Ashwagandha: Evidence for Stress, Sleep and Safety

Ashwagandha (Withania somnifera) is used in traditional medicine and marketed for stress, sleep and other outcomes (NIH ODS). A 2025 conference abstract reports pooled stress, anxiety and cortisol changes; the sleep review found a pooled improvement across several measures, with a non-significant PSQI result. These findings do not establish a universal form, dose or independent-only benefit. One testosterone trial was manufacturer-funded, while a performance review disclosed author commercial ties. LiverTox documents clinically apparent liver injury, Denmark prohibits sale as food, and one AAN abstract reports a possible escitalopram interaction. Benefits, financing and harms are assessed separately below.

Table of contents

• Evidence summary
• What Ashwagandha is
• Forms and preparations
• How Ashwagandha works
• What works and what does not
• Benefits with evidence grades
• Risks and reported side effects
• Medication and condition cautions
• Who should avoid Ashwagandha
• Dosage and how to take Ashwagandha
• Infographics
• Related research
• Frequently asked questions
• Sources and funding notes

Evidence summary

What Ashwagandha is

Withania somnifera, also called ashwagandha, Indian ginseng or winter cherry, is a shrub in the Solanaceae family. Its roots and, in some products, leaves are used in preparations marketed for stress and sleep. Traditional Ayurvedic use provides historical context, rather than proof of efficacy or safety for every modern extract (NIH ODS; RIVM).

Ashwagandha contains withanolides and other constituents. Concentrations differ by plant part, extraction method and product. Labels may report total withanolides or withanolide glycosides; these measurements should not be treated as interchangeable potency or clinical-effect scales (RIVM; Lopresti 2019).

Stress and sleep have short-term human-trial findings, while broader cognition and immune-health claims remain insufficiently established. Manufacturer support and untraced trial funding limit an independent-only benefit verdict (NIH ODS; NCCIH).

Forms and preparations

Different products use different plant parts, extracts and analytical specifications. A trial result applies most directly to the preparation, population and protocol actually studied; equal milligrams or percentages do not establish equal clinical effects (NIH ODS; RIVM).

The dose distinction is illustrated by Lopresti 2019: 600 mg/day of finished Shoden tablets delivered 21 mg/day of withanolide glycosides. Multiplying that entire tablet mass by the concentrated extract’s 35% specification would give the wrong exposure. Neither this calculation nor a label percentage proves cross-product clinical equivalence.

How it works

Ashwagandha may influence the hypothalamic-pituitary-adrenal (HPA) axis, which governs cortisol responses. Several stress trials and the 2025 conference-abstract meta-analysis report lower cortisol, but the manufacturer-funded Lopresti 2019 trial in overweight men with fatigue found no significant salivary-cortisol difference. Lower cortisol alone does not demonstrate the mechanism or a clinical stress benefit (Bachour et al. 2025; Lopresti et al. 2019).

GABA-related effects are proposed from laboratory and animal research discussed in the sleep review. Those findings do not demonstrate benzodiazepine-like receptor action or a confirmed anti-anxiety mechanism in people.

The fertility pilot measured LH and testosterone, while Lopresti’s trial measured salivary testosterone and DHEA-S. Changes in these markers do not establish a reproductive mechanism or clinical fertility, libido or vitality benefit.

The sleep review compared different preparations and durations, rather than demonstrating an eight-week onset threshold or cross-brand bioequivalence. Its subgroup comparisons cannot establish one product’s absorption or potency relative to another.

What the “adaptogen” label does and does not show

NIH describes ashwagandha as commonly called an “adaptogen” (NIH ODS). That broad label is not an outcome measured in a clinical trial. It cannot substitute for evidence about a particular stress, sleep or other endpoint, at a specified preparation and dose.

Comparing stress-focused options

For context, magnesium and L-theanine are often compared with ashwagandha for stress and sleep. These brief comparisons do not establish that the ashwagandha benefit trials pass an independent-only funding filter.

Magnesium. Its evidence, dose and safety need a separate assessment. The ashwagandha trials reviewed here do not establish magnesium’s comparative efficacy or a head-to-head safety advantage.

