Ashwagandha: effects, dose, liver, and what it switched off in me

I took ashwagandha for three months straight, 500 mg per day of KSM-66 extract, without a break, and without knowing I needed one. The result came slowly. Total composure first, which looked like what the box promised. Then emotions almost absent, when I do not have many to begin with. Then no desire left to build anything for myself.

I stopped. The drive came back. Today I only take it in one specific stressful situation, for a few days, never continuously.

This is an observation, not proof, and I will treat it as such. But it made me reread the literature with a simple question: what is actually measured, over how long, and what is not measured at all?

What the trials measure

The trial everyone cites is Chandrasekhar's, in 2012: 64 adults under chronic stress, 300 mg of root extract twice a day or a placebo, for 60 days. Stress scores drop clearly, blood cortisol too, and adverse effects are mild and comparable in both groups. That is the KSM-66 extract, the one I was taking.

Lopresti, in 2019, gets a similar result with another extract, 240 mg once a day for 60 days in 60 stressed adults: measured anxiety drops, morning cortisol too, and a second stress hormone, DHEA-S, with it. The authors propose the mechanism: the extract moderates the axis that commands the stress response, between the brain and the adrenal glands.

Put end to end, these trials give a meta-analysis of 12 studies and 1,002 participants, published in 2022. The measured effect is large, on anxiety as on stress. But two figures on the same page put it in perspective: the studies disagree with each other at 94%, and the authors rate the certainty of the evidence as low. The favourable effect on stress sits between 300 and 600 mg per day.

On sleep, five trials and 400 participants give a small effect, clearer in people with insomnia, at 600 mg per day or more, and after eight weeks. The authors add a sentence I had not read at the time: data on serious adverse effects are limited, and more safety data would be needed to judge long-term use.

Eight weeks: the length of the trials, not of the intake

Look at the durations. 60 days in Chandrasekhar. 60 days in Lopresti. Eight weeks in Langade's sleep trial, in 40 healthy people and 40 with insomnia. Eight weeks in Verma's safety study: 80 healthy volunteers, 300 mg twice a day, full blood work including liver and thyroid, no abnormality. Its authors conclude that intake is safe over eight weeks, and call for long-term studies.

There are none. The window in which ashwagandha has been measured stops at two months. My three months without a break were beyond it, in a zone where nobody has measured anything, neither benefits nor risks.

The liver

This is the documented risk, and it is getting better documented. LiverTox, the reference database of the US National Institutes of Health on substance-related liver injury, rates ashwagandha in category B: a likely cause of clinically apparent liver injury. Its entry was updated in December 2024.

The first case series dates from 2020: five patients, three in Iceland, two in the US surveillance network. Mean age 43. All developed jaundice, with nausea, lethargy and itching, after two to twelve weeks of use. None went into liver failure, and blood tests normalised within one to five months. The products were analysed: ashwagandha, with no other toxic compound.

In 2023, an Indian series changed the scale. Of 23 patients with liver injury attributed to ashwagandha, eight were taking a single-ingredient product. Five already had chronic liver disease, three arrived in acute-on-chronic liver failure, and those three died. Chemical analysis of the products found only the plant's natural compounds, with no adulteration.

The 2026 review gathers 13 publications and 25 patients. The picture is reproducible: an injury that declares itself after weeks of use, mostly cholestatic, with jaundice and itching, recovery over weeks or months for most, one transplant, and three deaths in people who already had cirrhosis. The authors ask that any unexplained hepatitis prompt a check for supplements, and that people with liver disease avoid the plant.

Thyroid, hormones, pregnancy: the Anses opinion

Anses issued its opinion on 19 April 2024. Eight adverse effect reports have reached it since 2009, six of them analysed. It adds what the trials say: sedative effects, and effects on thyroid and sex hormones, with cases of the thyrotoxicosis type.

That last point is not theoretical. In 50 people with mild hypothyroidism, 600 mg per day for eight weeks changed TSH, T3 and T4 compared with placebo. The authors see a benefit. One can also read that a plant sold for stress moves thyroid hormones, which is exactly what you do not want when a thyroid is already treated or out of balance.

Hence the agency's list. It advises against ashwagandha for people with thyroid, liver or heart disease, hyperandrogenism, pregnant women, because of a traditional use as an abortifacient, and people on sedative or central nervous system depressant treatments. As a precaution, breastfeeding women and anyone under 18. And no driving after taking it. Denmark, on the basis of a 2020 assessment, considers that the plant should not be used in food, and the European Medicines Agency judged in 2013 that it could not write a monograph for lack of sufficient data.

What it switched off in me

The calm, I got. Too much of it. What I called composure was mostly an absence: little annoyance, little joy, little curiosity, and above all no desire to launch the projects I do for myself. I did not link it to the capsule for weeks, because nothing hurt. Stopping is what made it legible: the desire came back.

I looked for what the literature says about it. Nothing. A PubMed search crossing ashwagandha or Withania with anhedonia, emotional blunting or apathy returns no article, no trial, no published case. What exists are user reports, many of them, describing the same thing in the same words. Reports are not data. They are questions nobody has yet put to a trial.

Two mechanisms make the story plausible without proving it. The first is cellular: the root extract activates GABA receptors, the brain's main brakes, with a particular potency on one of their subtypes. That is what explains the drowsiness in trials. The second is hormonal: in Lopresti's trial, morning cortisol and DHEA-S drop more than under placebo. A brake on the receptors and a brake on the stress axis, taken every day for three months, can produce exactly what I felt. Can. I do not know whether that is it.

How I use it now

In practice: one specific stressful situation, a few days, then I stop. Never with alcohol or a sedative. A liver panel if I ever went back to several weeks. And if the desire to build drops, that is the stop signal, before anything else.

What this article does not say

• That ashwagandha does not work. Twelve trials find an effect on stress, and two of them measure the drop in cortisol. What is missing is certainty, not signal.
• That ashwagandha switches off emotions. No study measures it. I report an observation consistent with two documented mechanisms, nothing more.
• That the liver is at risk for everyone. Twenty-five patients described, most recovered within months. The serious risk concentrates on livers that are already diseased.
• That 500 mg is the right or the wrong dose. It sits within the range of the trials. What fell outside the frame was the duration, not the dose.
• That my stopping proves the cause. Drive returning after stopping a capsule, without alternation, remains compatible with other explanations. I have not found a better one.

« A plant that calms stress calms the rest along with it. The question is not whether it brakes, it is what it brakes when you never lift your foot off the pedal. »

What Helix does

Ashwagandha is a textbook case of what Helix tries to show on a stack line: a studied duration, eight weeks, that appears on no box; a rare risk concentrated on one precise population, the already diseased liver; and an adverse effect thousands of people describe without any trial having measured it.

The engine shows the maximum studied duration next to the dose, it flags the Anses contraindications at the same level as the effect you are after, and when an observation exists only in reports, mine included, it marks it as such.