A Review of 12 Ashwagandha Trials Rated the Evidence Low in Certainty, and Reviews of Turmeric, Triphala and Brahmi Report Small Studies, Mixed Results and Varying Quality

A 2022 review in Phytotherapy Research pooled 12 randomised trials of ashwagandha with 1,002 adults and found lower anxiety and stress scores than on placebo. Its authors rated the certainty of the evidence low for both outcomes. A 2025 meta-analysis of seven trials on cortisol and six on perceived stress, 488 participants in all, found lower cortisol but no significant change in perceived stress.

This article reads the published trials and reviews on four herbs used in Ayurveda: ashwagandha, turmeric, triphala and brahmi. For each it reports how many people were studied, how long the study ran and what the authors say about quality. It then sets out what Indian law and the Ministry of Ayush require of the people who make these products. Nothing here is guidance on using any herb. Classical texts such as the Charaka Samhita are a separate subject, and one passage from it is quoted in our article on diabetes across generations.

Ashwagandha and stress: lower scores in the trials, disagreement in the reviews

Akhgarjand and colleagues searched PubMed, Scopus and Google Scholar up to December 2021 and included 12 randomised trials, with participants aged 25 to 48. Scores were lower on ashwagandha extract than on placebo, with standardised mean differences of 1.55 for anxiety and 1.75 for stress. (That measure expresses the gap between groups in standard deviations, so trials using different rating scales can be pooled.) The trials differed widely in the size of effect they found, with an I-squared statistic of 93.8 percent for anxiety and 83.1 percent for stress. The authors wrote that further high-quality studies are needed.

Albalawi’s 2025 meta-analysis in Nutrition and Health searched from January 2012 to February 2024. Cortisol was lower in the ashwagandha groups by 1.16 micrograms per decilitre (95 percent confidence interval 0.69 to 1.64). The change in perceived stress was not statistically significant (p = 0.40), and its confidence interval ran from a fall of 1.19 standard deviations to a rise of 0.47. Risk of bias was generally moderate, and some trials lacked allocation concealment, the step that keeps the people enrolling patients from knowing which group comes next. Albalawi wrote that the gap between the cortisol and perceived-stress results calls for longer trials and broader groups of participants.

A third summary comes from the US National Institutes of Health’s Office of Dietary Supplements, in a fact sheet updated on 2 May 2025. It describes a 2021 review of seven studies with 491 adults, all in India, who took ashwagandha or placebo for 6 to 8 weeks, and reports lower stress and anxiety scores. Three of the seven tested one branded root extract, and two tested branded extracts of root and leaf together. The fact sheet also describes a 12-week trial of 120 overweight or mildly obese adults with low energy, in which the extract did not reduce perceived stress but did reduce fatigue.

Sleep: five trials, all run in India

Cheah and colleagues, in an unfunded PLoS One review from 2021, searched databases that included the Clinical Trial Registry of India and the Ayush Research Portal. They included five randomised trials with 400 participants, according to the abstract. Sleep measures were better on ashwagandha extract than on placebo, with a standardised mean difference of 0.59 (95 percent confidence interval 0.42 to 0.75, I-squared 62 percent), and the authors graded that evidence moderate. The effect was larger in adults diagnosed with insomnia and where treatment lasted at least eight weeks.

The review lists its own limits. All five trials were run in India, and one declared partial sponsorship by the manufacturer. Only two had been registered in advance, and in one of those the dose given was much lower than in the registered protocol. The authors wrote that data on serious adverse effects are limited and that more safety data would be needed before judging long-term use.

Liver injury reports and Denmark’s ban

In 2020 Björnsson and colleagues described five people with liver injury attributed to ashwagandha supplements: three in Iceland in 2017 and 2018, and two in the US Drug-Induced Liver Injury Network in 2016. Jaundice, nausea, itching and abdominal discomfort appeared after 2 to 12 weeks. The pattern was cholestatic or mixed (cholestatic means impaired bile flow). No patient developed liver failure, and liver tests returned to normal within one to five months in four of them; the fifth was lost to follow-up. Four were also taking other supplements, and in one case the herb rhodiola was a possible cause alongside ashwagandha.

Germany’s Federal Institute for Risk Assessment (BfR) issued a communication on 10 September 2024. It says human studies of ashwagandha mostly looked for benefits and did not systematically record adverse effects, and that active constituents vary widely between supplements, depending on the plant part and extraction method. The claimed positive effects, it states, “have not been scientifically proven.” Case reports of liver damage since 2017, including acute liver failure, give reason for caution, the BfR says, though the data are too thin for a conclusive risk assessment or a safe intake value. It advised that children, pregnant and breastfeeding women, and people with acute or existing liver disease should not take ashwagandha supplements.

The NIH fact sheet says Denmark banned ashwagandha in 2023, after a 2020 risk assessment by the Technical University of Denmark that discussed a possible risk of miscarriage and possible effects on thyroid and sex hormones. Critics dispute that report. Patwardhan and four co-authors wrote in the Journal of Ayurveda and Integrative Medicine in 2024 that it does not appear to be peer-reviewed, cites a few articles from predatory journals, and draws on studies of leaves and other plant parts although its brief concerned the root. They also wrote that no conclusive causal link has been established between ashwagandha and the sporadic liver, thyroid and gut reports. The NIH fact sheet says the herb appears well tolerated for up to about three months and that long-term safety is not known.

Turmeric: 16 knee trials, and imaging that did not change

Most trials of turmeric extract or its compound curcumin involve knee osteoarthritis, the wearing down of joint cartilage. Wang and colleagues pooled 16 randomised trials in 2021, with 1,810 adults and trials lasting up to 16 weeks. Eleven compared the extract with placebo and five with anti-inflammatory painkillers, and overall risk of bias was moderate. Against placebo, knee pain scores were lower (standardised mean difference 0.82, I-squared 86 percent) and physical function scores were better (0.75, I-squared 90 percent). Against the painkillers the effects were similar. The review reported no significant between-group differences in blood markers or imaging, and 12 percent fewer adverse events than with the painkillers, at rates similar to placebo.