L-theanine. This is a different ingredient with a separate evidence base. The ashwagandha trials reviewed here do not establish a faster clinical benefit or a cleaner comparative safety profile for it.

Choosing between these options requires the person’s symptoms, medicines and the evidence for each ingredient. No universal winner follows from the ashwagandha sources reviewed here.

What the withanolide-content variability actually means for a shopper

To compare a product with a trial, check the plant part, named extract, actual extract mass and analytical specification. A clinical study of a branded preparation does not verify every retail batch. FDA does not generally approve supplements before sale; any quality seal must be checked for the exact product and scope, and does not establish clinical efficacy (FDA; RIVM).

What works and what does not

Stress and cortisol: the best-supported claim

The Bachour 2025 conference abstract reports 15 trials and 873 participants. At eight weeks it reports a cortisol mean difference of −2.36 (95% CI −3.26 to −1.46; units not specified in the abstract) and a PSS difference of −4.88 (95% CI −7.84 to −1.91; p=0.0013). It does not provide a full trial list, dose assessment or financial audit. In the Chandrasekhar 2012 trial, 64 people were randomized and 61 entered the efficacy analysis; after 60 days, cortisol fell 27.9% versus 7.9% and PSS fell 44% versus 5.5% from baseline. Those percentages are within-group changes. Salve 2019 reported cortisol reductions at 250 and 600 mg/day, but a significant HAM-A result only at 600 mg/day. Both stress trials disclose Ixoreal-supplied extract.

Albalawi’s 2025 review abstract reports 488 participants across seven cortisol and six PSS studies: cortisol mean difference −1.16 µg/dL (95% CI −1.64 to −0.69), and PSS standardized mean difference −0.355 (95% CI −1.188 to 0.47; p=0.40). Full methods, funding and author conflicts remain unavailable here, so this is an abstract-level result and does not establish reliable individual symptom relief.

The Bachour publication states that abstracts were reviewed by the RCPsych Academic Faculty rather than the standard BJPsych Open peer-review process. Its complete publication is still a conference abstract. The Chandrasekhar and Salve trials were small, short Indian studies; the abstract alone does not identify all 15 included trials or establish their populations and finances.

Anxiety symptoms

The Bachour abstract reports an eight-week HAM-A difference of −3.52 (95% CI −6.00 to −1.04; p=0.0053). Chandrasekhar measured PSS, DASS and GHQ, not HAM-A or STAI; Salve did measure HAM-A and reported a significant benefit only for its 600 mg/day group. The 2022 WFSBP/CANMAT guideline provisionally recommends ashwagandha as monotherapy or adjunctive treatment for GAD, with 300–600 mg/day of a 5%-withanolide extract listed in its table. That is a provisional recommendation requiring more research, not a replacement for individualized clinical care. Guideline author Adrian Lopresti disclosed grants or honoraria from several extract/supplement businesses, including Ixoreal and Arjuna.

Sleep quality

The Cheah 2021 review included five trials with 400 participants. Pooled sleep measures favored ashwagandha (SMD −0.59, 95% CI −0.75 to −0.42; I²=62%), while the PSQI-only result was non-significant (SMD −0.90, 95% CI −2.13 to 0.33; p=0.15). Insomnia and healthy-participant subgroup estimates were −0.84 and −0.63. Estimates were also larger at ≥600 mg/day and ≥8 weeks, but these comparisons were across different studies and preparations; they do not prove a minimum, optimal dose or causal duration rule. Review authors declared no funding or conflicts. At least one included trial had manufacturer part-sponsorship; every underlying trial’s financial chain was not verified.

Testosterone, libido and fertility: small, preparation-specific studies

A 2013 pilot trial tested named KSM-66 root extract (≥5% withanolides), 675 mg/day for 12 weeks, in 46 men with oligospermia. In the active group, testosterone rose about 17%, sperm concentration 167%, semen volume 53% and motility 57% from baseline; these percentages are not placebo-relative treatment effects. The paper declares no conflicts but does not establish external financial funding, and it did not test pregnancy or live birth (Ambiye et al. 2013).

