Ayurvedic herbs linked to liver injury, and what the reports show

Seven herbs on this site have published reports of liver injury in people who took them: Giloy (Tinospora cordifolia), ashwagandha, turmeric and curcumin supplements, bakuchi (Psoralea corylifolia), Garcinia cambogia (a relative of vrikshamla, Garcinia indica), oral aloe vera and senna. The NIH LiverTox database, which grades every substance it reviews from A (well established cause of liver injury) to E (unlikely cause), rates Tinospora and turmeric A, ashwagandha, Garcinia and aloe vera B, and senna D.

Two things have to be said together. Most people who take these herbs do not develop liver injury: LiverTox puts the frequency for Garcinia at under 1 in 10,000 users and gives no figure for the others, because nobody has one. And the injury, when it happens, can be serious: the published series include deaths and liver transplants, and in the Indian series these fell almost entirely on people who already had chronic liver disease. The reports below show association, assessed with causality tools such as RUCAM, not proof; many involved multi-ingredient products, other medicines or possible adulteration, and the authors say so.

The herbs, and what was reported

The two review papers that collect this literature for Indian herbs are Philips 2020 in the World Journal of Hepatology, and Philips and Theruvath 2024 in Medicine, which names Tinospora cordifolia, Withania somnifera, Curcuma longa and Psoralea corylifolia as the herbs with the clearest signal and Centella asiatica (Mandukparni), Garcinia cambogia, Cassia angustifolia and Morinda citrifolia as ones with an emerging profile. A systematic review of herb-induced liver injury across all traditions counted 936 published cases involving 79 herbs or products (Ballotin 2021).

Reports of poisoning rather than medicinal use

Three further plants on this site appear in the literature through poisonings, not through use at any medicinal dose. Children in western Uttar Pradesh who ate the beans of Kasamarda (Cassia occidentalis) developed a fatal hepatomyoencephalopathy; a case-control study found an odds ratio of 12.9 for bean eating (Vashishtha 2007; further cases in Chhapola 2018 and Getsuwan 2021). Infants who swallowed 10 to 30 mL of clove oil (Laung) developed severe liver and kidney injury within a day, which LiverTox rates C in overdose while finding no injury at ordinary doses. Neem seed oil (Neem) given to infants has caused a Reye-like syndrome with encephalopathy and liver involvement (Sinniah 1982; Lai 1990).

Herbs LiverTox has reviewed and rates unlikely

For balance: LiverTox has also reviewed several Ayurvedic herbs and rates them E, unlikely to cause clinically apparent liver injury, having found no convincing published case: Bacopa (Brahmi), fenugreek (Methi), Boswellia serrata (Shallaki), licorice (Yashtimadhu) and berberine, the main alkaloid of Daruharidra though not the plant itself. Tribulus (Gokshura) is rated E with an asterisk: it poisons grazing animals, but human injury is unverified.

How to read these reports

Association is not proof. A case report records that a person took a product and then developed liver injury, after other causes were looked for. Causality scores such as RUCAM grade how well the timing, the recovery on stopping, and the exclusion of other causes fit. In the Giloy study, two-thirds of patients scored “probable” and one-third “possible” (Kulkarni 2022); in the turmeric pathology series, eight of eleven were probable (Papke 2025). No trial could settle the question, and none will be run.

The product is often not just the herb. Four of the five ashwagandha patients in the Iceland and US series were taking other supplements, and in one case rhodiola was a possible cause alongside ashwagandha (Björnsson 2020). Most of the turmeric cases in the literature involved another medicine too (Lombardi 2021). The early Garcinia cases were multi-ingredient weight-loss products (LiverTox). In the Giloy study, two of six retrieved samples had mercury, arsenic or lead above prescribed limits and some contained industrial solvents (Kulkarni 2022). A 2026 Brazilian series found a product labelled Curcuma longa with negligible turmeric constituents and a weight-loss product with undeclared synthetic drugs (Góis 2026). The reverse also holds: in the Indian ashwagandha series and several of the US and Icelandic cases the product was tested and contained what the label said and nothing else (Philips 2023).

Species disputes get settled by testing. When the Giloy reports appeared, it was argued that patients had taken Tinospora crispa, a related species with its own published hepatitis cases (Langrand 2014; Huang 2019). The Mumbai group retrieved the plant material its patients had used and identified it as T. cordifolia by morphology, chromatography and DNA barcoding (Nagral 2023). This site’s page on why one herb has several botanical names explains why such disputes arise.

Existing liver disease changes the stakes. All four deaths in the Giloy study, the three deaths in the Indian ashwagandha series, and most of the fatal Tinospora cases that LiverTox describes were in people with chronic liver disease, several of whom did not know they had it (Kulkarni 2022; Philips 2023). LiverTox says plainly that ashwagandha should be avoided in cirrhosis or advanced chronic liver disease (LiverTox: Ashwagandha).

