Place · Level 3

Herb Safety · natural doesn't mean safe

有药理就有毒理 · 马兜铃酸留下 DNA 指纹 · 何首乌看你的基因 · 丹参会顶华法林 · 告诉医生你在吃什么

Story path

1Where 'natural = safe' comes fromWhere 'natural = safe' comes from
2Aristolochic acid · a fingerprint in DNAAristolochic acid · a fingerprint in DNA
3Why it hurt me and not themWhy it hurt me and not them
4Danshen × warfarin · two paths stackingDanshen × warfarin · two paths stacking
5The liver can't read the labelThe liver can't read the label
6What to do · red flags · honest closeWhat to do · red flags · honest close

Chapter 1

Where 'natural = safe' comes from

Natural and safe have never been the same thing.

To be clear up front: this island is not saying Chinese herbs don't work. Quite the opposite. Artemisinin came from sweet wormwood, and Tu Youyou won the 2015 Nobel Prize in Physiology or Medicine for it; aspirin's prototype came from willow bark; digoxin came from foxglove. There is real substance in herbs.

But precisely because they genuinely do pharmacology, they can genuinely do toxicology too. A molecule that can change something inside you can change the wrong thing. Where there's pharmacology, there's toxicology — that holds for pharmaceuticals, and it holds word for word for herbs.

So where does the natural = safe intuition grow from? Mainly three illusions:

The dose illusion. A tablet says 500 mg because someone measured it. An herb is a mixture of dozens to hundreds of compounds, and their content drifts with growing region, harvest season, and processing method. This isn't an abstract worry: someone bought 28 red yeast rice products and tested them — taking one day's labeled dose, the monacolin K inside ranged from 0.09 mg all the way to 10.94 mg (Cohen 2017). Incidentally, monacolin K's chemical structure is lovastatin — eating red yeast rice means taking a statin, you just don't know how much (EFSA 2018). For the detail, see the red yeast rice island.The purity illusion. You think you bought a pure plant. But herbal preparations have been found to contain heavy metals, and also pharmaceutical drugs that weren't on the label (Ernst 2002). One honest note: how common this actually is, the data can't tell us — most of it is scattered reports, not rigorous epidemiology.The regulation illusion. China does in fact have adverse-event monitoring for Chinese medicine — a national system was built in 1989, and roughly 10-15% of the reports it receives involve TCM drugs (Zhang 2012). But regulation watches the product. You taking it long-term on your own, stacking it with prescriptions, not telling your doctor — all of that falls outside its field of view.
So this island argues one thing only: an herb inside your body is doing what a drug does. If it's a drug, treat it like one.

Open this on the map

Chapter 2

Aristolochic acid · a fingerprint in DNA

One class of herbs leaves a fingerprint on your DNA. Decades later, a doctor can still recognize it inside a tumor and trace it back to that bowl of medicine.

These herbs contain aristolochic acid (AA), from plants of the genus Aristolochia. In 2012 the WHO's International Agency for Research on Cancer (IARC) classified both plants containing aristolochic acid and aristolochic acid itself as Group 1 carcinogens — the top tier, meaning the evidence of causing cancer in humans is conclusive, the same tier as aflatoxin and tobacco (IARC 2012).

How does it manage that? Follow it through:

1. Aristolochic acid goes in and is metabolically activated into a reactive molecule looking for something to bind.
2. It lunges at DNA and sticks onto adenine (the base written A), forming an adduct that won't leave. It can sit in the kidney for decades.
3. Why the kidney specifically? Because cells lining the kidney's proximal tubule carry a set of transporters (OAT1 and OAT3) that haul molecules like this out of the blood and into the cell. The more they haul, the more piles up (Han 2019).
4. When the cell divides it must copy its DNA, and the copying enzyme reaches that gummed-up A and misreads it, writing down a T instead.