Lopresti 2019 enrolled 57 overweight men aged 40–70 with mild fatigue; 43 completed a 16-week crossover with eight weeks per treatment and no washout. Manufacturer Arjuna Natural funded the study and supplied Shoden. Salivary testosterone and DHEA-S were 14.7% and 18% higher relative to placebo, but cortisol, fatigue, vigor and aging-male-symptom outcomes were not significantly different. These biomarker findings do not demonstrate libido improvement, clinical androgen treatment or benefit in all men (NIH ODS).

These two trials address different populations: oligospermia in Ambiye and overweight older men with fatigue in Lopresti. Small samples, short exposures, sponsor support or unreported procurement terms limit generalization. They do not establish a general-population testosterone, fertility or libido benefit.

Strength and exercise performance: pooled evidence with commercial ties

Bonilla 2021 included 13 studies with 615 participants; 12 studies entered quantitative analyses. The strength/power analysis covered five studies and 198 people (effect size 0.68, 95% CI 0.40 to 0.95), while cardiorespiratory estimates varied from 1.37 with outliers to 1.85 after an outlier exclusion. Outcomes and populations varied; these were not all resistance-training trials. The paper disclosed that Bonilla was MTX Europe’s science product manager, a MET-Rx consultant and an AlzChem scientific affiliate; Kreider disclosed industry-sponsored ashwagandha research, consulting and an AlzChem advisory-board role. DBSS International funded the publication charge.

The review included varied training status, sexes, doses, durations and outcome definitions; its strength/power grouping also included testosterone measures. The pooled estimate is not a universal muscle-mass or resistance-training effect. Author commercial ties warrant caution, but do not prove every included trial had the same sponsor or invalidate its data. The financial chain of every included trial was not traced here.

Thyroid function: a genuine double-edged sword

The Sharma 2018 original abstract reports 50 people with subclinical hypothyroidism randomized to 600 mg/day or placebo; 46 completed eight weeks. TSH, T3 and T4 differences were significant (p<0.001, p=0.0031 and p=0.0096). The full methods, funding and conflicts were not accessible here. A short marker trial does not establish self-treatment, clinical symptom benefit or long-term safety.

NIH ODS and RIVM discuss thyrotoxicosis case reports. RIVM includes a published Dutch case after 500 mg/day for about a month, and a separate adrenal-suppression case reported from the UK. These are clinical reports, not two Netherlands pharmacovigilance notifications, and do not establish incidence or a dose threshold. Denmark’s food authority also cites metabolism and reproductive-system concerns in its current advice.

Changes in thyroid markers and thyrotoxicosis reports justify caution, without proving a predictable dose mechanism. NCCIH advises against use in people with thyroid disorders and warns of possible thyroid-medication interactions. People with thyroid disease should use clinician-led treatment rather than self-treat with ashwagandha.

Cognition and memory

NCCIH describes insufficient evidence for conclusions about cognitive function. The sources reviewed here do not establish a clinical memory treatment or a dependable benefit for healthy adults; this is distinct from claiming that every cognition trial has been audited or found commercially funded.

Immune function and general “adaptogen” claims

A broad “adaptogen” or immune-support label does not establish protection against infection or a clinical immune benefit. NCCIH finds insufficient evidence for several broader uses and advises caution with autoimmune conditions and immunosuppressants. RIVM treats laboratory and animal findings as safety/mechanistic context, rather than proof of human wellness benefit.

Benefits by claim

Risks and reported side effects

Safety evidence comes from short trials, clinical case reports and official reviews. Trial samples may miss uncommon harms; reports without a denominator cannot quantify population incidence (NIH ODS; LiverTox; TGA).

Reported mild adverse effects

• Gastrointestinal upset — nausea, diarrhea, vomiting, abdominal pain, loose stools
• Drowsiness or sedation; medicine combinations need individual advice
• Headache
• Dizziness
• Nasal congestion

Mild effects reported in the trials and NIH summary include gastrointestinal symptoms, drowsiness and some headache, dizziness or nasal symptoms. Their frequency varies by study and preparation; they are not a guarantee that every product is well tolerated (NIH ODS; Chandrasekhar; Salve).