Some of it looks like autoimmunity. Giloy injury came with autoantibodies and biopsy features of autoimmune hepatitis, and the authors suggest it can unmask silent autoimmune liver disease (Nagral 2021). Seven of ten US turmeric patients carried HLA-B*35:01, an allele also linked to green tea extract and Garcinia injury, at an allele frequency of 0.45 against 0.06 in the population (Halegoua-DeMarzio 2023). An idiosyncratic, immune-mediated injury in a susceptible minority fits what is seen better than a poison that harms everyone who takes enough.

Multi-herb and heavy-metal products

Rasa shastra preparations deliberately contain processed metals and minerals, and the Ayurvedic Formulary of India sets no heavy-metal limits. What has been published about them is mostly about lead, and mostly not about the liver.

• Content. 14 of 70 Ayurvedic products bought in Boston shops contained lead, mercury or arsenic (Saper 2004). Of 193 products bought online, 20.7% contained a detectable metal, with no difference between US-made and Indian-made; rasa shastra products were more likely to (40.6% against 17.1%) and carried far higher mercury, and every metal-containing product exceeded at least one daily-intake standard (Saper 2008). In an Iowa poisoning cluster, lead was found in 65% of 252 product samples (Mikulski 2017).
• Lead poisoning. 12 adult cases across five US states, 2000 to 2003 (CDC 2004); 6 pregnant women in New York with blood lead of 16 to 64 µg/dL from products containing up to 2.4% lead (CDC 2012); 46 of 115 screened members of one Ayurveda community with elevated blood lead (Breeher 2015); and, across ten years of New York City poisoning investigations, higher blood lead in rasa shastra users than in users of other remedies (Hore 2022). Lead harms blood and the nervous system first; the anaemia in these patients was more severe than in occupational lead poisoning (Kales 2007).
• Kidney. Anuric kidney failure in a 2-year-old after chronic intake of a mercury-containing Ayurvedic medicine (Sathe 2013); lead poisoning from an Ayurvedic product in a patient with chronic kidney disease (Prakash 2009).
• Liver. Of 33 south Indian patients with severe liver injury attributed to Ayurvedic and herbal medicines, five of the 27 followed with a biopsy died, and arsenic and mercury in the products they had taken were significantly associated with death (Philips 2018). In 386 products retrieved from 91 patients at the same centre, mercury exceeded WHO limits in 34%, cadmium in 25%, arsenic in 21% and lead in 14%; 40% of patients had acute-on-chronic liver failure, with 39% mortality in that group (Philips 2026). Chronic arsenic from Ayurvedic preparations has been reported with non-cirrhotic portal hypertension, in an 11-year-old on eight preparations containing 5 to 248 mg/L arsenic (Khandpur 2008) and in a 1979 series where two patients had taken bhasmas (Datta 1979).

The pages here for Naga Bhasma (lead), Haratala Bhasma and Rasamanikya (arsenic), Tamra Bhasma (copper) and Arogyavardhini Vati (mercury and copper) report animal studies in which correctly prepared material showed no toxicity at the usual dose. Those studies test a preparation made by the method; the surveys above tested what was actually on sale, and the two are not the same question. In the Giloy study, the authors could not separate the herb from the metals and solvents some retrieved products also contained, and said so (Kulkarni 2022).

If you suspect liver injury

Stop the product and see a doctor promptly if you notice yellowing of the eyes or skin, dark urine, pale stools, itching, pain or tenderness under the right ribs, nausea, loss of appetite or tiredness out of proportion to anything else going on. In the published series the first signs came weeks to months after starting, so a product begun long ago is not cleared by the delay.

Take the packaging with you, and name every product you have taken: herbs, supplements, powders from a practitioner, teas, and anything bought online. Several of the cases above were diagnosed late because the patient did not think of a herbal product as a medicine (Bottenberg 2007). Most patients in these reports recovered once the product was stopped, and that is also the first thing the doctor will want to know you have done.

Anyone who already has liver disease, or a diagnosed autoimmune condition, has the strongest reason to discuss any of these herbs with their doctor before taking them, because the serious outcomes in the Indian series were concentrated in those patients.

What this page does not do

It does not say that any herb is safe, or that any herb is dangerous. It reports what has been published, with the number of patients and the authors’ own caveats, so that the strength of each link can be judged. It does not estimate how often injury happens, because the denominator, the number of people taking each herb, is unknown. It does not give doses or say what amount is safe; the reports do not support such a figure. It does not test products, so it cannot say whether a particular packet contains what its label says. And it is a reading of the literature on 26 October 2026, not medical advice for any one person: the FAQ and the section above say when to see a doctor, and the doctor, not this page, decides what to do next.