Psilocybin for depression: what 20 years of research shows, and what changed in 2026

Introduction: how a prohibited compound reached the NEJM

Twenty years ago a clinical trial of psilocybin was nearly impossible to run — not for scientific reasons but regulatory ones. Today psilocybin-assisted therapy appears in The New England Journal of Medicine, JAMA and Nature. Johns Hopkins, Imperial College London, NYU Langone and the international COMPASS Pathways consortium have run dozens of trials in depression, addiction and existential distress.

This is one of those topics where enthusiasm runs ahead of the data. Popular reviews tend to list positive results one after another, and the picture comes out close to triumphant. But three papers published in 2025–2026 change that picture substantially, and none of them appear in the popular round-ups.

The key point: the evidence base for psilocybin is neither "strong" nor "weak" — it is inconsistent. The more rigorous the methodology, the more modest the effect. Below, indication by indication with levels of evidence, plus what an endocrinologist is obliged to exclude in "treatment-resistant" depression before a patient goes looking for experimental answers.

Depression: where the evidence is strongest and where it cracked

The modern era begins with an open-label pilot in treatment-resistant depression (Carhart-Harris RL, et al. Lancet Psychiatry 2016;3:619–627. DOI 10.1016/S2215-0366(16)30065-7, level 2b — uncontrolled open-label, n=12). All participants received two supported sessions; most improved, and in some the effect held for three months. The study established feasibility and safety — not efficacy: without a control arm there is no way to separate the drug from the effect of intensive attention.

A randomised trial in major depressive disorder followed (Davis AK, et al. JAMA Psychiatry 2021;78:481–489. DOI 10.1001/jamapsychiatry.2020.3285, level 1b, n=27). More than half the participants reached remission by week four. The important caveat: the control was a waiting list — a group receiving nothing at all. That design systematically inflates the apparent effect of any active intervention.

The first attempt at a fair comparison was the escitalopram trial (Carhart-Harris R, et al. N Engl J Med 2021;384:1402–1411. DOI 10.1056/NEJMoa2032994, level 1b, n=59). Here is what popular summaries mention only in passing: on the primary endpoint there was no difference between psilocybin and a standard antidepressant. Most secondary measures — remission rate, social functioning — favoured psilocybin, but secondary endpoints do not substitute for the primary one; they generate hypotheses rather than prove effects.

The largest trial of its time came from COMPASS Pathways (Goodwin GM, et al. N Engl J Med 2022;387:1637–1648. DOI 10.1056/NEJMoa2206443, level 1b, n=233). A 25 mg dose significantly outperformed the 1 mg comparator by week three, while 10 mg showed no advantage. A single 25 mg dose was then confirmed against active placebo in major depressive disorder (Raison CL, et al. JAMA 2023;330:843–853. DOI 10.1001/jama.2023.14530, level 1b, n=104).

Durability looked convincing: twelve months after two sessions, 75% of participants maintained a clinical response and 58% were in remission (Gukasyan N, et al. J Psychopharmacol 2022;36:151–158. DOI 10.1177/02698811211073759, level 2b — open prospective follow-up, n=27). In the larger COMPASS cohort, however, effect size declined over time and some patients required repeat courses (J Clin Psychiatry 2025. DOI 10.4088/JCP.24m15449, level 2b — observational follow-up).

What 2026 changed

In May 2026 the EPISODE trial was published: two centres, triple blinding (investigator, participant, rater), active placebo, in patients with treatment-resistant depression withdrawn from antidepressants (Mertens LJ, et al. JAMA Psychiatry 2026;83:448–460. DOI 10.1001/jamapsychiatry.2026.0132, PMID 41848690, level 1b, n=144).

On the primary endpoint (≥50% reduction on HAMD-17 at week 6) response rates were 17.0% on 25 mg, 12.5% on 5 mg and 10.6% on nicotinamide. The adjusted odds ratio for 25 mg versus placebo was 1.73 (95% CI 0.53–6.23), p=0.19. The difference did not reach significance and, under the hierarchical testing plan, no further formal testing was performed. The authors note a clinically meaningful symptom reduction on secondary measures but describe the trial itself as inconclusive.

Note the absolute numbers: 17% versus 10.6%. That is a long way from "more than half of patients in remission" reported against a waiting list. The gap between those figures is not a difference between compounds — it is a difference between study designs.

Six- and twelve-month follow-up of the same cohort was published separately (Psychother Psychosom 2026. DOI 10.1159/000552272, PMID 42201843, level 2b — naturalistic follow-up).

COMPASS phase 3: endpoint met, effect size underwhelming

While EPISODE was being discussed, phase 3 results appeared — and they do not contradict it; they fit the same picture.

COMP005 (NCT05624268, n=255, a single 25 mg dose versus placebo in treatment-resistant depression) met its primary endpoint: the difference on the MADRS was −3.6 points at week 6 (95% CI −5.7…−1.5; p<0.001). COMP006 (NCT05711940, n=572, three arms — 25 mg, 10 mg and 1 mg, two administrations three weeks apart) also met its primary endpoint; per 26-week data released in July 2026, 39% of participants in the 25 mg arm sustained response at six months. Serious adverse events were 5.7% versus 6.3% in the comparator arm.