Clinically important safety signals

Drug-induced liver injury. LiverTox assigns likelihood score B: a likely cause of clinically apparent liver injury. Typical reported onset is 2–12 weeks, often cholestatic or mixed injury with jaundice and itching. Many cases resolve after stopping, but recovery can be prolonged; the broader reference includes rare fatal or transplant-requiring cases, especially with pre-existing liver disease. The Björnsson 2020 series itself contained five cases with no deaths, liver failure or transplantation; four recovered and one was lost to follow-up. It reports NIH/NIDDK and NCI support and no author conflicts. The February 2024 TGA alert recorded 12 liver reports through 5 February, seven possibly caused by ashwagandha and four liver-related hospitalizations. RIVM 2024 separately described four Dutch Lareb liver notifications and reviewed nine published cases of different harms. These datasets may overlap and must not be added as a worldwide incidence count.

Thyrotoxicosis. NIH and RIVM discuss published thyroid-hormone excess cases, including a Dutch case after 500 mg/day for about one month. These are safety signals without a denominator or established dose threshold (NIH ODS; RIVM).

Adrenal suppression. RIVM summarizes a published UK case after approximately 860 mg/day for ten weeks, with recovery after stopping. It was not a Netherlands pharmacovigilance notification. One report does not establish frequency or prove an HPA-axis mechanism.

Serotonin syndrome (possible interaction reported in one conference abstract). An AAN 2025 abstract in Neurology describes a 22-year-old woman taking escitalopram 10 mg/day who developed myoclonus, eyelid flutter, tachycardia, fever, vomiting, dilated pupils and arrhythmia after a 600 mg ashwagandha pill and 1520 mg tea two days before admission. Reported treatment included stopping serotonergic agents, benzodiazepines and supportive care; the abstract does not state a complete recovery outcome. Authors reported nothing to disclose, without a separate funding statement. Neurology explicitly states that these meeting abstracts were not reviewed by its editors. This is a safety signal, not proof of causality, incidence or a confirmed interaction mechanism across antidepressants. (source)

Severe gastrointestinal reactions. The TGA’s February 2024 alert reports 16 gastrointestinal cases requiring hospitalization; some occurred after one or two doses. Other product ingredients were potential contributors in some cases. These reports cannot establish a population frequency.

Short trials cannot establish safety of daily use over years. NIH ODS identifies limited long-term safety information; RIVM found only two clinical studies lasting more than three months in its reviewed dataset. That is a limitation of the evidence assessed, rather than an exhaustive claim that no longer study can exist.

Medication and condition cautions

NCCIH also warns of possible interactions with anticonvulsants. The efficacy trials reviewed here are not comprehensive interaction studies. They cannot clear combinations with anticoagulants, statins, acid-suppressing drugs or other prescriptions; a pharmacist or clinician should assess the actual medicine list.

Who should avoid Ashwagandha

• Pregnancy. NIH, RIVM and BfR advise against use. Traditional abortifacient accounts do not prove a human effect, but adequate reproductive safety has not been established (NCCIH; RIVM; BfR).
• Breastfeeding. NIH and BfR advise against use because safety information is insufficient (NCCIH; BfR).
• Pre-existing liver disease. LiverTox and TGA advise avoidance; severe outcomes in the wider case literature are especially concerning in this group (LiverTox; TGA).
• Thyroid disorders, including hyperthyroidism. NCCIH advises against use; symptoms and medicines should be managed with a clinician (NCCIH).
• Thyroid medicines. NIH lists a possible interaction; discuss the actual medicine and thyroid condition with the prescriber (NCCIH).
• Autoimmune conditions or immunosuppressants. NIH cautions against use in autoimmune conditions and identifies possible medicine interactions (NCCIH).
• Sedatives or alcohol use. NIH and RIVM identify potential sedation-related combinations; seek advice before combining them (NCCIH; RIVM).
• People taking SSRIs/SNRIs or other serotonergic medicines should discuss ashwagandha with their prescriber. One 2025 conference abstract reports possible serotonin syndrome with escitalopram and high-dose ashwagandha; causality and risk frequency are unestablished. (source)
• Hormone-sensitive prostate cancer. NCCIH advises avoidance. Other cancers and hormone treatments need treatment-specific oncology advice (NCCIH).
• Scheduled surgery. Tell the surgical team about ashwagandha and follow its stopping instructions; NCCIH advises against use around surgery (NCCIH).
• Children. Germany’s BfR explicitly advises against use in children given the absence of pediatric safety data (BfR 2024).