That error has a very distinctive shape: A swapped for T. This A:T→T:A transversion is a pattern virtually no other carcinogen leaves behind. In upper-urinary-tract tumors from people with documented exposure, 72% of the mutations were exactly this kind (Hoang 2013). That's why it works as a fingerprint — unique, and it doesn't fade for decades.

Evidence · kidney is settled, liver is still argued

Kidney and upper urinary tract: causation is solid.

The Belgian patients underwent preventive surgery: of 39 people, 18 (46%) already had upper-tract urothelial carcinoma; every tissue sample analyzed contained aristolochic-acid-related DNA adducts (Nortier 2000).Taiwan has the highest upper-tract urothelial carcinoma incidence reported anywhere. A molecular-epidemiologic study of 151 patients found that among those carrying A:T→T:A mutations, 83% had aristolochic acid DNA adducts detectable in the renal cortex. The conclusion is blunt: aristolochic acid exposure contributes significantly to Taiwan's upper-tract urothelial carcinoma incidence (Chen 2012).
Liver cancer: here we have to be honest.

One study sequenced the exomes of 98 hepatocellular carcinomas (HCC) from Taiwan; 78% carried the aristolochic acid mutational signature. Widening to 1400 liver cancers from many regions, the figure for mainland China was 47% (Ng 2017). That number is striking.But a signature means exposure happened, not that it was the culprit. A review argues aristolochic acid's target organ may depend on age: young animals develop liver tumors, whereas exposed adults mainly develop urinary-tract tumors (Li 2024). So whether aristolochic acid directly drives these liver cancers is still being argued.The honest statement: for kidney and upper urinary tract, the evidence is tier A; for liver cancer, we can only say the association is strong and causation is unsettled.
The rules did move. The 2020 Chinese Pharmacopoeia removed aristolochic-acid-containing herbs, with Asarum the exception (Li 2024). But old prescriptions, old stock, and overseas products still circulate — so names like Guan Mu Tong, Guang Fang Ji, and Qing Mu Xiang are worth pausing on.

For how the kidney actually works, and why 'kidney-nourishing tea' is mostly a non-question, see the kidney island.

nortier-2000-nejm-aristolochia-urothelialchen-2012-pnas-aa-urothelial-taiwanng-2017-aa-liver-cancer-asia

The Fang Ji lesson · it wasn't counterfeit

Between 1990 and 92, a batch of Belgian weight-loss formulations had Stephania tetrandra (fen fang ji) replaced by Aristolochia fangchi (guang fang ji). Both names contain 'fang ji'. The result was a cluster of women progressing to kidney failure, and then upper-tract urothelial carcinoma found across the group (Nortier 2000; IARC 2012).

This story often gets told as 'counterfeits hurt people'. It wasn't.

Aristolochia fangchi is a real herb — it genuinely appeared in pharmacopoeias, with its own name and growing regions. The problem was never authenticity; it was that two herbs with similar names had wildly different toxicity. One got swapped for the other, and whoever swapped them may not have known what they'd swapped in.

This is exactly what makes 'the ones that cause trouble are all fakes' so dangerous: it lets you believe that buying genuine, buying expensive, or buying from an old established brand makes you safe. Aristolochic acid does not care what you paid. Genuine Aristolochia is a Group 1 carcinogen too (IARC 2012).

So the question to ask isn't 'is this real?' — it's 'what is actually in this?'

nortier-2000-nejm-aristolochia-urothelial

Open this on the map

Chapter 3

Why it hurt me and not them

The same herb: your uncle took it for ten years and was fine, you took it for three months and your liver gave out. This is not luck, and it is not you being frail. Often it's the shape of a lock you happen to carry.

Drug-induced liver injury has two temperaments. One tracks dose — take enough and it hurts, and it hurts anyone. The other is called idiosyncratic: it doesn't much track dose or duration, it tracks the person. Polygonum multiflorum belongs to the second kind.

Polygonum multiflorum (He Shou Wu) is commonly used for greying hair and as a liver-kidney tonic. The US National Institutes of Health drug-induced-liver-injury database rates its causal likelihood at grade A, meaning it is a well established cause of clinically apparent liver injury (LiverTox 2020).