Three caveats without which those numbers cannot be read. First, neither trial has been peer-reviewed: as of late July 2026 there is no primary publication in PubMed, and on ClinicalTrials.gov both studies show no posted results. Everything known comes from the sponsor. Second, the effect size of COMP006 has not been disclosed: the company calls it "very similar" to the −3.6 of COMP005, meaning the single most important figure is unpublished. Third, −3.6 points on the MADRS is not much: the threshold for a clinically noticeable difference on that scale is usually taken to be around 2 points, and the result exceeds it only modestly.

Putting EPISODE and phase 3 together yields a consistent conclusion: the effect exists, it is reliably small, and the tighter the control of expectation, the smaller it gets.

The best synthesis available today confirms this: a living systematic review of 30 RCTs (n=1480, Eur Neuropsychopharmacol 2025. PMID 41205366) reports an SMD of −0.62 (95% CI −0.97…−0.28) in major depressive disorder with heterogeneity I²=55%, and grades the certainty of evidence under GRADE as very low — 83.3% of included RCTs were rated at high risk of bias.

How much the result depends on design is shown by a separate meta-analysis of 6 RCTs (J Affect Disord 2026. PMID 41876058): at the standard dose the SMD is −1.05, but excluding waiting-list studies it falls to −0.70, while heterogeneity drops from I²=75% to 43%. The gap between those numbers is, once again, a difference not between compounds but between designs.

The regulatory fork

On 24 April 2026 the FDA granted the company a Commissioner's National Priority Voucher, compressing review to one or two months; the final application is expected in the fourth quarter of 2026. On 14 July 2026 the FDA published its final guidance on clinical investigations of psychedelic drugs: low-dose psychoactive comparators are permitted in place of inert placebo, and early characterisation of the dose–response relationship is required.

The collision here is worth naming out loud: the regulator is accelerating review of a drug whose evidence base, by the best available synthesis, is graded "very low certainty".

A useful gauge of how many people actually reach this therapy where it is permitted: Australia has allowed prescribing by authorised psychiatrists since July 2023, and per regulator data, by September 2025 the programme had treated 47 patients with psilocybin and 87 with MDMA, with zero reported serious adverse events. The main lesson of the Australian experience so far is not about efficacy but about scale — real-world uptake turned out to be orders of magnitude smaller than expected.

Why one negative trial outweighs five positive ones

This is not special pleading against an inconvenient result. It is a basic principle of evidence-based medicine: a study weighs according to its protection from bias, not the direction of its conclusion.

The central problem across this field is blinding. A systematic review of 112 randomised psychedelic trials (JAMA Psychiatry 2026. DOI 10.1001/jamapsychiatry.2026.0255, PMID 41984443) found that only 29.5% assessed blinding integrity, although 57.1% cited blinding as a limitation. In psilocybin, LSD and ayahuasca studies, unblinding among participants and raters frequently exceeded 90%. No control strategy reliably preserved blinding.

Why this matters so much here: someone who has received 25 mg of psilocybin knows it almost without error — the experience cannot be mistaken for placebo. They understand they are in the "real" arm, and the expectation of improvement becomes part of the treatment. With subjective endpoints such as depression rating scales, that expectation is recorded as effect.

A second methodological thread concerns how control groups behave. A meta-analysis compared outcomes in control arms across trials of psilocybin, SSRIs and esketamine (JAMA Netw Open 2025;8:e2524119. DOI 10.1001/jamanetworkopen.2025.24119, PMID 40736734). When control arms behave differently across fields, directly comparing "response percentages" between trials loses meaning — which is exactly how popular summaries are usually assembled.

The practical conclusion as of July 2026 is "an effect probably exists, it is moderate, and its size is not yet established" — not "psilocybin outperforms antidepressants".

Addiction: the most unexpected result

The paradox is that the most convincing numbers come not from depression but from tobacco dependence. The first pilot produced a figure that looked implausible: 80% of participants were not smoking at six months (Johnson MW, et al. J Psychopharmacol 2014;28:983–992. DOI 10.1177/0269881114548296, level 2b — open pilot, n=15). At roughly 2.5 years, 60% remained abstinent (DOI 10.3109/00952990.2016.1170135, level 2b, n=15).

The key confirmation arrived in 2026 as a head-to-head against standard care. All participants completed the same 13-week cognitive behavioural programme, then received either a single psilocybin dose or a course of nicotine patches (JAMA Netw Open 2026. DOI 10.1001/jamanetworkopen.2026.0972, level 1b — pilot RCT, n=82). At six months, complete abstinence was 40.5% versus 10% — more than a sixfold higher probability of quitting. No serious treatment-related adverse events were recorded.