Regulatory assessments and advice by jurisdiction

Different authorities serve different roles: risk assessment, consumer advice, food-sale restrictions and therapeutic-goods monitoring. Their publications can rely on overlapping literature. Agreement is relevant context, but is not proof of financially independent datasets or equivalent laws.

Denmark’s current food-authority page states that selling ashwagandha as food is illegal, including supplements and tea. It cites DTU’s concerns about reproduction/sex hormones, metabolism, immune effects and the central nervous system. This is the current food-sale advice; historical onset and comparisons with every other jurisdiction are separate questions.

Patwardhan et al. 2024 disputes the Danish assessment in an editorial. The authors explicitly report no funding for the work. Patwardhan’s Ayush research professorship is described as non-remunerative and honorary; Tillu reports an honorary World Ashwagandha Council association, and three authors report ashwagandha research involvement. These are relevant professional interests, but they do not establish payment or a financial incentive. The editorial’s arguments and the underlying safety evidence need separate assessment; the editorial alone does not invalidate the official advice.

RIVM’s January 2024 report reviewed nine published cases: seven liver injury, one thyrotoxicosis and one adrenal suppression, alongside four Dutch Lareb liver notifications. These were not all Dutch cases. It could not establish a health-based safe intake and advised against herbal preparations, especially during pregnancy. Its review also uses prior BfR and DTU assessments, so it is not a wholly separate safety dataset.

BfR’s September 2024 communication stresses children, pregnant/breastfeeding people and those with liver disease as groups that should avoid ashwagandha. Its introduction urges broader caution, and its later conclusion advises against consuming ashwagandha supplements. The evidence is insufficient to derive a safe intake. This is risk advice, not proof of a universal ban or of safety for every other adult.

The TGA’s February 2024 alert reports 12 liver notifications, seven possibly caused by ashwagandha; four of those seven had no other likely contributing ingredients. Four liver cases and all 16 reported severe gastrointestinal cases required hospitalization. These reports do not establish incidence. Separately, FSANZ’s August 2024 record lists root and root extract as non-traditional/novel food requiring a safety assessment before food sale. That record is general guidance and depends on product presentation and jurisdictional enforcement; it is distinct from therapeutic-goods regulation.

BfR’s 2024 communication describes the Annex III/Article 8 scrutiny framework and BfR’s earlier recommendation. A recommendation should not be described as a completed EU-wide restriction. In the UK, the FSA’s June 2026 response summary states that its call ran from 8 July to 2 September 2024 and received 19 responses, including businesses and trade groups. The material was supplied to the Committee on Toxicity; a final risk-assessment outcome and any advice were still to be published in that update.

ANSES’s February 2024 committee record recommends against use in pregnant/breastfeeding people, children under 18 and those with endocrine disorders, especially thyroid disorders or hyperandrogenism. This is the committee’s risk recommendation; it does not demonstrate that every national process used distinct data.

In the United States, the FDA does not generally approve dietary supplements before sale. However, qualifying new dietary ingredients require advance safety notifications, and FDA can inspect manufacturing, monitor adverse events and enforce adulteration or labeling rules. It is incorrect to say FDA can act only after injury. ODS and NCCIH publish evidence summaries; they are NIH research/information bodies, not food-sale regulators.

Debunking the commercial noise

The following examples show why broad marketing-style claims need qualification.

“Clinically proven testosterone booster.” The Ambiye pilot explicitly tested KSM-66, with no conflicts declared but external funding/supply terms unestablished. The Lopresti trial was Arjuna-funded and found salivary hormone changes without significant fatigue, vigor, aging-symptom or cortisol benefit. Neither result proves improved libido or androgen treatment for all men.

“It cannot affect the liver because it is natural.” LiverTox identifies clinically apparent liver injury and rare severe outcomes in its wider literature. The five-case Björnsson series had no deaths or transplantation. Clinical reports and TGA notifications are safety signals, rather than a population incidence estimate.