Here's the lock. Your cell surfaces carry a kind of display rack (HLA), whose job is to hold up fragments from inside the cell so the immune system can glance at them: friend or enemy? Racks come in many shapes, and HLA-B*35:01 is one of them.

Some component of Polygonum multiflorum apparently slots into that particular shape. Once it's held up, the immune system doesn't recognize what it's looking at — so it attacks it along with the liver cell holding it up. The liver cell becomes collateral damage.

The numbers are hard: among 26 patients with Polygonum-induced liver injury, 88% carried at least one copy of HLA-B*35:01; among patients whose liver injury came from other drugs only 12% did, and among healthy controls, 5% (Li 2019). The association replicated in two further independent cohorts.

So 'why me?' has an answer: not a fake herb, not a weak constitution. Your immune system simply recognizes it.

Not a Chinese-herb problem — a herb problem

It's easy to read the previous page as 'Chinese medicine has a problem'. It doesn't. This is a property of herbs, and it has nothing to do with culture.

Take an example from an entirely different tradition: ashwagandha, the flagship Ayurvedic supplement, hugely popular in Europe and the US. The drug-induced liver injury networks of Iceland and the US reported 5 cases — cholestatic or mixed liver injury after 2 to 12 weeks of use (Björnsson 2020). India then reported a series of 23 cases, in which people who already had chronic liver disease had severe, even fatal, outcomes (Philips 2023).

Same script, different continent: a plant regarded as a gentle tonic, with real pharmacological activity, therefore with real toxicological risk, therefore colliding with certain people.

For how the whole adaptogen marketing frame got built, see the ashwagandha and adaptogens islands.

One practical note. After the previous page you may want to get tested for HLA-B*35:01. This is not a routine test today, and most places can't run it. So the action available to you isn't genotyping first — it's: don't take it long-term unmonitored, watch your symptoms, and check liver enzymes periodically. Latency can stretch from days to about 6 months — you cannot infer safety from 'I've taken it a while and I'm fine' (LiverTox 2020).

bjornsson-2020-ashwagandha-diliphilips-2023-ashwagandha-dili

Open this on the map

Chapter 4

Danshen × warfarin · two paths stacking

'Chinese and Western medicines don't interfere with each other' — with danshen and warfarin, that sentence is wrong enough to bleed.

Meet the pair. Warfarin is an anticoagulant with an unusually narrow therapeutic window: a bit too much and you bleed, a bit too little and you clot, which is why the dose is tuned by repeated blood draws. Danshen (Salvia miltiorrhiza) is widely used in China for cardiovascular conditions — and many people meet both precisely because something is wrong with their heart or vessels. That's not coincidence; it's a high-risk intersection.

They come at you down two separate paths, both pointing the same way.

Path one · pharmacokinetics (more drug). In animal studies, adding danshen raised warfarin's peak concentration and total exposure (AUC), slowed its clearance, and lengthened its half-life (Chan 2001). In plain terms: same dose going in, more drug staying inside. The number your doctor carefully tuned has quietly stopped being true.

Path two · pharmacodynamics (it thins blood itself). Danshen acts on blood directly: it holds platelets back from aggregating, interferes with the extrinsic coagulation pathway, has antithrombin-III-like activity, and pushes the fibrinolytic system to break down clots already formed (Chan 2001).

One path raises the drug level; the other thins blood on its own. Stacked, the result is easier bleeding. Several serious bleeding cases have been reported. The reviewer's conclusion leaves no room: people on warfarin should avoid danshen (Chan 2001).

The direction varies · that's the scary part

If herbs only ever thinned blood, this would be simple — remember 'don't stack' and you're done. The real trouble is that the direction varies.

