 by a report around Ruschitzka in the Lancet: in two people after heart transplantation who were taking ciclosporin and St John's wort, acute rejection occurred, which the authors attributed to a metabolic interaction. Two cases are not statistics, but they are a clear warning sign.

The pill, antidepressants, anticoagulants and operations

The pill. In a study around Hall, 7 of 12 women had breakthrough bleeding during the St John's wort phase, compared with 2 of 12 in the control phase (Hall 2003). In a second study around Murphy with 16 women, hormone exposure from the pill fell by 13 to 15 percent, with signs of follicle growth and probable ovulation (Murphy 2005). A systematic review of four studies saw this as a reason for concern about contraceptive reliability (Berry-Bibee 2016), and the monograph recommends additional contraceptive measures. Please do not stop the pill or a prescribed medicine on your own; discuss the combination with your doctor or pharmacist.

Antidepressants. According to the monograph, St John's wort together with serotonin reuptake inhibitors or buspirone can contribute to serotonergic effects, very rarely up to serotonin syndrome. Such a combination never belongs in self-management.

Anticoagulants. With vitamin K antagonists, the window is narrow: deviations can lead to bleeding or thrombosis, and herbs and over-the-counter remedies can shift the benefit-risk balance (Di Minno 2017). For standardized ginkgo extract, a meta-analysis of 18 studies found no abnormal coagulation values (Kellermann 2011). However, laboratory values were measured, not bleeding events, and not specifically in people taking anticoagulants. The combination should therefore still be discussed with a doctor.

Operations. A review in JAMA named echinacea, ephedra, garlic, ginkgo, ginseng, kava, St John's wort and valerian as remedies that can cause problems around an operation, from a tendency to bleed to an increase in the sedative effect of anesthetics, and called for herbal remedies to be asked about explicitly before every operation (Ang-Lee 2001). So mention them before every planned procedure, even if nobody asks.

Pregnancy and breastfeeding. For St John's wort, the monograph states: “Safety during pregnancy and breast-feeding has not been established.” It adds that animal studies have shown signs of reproductive toxicity and states: “The use is not recommended during pregnancy and lactation.” Use during pregnancy and breastfeeding is therefore explicitly not recommended. If you are already taking St John's wort and are pregnant or breastfeeding, please discuss the next steps promptly with a doctor rather than deciding alone. Any use of herbal remedies during this time should be discussed with a doctor.

The liver: rare, but real

The liver processes almost everything we swallow. With herbal remedies it is worth taking a close look, without panic and without playing things down.

Kava is potentially hepatotoxic according to a clinical review around Teschke, based among other things on a positive rechallenge test, with overdose, long-term use and concomitant medication as risk factors (Teschke 2010). In Germany, kava is prescription-only.

For pelargonium the link is disputed: an analysis around Teschke found no case with highly probable or probable causality among 15 spontaneous reports (Teschke 2012), and other experts disagreed in a letter to the editor. In the US Drug-Induced Liver Injury Network, the share of cases due to herbal remedies and dietary supplements rose from 7 to 20 percent over the study period. The reference is all liver injuries recorded in the network, not all people who take such products (Navarro 2014).

What the studies on milk thistle show is covered in Milk thistle and the liver. What matters for the liver in everyday life I describe in what matters for the liver.

And now you know why the friendly green box belongs on your medication list.

Quality: adulteration, heavy metals, pyrrolizidine alkaloids

“How do I know that what is on the label is inside?” The following findings mainly concern food supplements, imports and products bought online, not the medicines authorized or registered in Germany. Still, they belong in an honest assessment.

Adulteration. An analysis around Ichim collected DNA-based authentication tests of 5957 commercial herbal products from 37 countries. According to this analysis, 27 percent were adulterated, meaning they contained undeclared admixtures, substitute plants, fillers or not the stated plant at all (Ichim 2019). The data mix product categories and countries and cannot be transferred to German pharmacy products.

Heavy metals. A group around Saper analyzed 193 Ayurvedic preparations ordered online. 20.7 percent contained lead, mercury or arsenic, 75 percent of the contaminated products advertised good manufacturing practice, and all contaminated products exceeded at least one standard for tolerable daily intake (Saper 2008). A quality promise on a website is therefore not a test result. More in Heavy metals in the body and Lead exposure: symptoms and sources.

Pyrrolizidine alkaloids are plant toxins found in many plant species that are suspected of being genotoxic and carcinogenic. In a European market survey around Mulder with 1105 samples from 2014 and 2015, at least one of them was found in 91 percent of teas and herbal teas and in 60 percent of food supplements (Mulder 2018). In 2017, the European Food Safety Authority saw a possible concern mainly for people who frequently drink large amounts of tea and herbal tea, and, for food supplements made from plants that produce such substances, a possible intake that could be too close to doses with severe acute toxicity. I deliberately do not convert this into cups, and I have not separately checked the current level of contamination.