“Traditional use proves safety.” Historical use is relevant context, but does not prove safety of every preparation or long-term exposure. RIVM found that historical safety and exposure were inadequately documented and that reproductive concerns could not be resolved. Traditional abortifacient accounts are not demonstrated human effects.

“Every clinically studied milligram dose is interchangeable.” Trial matching requires the extract, plant part and actual extract dose. For example, Lopresti’s 600 mg finished-tablet dose delivered 21 mg glycosides; it was not 600 mg of 35% concentrated extract. Label percentages are not a validated clinical equivalence scale.

Dosage and how to take Ashwagandha

There is no single, universally validated ashwagandha dose. What follows are the dose ranges actually used in the human trials this article relies on — not manufacturer marketing claims.

• Stress and cortisol: One frequently cited 60-day trial tested 300 mg twice daily (600 mg/day total) of a standardized root extract (≥5% withanolides). The 2025 conference abstract does not establish an optimal dose (Chandrasekhar et al. 2012; Bachour et al. 2025).
• Anxiety: The WFSBP/CANMAT guideline provisionally lists 300–600 mg/day of a 5%-withanolide extract for GAD as monotherapy or adjunctive treatment. This is clinical guideline context, not a firm self-treatment instruction or proof all preparations are equivalent.
• Sleep: The Cheah review found larger pooled estimates in ≥600 mg/day and ≥8-week subgroups. Different trials and preparations underlie those comparisons, so they do not establish a minimum dose, optimal duration or causal dose-response.
• Concentrated extracts: Finished tablet/capsule mass and actual extract mass differ. In Lopresti, 600 mg/day of finished tablets provided 21 mg/day glycosides. Do not infer equal effects by converting different assay percentages.
• Root powder: NIH’s stress-trial summary includes a granule preparation equivalent to 6 g/day root powder. This is a studied protocol, not proof of an optimal traditional dose or equivalence to concentrated extracts (NIH ODS).

RIVM and BfR could not derive a safe intake from available evidence. That is an evidence gap, not permission to increase the dose. RIVM reports adverse cases at exposures overlapping marketed and studied ranges; a lower or “trial-like” dose cannot guarantee avoidance of liver or thyroid injury. FDA’s supplement framework also includes safety notification and enforcement powers (FDA).

Timing varied by study. Chandrasekhar used twice-daily dosing, while Lopresti used tablets two hours away from food, preferably after dinner. Neither directly compared morning with evening dosing. These protocols do not establish one superior time of day for stress or sleep.

Infographics with full text versions

Infographic 1: Evidence strength by claim

Infographic 2: Global regulatory status, 2020–2026

Related research

Stress and anxiety care extends beyond a supplement. For a broader review of evidence and care options, see the Pure City Research stress, anxiety and depression guide.

For broader sleep evidence and practical guidance, see the Pure City Research sleep guide.

For blood-pressure research and medicine-related considerations, see the Pure City Research blood-pressure guide. A possible interaction warning is not proof of a tested combination effect.

For broader blood-sugar evidence, see the Pure City Research diabetes guide. Discuss supplement combinations with the treating clinician rather than inferring clinical hypoglycemia from unrelated trials.

Frequently asked questions

Sources and funding notes

Cash funding, author relationships, supplied products and institutional mandates are different. The table records actual declarations and source-access limits. An unfunded review does not establish that every included trial is independent; official safety advice also does not certify each underlying study’s finances. Unknown procurement terms or funding remain unknown.

Role of non-human evidence

The reviews below discuss laboratory and animal evidence. Such findings do not establish a clinical wellness benefit in people, but can contribute to mechanism hypotheses and regulatory safety assessment. This is review-level context, rather than an independent assessment of every underlying animal paper (Cheah; RIVM).

Laboratory findings and clinical conclusions

Laboratory findings discussed in the cited reviews can inform hypotheses or safety concerns. They do not prove clinical cognition, immune protection, fertility or symptom benefit, and their financial independence cannot be inferred from the reviewing institution (RIVM; Cheah).