All colliding with warfarin:

Danshen: pushes warfarin's effect up → bleeding risk (Chan 2001).American ginseng: a randomized controlled trial found it pulled warfarin's effect down (Yuan 2004). The opposite direction — and the risk on this end isn't bleeding, it's clotting.Goji: reported cases of markedly raised INR with bleeding (Rivera 2012).Natto: extremely high in vitamin K2, so it opposes warfarin. See the natto island for detail.
So the conclusion is not that herbs thin your blood. It's that some push, some pull, the magnitude is unpredictable, and you can't guess it. That's exactly why your doctor has to know — not so they can stop you, but so they know which way to adjust and how often to check.

Grade the evidence honestly. Most of this rests on case reports plus animal pharmacokinetics, not large randomized trials. That's enough to support the decision 'don't stack' — because the downside is bleeding or clotting, too costly to gamble on. It is not enough to support quantitative prediction: nobody can tell you how much danshen raises your INR by how much.

For systematic drug-nutrient interactions (metformin and B12, proton pump inhibitors, and so on), see the drug-nutrient interactions island; for the liver's phase-one metabolism and why grapefruit is dangerous, see the liver island.

yuan-2004-aim-ginseng-warfarinrivera-2012-goji-warfarin

Open this on the map

Chapter 5

The liver can't read the label

The word natural is printed for you to read. Your liver can't read it.

There is no department in your liver that sorts molecules by origin — synthesised in a lab, or simmered out of a root. It does one thing: it reads structure. And any molecule with pharmacological activity is, by definition, one that binds something inside you. They all go down the same line.

That line has two stages, and the danger sits between them.

Stage one (phase I) often makes a molecule MORE dangerous, not less. Cytochrome P450 enzymes bolt an oxygen onto it — the point is to give the next stage a handle to grab. But that step frequently produces an intermediate more reactive than the original: it carries a site desperate to react with something, and it is sitting inside a liver cell, surrounded by protein and DNA (Guengerich 2008).

Stage two (phase II) is the rescue. Conjugating enzymes weld glutathione (or similar) onto that reactive site, neutralising it and making the whole molecule water-soluble enough to leave in bile or urine (Jancova 2010).

So injury almost always happens the same way: stage one outruns stage two. Reactive intermediates appear faster than they can be quenched, and the surplus reacts with whatever is nearest — the liver cell itself. The molecule's origin has nothing to do with it. Paracetamol overdose (a Western drug) kills livers by this mechanism. Aristolochic acid (a herb), the previous scene's DNA signature, is the same mechanism: phase I activates it, and the activated form goes straight at DNA.

Natural buys nothing on this production line. If anything it cuts the other way: a plant has pharmacological effects precisely because it evolved molecules that interfere with animal physiology — that was the anti-pest strategy.

This was never a Chinese-versus-Western team sport. It's pharmacology versus a body. For how the liver actually handles these molecules, and where the word 'detox' stops being real, see the liver island.

Why the doctor can't see it either

This production line is also hard to see in the statistics — and the reason is mechanical too.

Drug-induced liver injury has no single confirming lab test. It's a diagnosis of exclusion: rule out viral hepatitis, alcohol, autoimmune disease and biliary obstruction, and only then does suspicion land on a drug. So the doctor has to know what you took before they can suspect it at all.

But many people don't say. In a US survey, only about one third of herb and supplement users had told their physician; the usual reasons were that the doctor never asked, and the patient didn't think it needed mentioning (Kennedy 2008). That's US data and doesn't transfer directly, but the psychology is universal: you don't think of it as a drug, so you don't think it needs declaring.

Which closes a loop: you believe herbs are harmless → you don't tell your doctor → the doctor never looks there → the cause is filed as 'unknown' → herbs' share in the statistics drops → they look safer.

For background: in a study spanning mainland China, the single largest identified class among liver-injury cases was traditional Chinese medicine plus herbal and dietary supplements (26.81%), with anti-tuberculosis drugs second (21.99%) (Shen 2019). That's a share of causes, not a rate of poisoning — in a population that uses Chinese medicine enormously widely the denominators differ, so it can't be read as a toxicity comparison. It settles exactly one thing, and that one thing is enough: if Chinese medicine were truly gentle on the liver, it could not be sitting at the top of that list.

