Ashwagandha's Evidence Problem: What the RCTs Show, What the Safety Signals Say, and Why the Two Conversations Rarely Happen Together

Ashwagandha is unusual in the supplement world. Unlike most adaptogens, which rest almost entirely on traditional use and animal data, it has a genuine body of randomized controlled trials behind it. Multiple meta-analyses, published in peer-reviewed journals, show measurable reductions in cortisol and anxiety scores in stressed adults. That evidence is real and deserves to be taken seriously. The problem is that the conversation about ashwagandha tends to stop at the efficacy signal, without getting to two questions that matter just as much: whether emotional blunting is a documented adverse effect pattern, and whether the liver injury case reports appearing in the medical literature represent a genuine hepatotoxicity signal. This episode works through all three threads in sequence — what the RCTs actually show and how to read them correctly, what the safety signals amount to, and how the two main standardized extracts compare. Start with the mechanism, because it shapes how to interpret the trial data. Ashwagandha's active compounds — a class of steroidal lactones called withanolides — are not cortisol antagonists. They do not bind cortisol receptors or block cortisol synthesis the way a pharmaceutical would. The proposed mechanism is considerably more indirect. Withanolides are believed to improve the feedback sensitivity of the hypothalamic-pituitary-adrenal axis, the HPA axis — the cascade that runs from the hypothalamus through the pituitary gland down to the adrenal cortex, where cortisol is produced. When the HPA axis is chronically dysregulated, as it tends to be in people under sustained psychological stress, the negative feedback loop that should throttle cortisol production becomes less effective. The evidence suggests that ashwagandha helps recalibrate that feedback loop, rather than pharmacologically suppressing cortisol from the outside. This distinction matters because it predicts who should respond and who probably will not. If cortisol is already well-regulated, there is little dysregulation to correct. Most of the trials that show the strongest effects enrolled people with elevated baseline cortisol — adults reporting chronic stress, not healthy adults with normal HPA function. That is a critical qualifier when reading any headline that says ashwagandha "lowers cortisol." Now to the trial data. The most recent comprehensive meta-analysis on cortisol, from 2024, covered seven RCTs and 488 participants. All studies used placebo controls, and the pooled analysis found statistically significant cortisol reductions in the ashwagandha arms compared to placebo. A separate meta-analysis — fifteen studies, 873 participants, databases searched through September 2024 — found significant anxiety reductions. These are not outlier findings. The direction of effect is consistent across trials. Two individual trials are worth examining in some detail. The Chandrasekhar 2012 study, indexed under PMID 23439798, used KSM-66 at 300 milligrams twice daily — 600 milligrams per day total — for 60 days in chronically stressed adults. It found significant reductions in both serum cortisol and perceived stress scores compared to placebo. The Salve 2019 study, published in PMC under accession PMC6979308, enrolled 58 completers over eight weeks and tested two doses: 250 milligrams per day and 600 milligrams per day of ashwagandha root extract. Both reduced cortisol compared to placebo — the lower dose reaching statistical significance at p less than 0.05, the higher dose at p less than 0.0001. Sleep quality and stress scores also improved. Across the clinical literature, cortisol reductions in treatment groups range from roughly 23 to 33 percent compared to baseline in stressed adults, which is a meaningful effect size for a dietary supplement. There is an important distinction that tends to get blurred in consumer communication. The National Center for Complementary and Integrative Health — NCCIH — has reviewed the evidence and describes findings as "unclear" specifically for anxiety disorders. That is not a contradiction of the meta-analysis data; it is a different question. The RCTs largely enrolled people with high perceived stress or mild to moderate anxiety, not people with diagnosed anxiety disorders according to clinical criteria. Demonstrating efficacy for elevated stress scores in a generally healthy adult population is a lower evidentiary bar than demonstrating efficacy for a clinical disorder. One of the source materials for this episode explicitly flags that the association with cortisol changes "is not a claim that ashwagandha directly lowers cortisol as a drug would." That framing is accurate and important. The other issue with generalizing from trial data is standardization. The RCTs that produced these results used specific, characterized extracts. KSM-66 and Sensoril are the two that appear repeatedly in the clinical literature, and they are meaningfully different products. KSM-66 is a full-spectrum root extract standardized to at least five percent withanolides. At 600 milligrams per day — the most commonly studied dose — it delivers approximately 30 milligrams of withanolides, which sits within the range used across the trials that showed effects. Sensoril uses both root and leaf material and is standardized to a higher withanolide concentration, at least ten percent, plus 32 percent oligosaccharides. The two extracts have different phytochemical profiles, and while both have RCT support, no single trial has compared them head-to-head in the same population under the same conditions. Their relative efficacy is genuinely unresolved. What can be said with confidence is that neither is comparable to a generic, non-standardized ashwagandha powder of unknown withanolide content. A capsule labeled simply "ashwagandha root" with no standardization specification cannot replicate what the trials measured. Now to the safety questions, which rarely get the same airtime as the efficacy data. The standard contraindications are relatively well documented: ashwagandha is not advised during pregnancy, given evidence of uterotonic effects; it requires caution in people with thyroid conditions because some evidence suggests it modulates thyroid hormone levels; autoimmune disease is a concern