What you can look out for

Authorized and registered herbal medicines go through an official procedure, whereas food supplements are not checked by any authority before sale. That is not a verdict on individual products, but a difference in safeguards.

• Look at the number. Zul.-Nr. or Reg.-Nr. show that a medicinal product procedure has taken place.
• Prefer traceable sources such as a pharmacy over unclear imports from the internet.
• Look for a precise description of the preparation. If the type of extract and the solvent are missing, the basis for matching it to studies is missing.

I deliberately do not recommend quality seals or brands. What to make of detox programs and detox teas is covered in Detox programs, a critical look.

Why there are fewer large studies, and how guidelines include phytotherapy

“If plants have been used for centuries, why is so little known?” There are reasons that can be documented, and one that is often cited but that I could not substantiate in my research.

First: legally it is often not necessary. Traditional medicines are registered without clinical trials, an authorization can be granted on a bibliographic basis, and Commission E assessed mainly on a bibliographic basis. Anyone allowed to sell on these terms has, at least legally, little reason for an expensive study.

Second: every study applies to only one extract. Anyone who pools ten valerian studies with ten preparations is mixing unlike things. The DEGAM guideline calls such meta-analyses “methodologically difficult”, and the S3 guideline on irritable bowel syndrome considers pooling “not meaningful”. An early valerian meta-analysis of 16 studies described considerable methodological problems, very different preparations and signs of publication bias (Bent 2006).

Third: many studies do not report what exactly was tested. 406 randomized trials of herbal remedies reported on average only 38 percent of the items recommended in a checklist (Gagnier 2011), even though the CONSORT reporting standard had been extended for herbal trials back in 2006, with the most extensive addition, of all places, in the description of the intervention (Gagnier 2006).

Fourth: large studies often come from manufacturers. This applies to the large lavender trial and the lavender meta-analysis, to ten studies in the Cochrane review on saw palmetto and to the meta-analysis on the hexane extract. That is not proof of a wrong result: the manufacturer-funded GuidAge trial on ginkgo was negative and was published. It only means that independent replications are especially valuable.

How guidelines assess phytotherapy

A systematic overview around Klose screened 128 German S3 guidelines. Phytotherapy appeared in 40 of them. Of 204 recommendations, 139, or 68.1 percent, were negative or open, side effects and interactions were mentioned in 70 of the 204 recommendations, and fewer than half of the guidelines included phytotherapy in their literature search at all (Klose 2014). When reading, pay attention to the small words: in German guidelines “soll” (shall), “sollte” (should) and “kann” (can) are fixed grades of strength.

An image for it: guidelines are like a mountain guide who only recommends paths he has walked several times himself. That does not mean there are no other paths. He just does not vouch for them.

Reading the evidence yourself: five questions for every pack and every study

“How am I supposed to tell the difference in everyday life?” You do not need any knowledge of statistics, just a handful of questions you can ask yourself, your pharmacy or your practice.

You can also remember two warning signals. When small studies show clearly better results than large ones, as with St John's wort at 1.87 compared with 1.28, caution about the effect size is warranted. And when a review suspects publication bias, as with pelargonium, unfavorable studies could be missing from the analysis.

Four levels on which a statement can stand

• Supported by large RCTs and meta-analyses
  Certain St John's wort extracts for mild to moderate depression and peppermint oil in the short term for irritable bowel syndrome, each with limits. Equally well supported, but negative: saw palmetto for prostate symptoms and ginkgo for preventing dementia.
• Mechanistically plausible, human studies thin
  The assumption that low-hyperforin St John's wort preparations cause hardly any interactions: the mechanism fits, but the human data come from one study with 20 healthy men over two weeks.
• Clinical tradition without a strong study base
  Traditionally registered medicines: long use, plausible efficacy, no clinical trial required. An honest category, but not a study finding.
• What I observe clinically
  That herbal remedies are often missing from medication lists until you ask specifically. Experience from my consultations, not a study result.

Where to go next in the Medicinal Plants category

This page puts things into context and points onward. What the term adaptogens means is covered in Adaptogens: what the term means. What is known about frankincense I put into context in Frankincense for joints and gut. Intravenous administration of plant substances is a separate topic, described in Medicinal plants by infusion. How plant substances and the microbiome might interact is covered in Medicinal plants and the microbiome, and herbal remedies for small intestinal bacterial overgrowth I put into context in Oregano oil and berberine in SIBO.

And now you know why the best question before any herbal remedy is not “Is it natural?”, but “What exactly is this, and what is known about it?”.

Frequently asked questions about phytotherapy and herbal medicines

Continue reading from here

This pillar page points to individual articles in which single plants, organs and questions are discussed in detail. Twelve paths that lead on sensibly from here.