A 2020 trial in Tasmania by several of the same authors enrolled 70 people with knee osteoarthritis and ultrasound-detected fluid and swelling in the joint. After 12 weeks the turmeric group’s pain score was 9.1 millimetres lower on a visual analogue scale than the placebo group’s (95 percent confidence interval 0.4 to 17.8). Fluid volume on MRI did not change, and neither did a measure of cartilage composition. The authors listed a modest sample and short duration as limits, called for larger multicentre trials, and named the University of Tasmania and Natural Remedies Private Limited as primary funders.

Shi and colleagues graded the reviews themselves in a 2026 overview in Frontiers in Medicine. They rated 10 systematic reviews and meta-analyses of curcumin in osteoarthritis with the AMSTAR-2 checklist, and only three came out as high quality. The overview reported improvements in pain and joint function, with efficacy signals comparable to painkillers, but also differences between the formulations tested, no head-to-head comparisons of them and incomplete safety reporting. In the authors’ words these preclude definitive conclusions.

Triphala and brahmi: small trials, and a cholesterol result no better than placebo

Triphala is a mix of three plants. Phimarn and colleagues reviewed 12 randomised trials with 749 patients in 2021, looking at blood fats, blood sugar and body measures. Six trials were done in India, and the review calls most of them small, with two enrolling more than 100 patients. Triphala groups had lower LDL cholesterol, total cholesterol and triglycerides in six studies, and fasting glucose fell in people with diabetes but not in people without it. The authors wrote that triphala might improve these measures “under certain conditions” and that large, well-designed trials are needed.

One placebo-controlled trial tested triphala together with guggulu (Commiphora mukul), so it cannot isolate triphala. Donato and colleagues gave the combination three times daily for three months to 90 adults with high cholesterol. Total cholesterol fell 3.3 percent on the combination and 1.9 percent on placebo, while LDL cholesterol fell 4.8 percent and 4.9 percent, with no difference between groups. Two people on the combination developed a rash, against none on placebo.

Brahmi is Bacopa monnieri. Kongkeaw and colleagues in 2014 included nine placebo-controlled trials of standardised extract lasting at least 12 weeks, with 518 people, and judged overall risk of bias low. Among the 437 people in the meta-analysis, a timed trail-making test and choice reaction time both improved, and the authors described a potential to improve cognition, particularly speed of attention. They added that only a large head-to-head trial against an existing medicine would give definitive data.

Basheer and colleagues looked in 2022 at trials in people with Alzheimer’s dementia or mild cognitive impairment and found five. Three tested Bacopa in combination with other herbs, and two compared it with the drug donepezil. All five were judged at high risk of bias and the evidence was rated very low in certainty. Each trial reported a statistically significant difference in at least one outcome, yet the reviewers concluded there was no difference between Bacopa and placebo or donepezil, and that no major safety issues were reported.

What Indian law asks of an Ayurvedic maker

The Drugs and Cosmetics Act of 1940 gives Ayurvedic, Siddha and Unani drugs their own chapter, IVA. Section 3 defines such a drug as a medicine made exclusively according to formulae in the authoritative texts listed in the Act’s First Schedule. Section 33E deems one misbranded if its label, container or accompanying material makes a false or misleading claim, and section 33EEC bars manufacture for sale without a licence, from a date each state notifies. Under section 33EED the Central Government may ban a drug by notification if it is satisfied that the drug is likely to involve a risk to people or animals, or that it “does not have the therapeutic value claimed or purported to be claimed for it.” That text is from the Ministry of Health’s compilation of the Act, which may not include the latest amendments. India’s regulation of conventional medicines and its generic-drug industry is told in how India became the world’s pharmacy.

The Ministry of Ayush also runs a pharmacovigilance programme, which collects reports of suspected adverse reactions and of misleading advertisements. A Lok Sabha written reply of 7 August 2026 says it has run since 2018 through a national centre, five intermediary centres and 97 peripheral centres, with an online portal, Ayush Suraksha, added in 2025. The reply lists complaints about misleading advertisements for Ayush drugs of 7,771 in 2023, 10,161 in 2024 and 16,129 in 2025, a total of 34,061.

On 15 April 2026 the Ministry wrote to all Ayush drug manufacturers that only ashwagandha roots may be used, in any form, and that leaves are strictly prohibited in Ayush drugs and products. The letter recalls an October 2021 letter advising against leaves and cites a 2024 safety dossier from a Ministry expert committee that recommended roots. It says studies reported possible safety concerns for leaves because of higher concentrations of reactive withanolides, particularly withaferin-A, and it tells makers to follow Rule 161 of the Drugs Rules, 1945, which requires the plant parts used to be named on the label. Copies went to state licensing authorities and to the chief executive of the Food Safety and Standards Authority of India (FSSAI), the food regulator.

LiveLaw reported on 19 May 2026 that Justice S.R. Krishna Kumar of the Karnataka High Court had stayed an FSSAI advisory of 16 April 2026 on leaves in health supplements and nutraceuticals, and the Ayush directive of 15 April 2026, insofar as they apply to the petitioners in Sami-Sabinsa Group Ltd and others v Union of India. The petitioners say they have used root and leaf extracts for about 30 years, that a food-safety advisory cannot prohibit leaves without amending the 2016 regulations, and that the studies cited do not sufficiently establish a safety concern. The court allowed the authorities to seek to vacate or modify the stay and listed the next hearing for 8 June 2026. We found no published report of a later order.