given its immunomodulatory activity; and sedative drug interactions are plausible given its reported effects on GABA-related pathways. These warnings appear consistently across clinical reviews. What does not appear consistently is substantive discussion of two more recently flagged signals. The first is hepatotoxicity. Liver injury case reports associated with ashwagandha have begun appearing in the medical literature. These are case reports, not controlled data, which means they demonstrate that the association exists but cannot establish incidence rates or confirm causation with certainty. What makes liver injury signals worth taking seriously even at the case report stage is that herb-induced liver injury tends to be underreported, can range from transient enzyme elevation to severe cholestatic injury, and the underlying mechanism — whether immune-mediated, dose-dependent, or related to contaminants in lower-quality preparations — is not established. The current case report literature does not provide enough data to quantify the risk, and long-term safety data beyond twelve weeks does not exist in the trial literature. The trials run eight to twelve weeks in duration. What happens at six months or two years of continuous use is an open empirical question. The second signal is emotional blunting. Users report a flattening of emotional response — not sedation exactly, but a reduction in emotional range that some find uncomfortable or functionally impairing. This pattern has surfaced enough in clinical anecdote to warrant discussion, but it occupies an awkward evidential position. It is not a documented adverse effect in the RCT literature in any systematic way — the trials were not designed to capture it with validated instruments, and most were too short and too small to detect subtle psychological adverse effects. That absence of documentation is not the same as an absence of effect. The trials showing cortisol reductions used populations of 50 to 200 people over two to three months, with outcome measures focused on efficacy endpoints. They were not adequately powered or designed to characterize the full adverse effect profile. The Nutritional Medicine Institute has noted ashwagandha's potential amphoteric action — the idea that withanolides can both mildly activate and block hormone receptors depending on circulating levels — which is at least a plausible pharmacological substrate for why some users experience emotional narrowing. But that is a mechanistic hypothesis, not a confirmed explanation. There is also a broader structural issue worth naming. The supplement industry, and supplement-adjacent wellness media, have a consistent incentive to amplify efficacy signals and minimize safety signals. The RCT data on cortisol and anxiety is genuinely positive enough that it requires no embellishment. But the framing of ashwagandha as straightforwardly safe because it is natural and traditional is not supported by the evidence base. Herbs with potent physiological effects — and the cortisol data indicates that ashwagandha has genuine physiological activity — can produce both intended and unintended consequences. The same mechanism that may help recalibrate an overactivated stress axis could plausibly, in some individuals or at some doses, produce effects that extend beyond what the user intended. One counterargument deserves honest acknowledgment: the liver injury case reports and the blunting reports could reflect confounding from poor-quality or adulterated products rather than a signal inherent to well-characterized ashwagandha extracts. The supplement market is not uniformly regulated, and products sold under the ashwagandha label vary enormously in actual content and purity. It is entirely possible that some adverse reports trace back to contaminated or mislabeled products rather than to the compound itself. That is a legitimate possibility, but it is not exculpatory evidence — it is a gap in the data. Attributing adverse effects to product quality problems rather than to the active compound requires the same standard of evidence as any other causal claim, and that evidence does not currently exist. Pulling this together: the RCT evidence for ashwagandha's effects on cortisol and perceived stress is more substantial than what exists for almost any other adaptogen. The findings are consistent across multiple studies and two meta-analyses, the effect sizes are meaningful for a dietary supplement, and the mechanism — HPA axis recalibration rather than pharmacological cortisol suppression — is coherent and supported by the pattern of results, which show the strongest effects in people with elevated baseline cortisol. The extracts with actual trial backing are KSM-66 and Sensoril, both standardized products with characterized withanolide content, and neither can be substituted with generic ashwagandha powder. The distinction between efficacy for high perceived stress in healthy adults and efficacy for diagnosed anxiety disorders is real and should not be collapsed. Against all of that, the safety picture is genuinely incomplete. Liver injury case reports constitute a signal that has not been adequately investigated. Emotional blunting is reported by enough users to be taken seriously but has not been captured by the clinical trial apparatus. Long-term data beyond twelve weeks is absent. The specific risk profile of KSM-66 versus Sensoril — including adverse effects, not just efficacy — has not been compared in a head-to-head trial. And the question of who should not use ashwagandha, beyond the established contraindications for pregnancy, thyroid conditions, and autoimmune disease, is not resolved by the current literature. What the evidence currently supports is this: ashwagandha has the strongest efficacy data in its category, acting through HPA axis recalibration rather than direct cortisol suppression, with effects that are most reliable in people with elevated baseline cortisol using standardized extracts at studied doses. It also has an undercharacterized safety profile, with hepatotoxicity and emotional blunting as signals that the trial literature was not designed to detect and has not resolved. Both of those things are true simultaneously, and treating them as separate conversations — one for the efficacy advocates, one for the safety skeptics — is precisely the problem this episode has tried to correct.