Breaking the loop takes one extra sentence in the exam room. The next scene is that sentence.

shen-2019-dili-incidence-china

Open this on the map

Chapter 6

What to do · red flags · honest close

The single most useful move costs nothing and requires no pharmacology: let your doctor know what you're taking.

Five things you can do today:

1. Volunteer it; don't wait to be asked. At a visit, list the herbs, supplements, medicinal teas, and the formula a relative gave you — all of it. Your doctor probably won't ask (Kennedy 2008). Saying it is what gives them the chance to think in that direction.
2. Pause on any Aristolochia name. Guan Mu Tong, Guang Fang Ji, Qing Mu Xiang — these are confirmed sources of a Group 1 carcinogen (IARC 2012). The 2020 Pharmacopoeia removed aristolochic-acid-containing herbs (Asarum excepted), but old prescriptions, old stock, and overseas products still circulate (Li 2024). If unsure, ask the pharmacist: is there any Aristolochia in this?
3. Stop them around surgery. Herbs are generally advised to be stopped before an operation, since anesthesia, bleeding, and blood pressure can all get tangled with them. How long beforehand is a question for your anesthesiologist and surgeon — don't decide it yourself (Ang-Lee 2001).
4. Especially speak up if you take these: anticoagulants, glucose-lowering drugs, immunosuppressants, anti-rejection drugs, antiepileptics. What they share is a narrow therapeutic window — they can't absorb an unquantified variable (Chan 2001).
5. Don't take anything long-term unmonitored. Liver injury from things like Polygonum multiflorum can surface months in, with latency from days to about 6 months (LiverTox 2020). 'I've taken it a while and I'm fine' is not evidence of safety. If you're going long-term, someone should be checking your liver enzymes on a schedule.

Red flags · don't watch and wait, get liver enzymes checked now:

Yellowing of the skin or the whites of the eyesUrine turning dark, like strong teaPersistent fatigue, nausea, right upper abdominal pain
This is the classic presentation of drug-induced liver injury, and the order is often fatigue and nausea first, then dark urine and jaundice (LiverTox 2020). Among clinically apparent Polygonum-induced liver injury cases, about 10% ended fatally or required urgent liver transplant — that isn't scaremongering; it's why this needs to happen now rather than 'sometime' (LiverTox 2020). Stop the herb and see a doctor; don't give it a few more days.

The honest close · what this island claims

This island in one sentence: it isn't asking you to avoid Chinese medicine — it's asking you to treat it as medicine.

What does 'medicine' mean? It means there's a dose, there are interactions, there's individual variation, there are conditions for it to work, there are reasons to stop, and there are red flags. The caution you naturally bring to a pharmaceutical — reading the leaflet, asking the doctor, watching for adverse effects — carry it over to herbs unchanged, and you're done.

Why is it worth it? Because every item on this island is avoidable:

Aristolochic acid's cancers are avoided by reading a name (IARC 2012).The danshen-warfarin bleed is avoided by saying one sentence (Chan 2001).Polygonum's liver injury is caught early by checking liver enzymes and knowing the red flags (LiverTox 2020).
Not one of them asks you to give up a belief. They all just ask for one extra action.

Finally, the title's sentence, finished:

Natural doesn't mean safe. But natural doesn't mean dangerous either. It simply doesn't mean anything.

Origin can't certify safety — not willow bark, not a laboratory. Safety is earned one piece of evidence at a time: who took it, for how long, what happened, what it collided with. Artemisinin earned it, which is why it has saved millions. Aristolochic acid was investigated too, which is why it sits on the Group 1 carcinogen list. Both used the same ruler — and that ruler is evidence, not allegiance.

This page is education to help you understand the why; it is not medical advice. Whether to stop something, how to stop it, and what to test — ask your doctor.

Open this on the map