In alcohol use disorder, the first large RCT reported fewer heavy drinking days and complete abstinence in about 48% versus 24% on active placebo over eight months of follow-up (Bogenschutz MP, et al. JAMA Psychiatry 2022;79:953–962. DOI 10.1001/jamapsychiatry.2022.2096, level 1b, n=93). A Swiss phase 2 trial conducted after detoxification, however, found no significant difference in either abstinence duration or consumption — only a greater reduction in craving (eClinicalMedicine 2025. DOI 10.1016/j.eclinm.2025.103149, level 1b, n=37).

Cancer and existential distress

Palliative psychiatry is where the modern research programme began. Two 2016 trials — NYU (Ross S, et al. J Psychopharmacol 2016;30:1165–1180. DOI 10.1177/0269881116675512, level 1b, n=29) and Johns Hopkins (Griffiths RR, et al. J Psychopharmacol 2016;30:1181–1197. DOI 10.1177/0269881116675513, level 1b, n=51) — showed marked reductions in anxiety, depression and fear of death, sustained for at least six months.

Re-assessment of the NYU participants at an average of 4.5 years found persistently lower distress, and most described the experience as among the most significant events of their lives (Agin-Liebes GI, et al. J Psychopharmacol 2020;34:155–166. DOI 10.1177/0269881119897615, level 2b — observational follow-up, n=16). A group-therapy format in cancer patients with depression has been studied separately (Cancer 2024;130:1101–1112. DOI 10.1002/cncr.35010, PMID 38105655, level 2b — open-label, n=30).

The honest appraisal of this whole body of work comes from Cochrane: certainty of evidence for psychedelic-assisted therapy in anxiety, depression and existential distress in people with life-threatening illness was graded low (Cochrane Database Syst Rev 2024. DOI 10.1002/14651858.CD015383.pub2, PMID 39260823). Small samples, unachievable blinding, short follow-up.

Mechanism: what neuroimaging shows

Psilocybin is an agonist at serotonin 5-HT2A receptors, densely expressed in the cortex. The first fMRI study produced an unexpected result: instead of the anticipated activation, it found reduced activity in the medial prefrontal and posterior cingulate cortex — key hubs of the Default Mode Network (Carhart-Harris RL, et al. PNAS 2012;109:2138–2143. DOI 10.1073/pnas.1119598109).

This underpinned the "entropic brain" model: psychedelics temporarily loosen the rigidity of established neural networks, increasing flexibility of information processing (Front Hum Neurosci 2014;8:20. DOI 10.3389/fnhum.2014.00020). Direct confirmation came from high-resolution functional MRI: psilocybin desynchronises large-scale networks, most strongly the Default Mode Network, with synchrony subsequently restored (Siegel JS, et al. Nature 2024;632:131–138. DOI 10.1038/s41586-024-07624-5).

A separate line links the depth of subjective experience to outcome. A systematic review of 12 clinical studies found that in 10 of them the intensity of mystical-type experience statistically predicted symptom reduction (Front Psychiatry 2022;13:917199. DOI 10.3389/fpsyt.2022.917199). Interpret with care: this is a correlation, and it may partly express the same unblinding problem — the stronger the experience, the more obvious to the participant that they received the active compound.

What psilocybin does to hormones

No popular review contains this section, and for an endocrinologist it comes first: psilocybin is a serotonin receptor agonist, and the serotonin system directly governs pituitary secretion. "What does it do to hormones" is not an exotic question here but ordinary pharmacology.

Cortisol and ACTH: they rise, but only at a high dose

The classic dose-finding study (Hasler F, et al. Psychopharmacology 2004;172:145. PMID 14615876, double-blind placebo-controlled, n=8, doses 45/115/215/315 µg/kg) showed that ACTH and cortisol rise only at the maximum dose of 315 µg/kg — roughly 22 mg for a 70 kg person. The rise was confirmed at fixed doses of 15 and 30 mg (Holze F, et al. Neuropsychopharmacology 2022;47:1180. PMID 35217796, double-blind crossover, n=28).

A systematic review of nine studies measuring blood biomarkers (Prog Neuropsychopharmacol Biol Psychiatry 2025;136:111251. PMID 39788410) confirms that the rise in cortisol and ACTH is a stable finding. With a substantial caveat: all of these studies were done in healthy volunteers. The rise is acute and resolves as the drug wears off.

A curious detail from preclinical work: in mice the rise in glucocorticoids turned out to be necessary for the anxiolytic effect — blocking the glucocorticoid receptor abolished it (PMID 37588757). The hormonal response here may therefore be part of the mechanism rather than a side effect. This has not been tested in humans.

Prolactin: the most sensitive marker, and a practical trap

The same 2004 study contains something that is rarely quoted: prolactin rises already at 215 µg/kg, whereas ACTH, cortisol and TSH rise only at 315. The threshold for prolactin is lower, making it the most sensitive endocrine marker of psilocybin's action. Confirmed at 15 and 30 mg (PMID 35217796), and it is a class effect: ayahuasca (PMID 22005052) and DMT (PMID 8297216) do the same.

Hence a practical conclusion found in no guideline: prolactin measured in the hours after a session may be falsely elevated. And immediately the honest boundary of knowledge: not one study reports absolute prolactin values, the duration of the rise, or the time to return to baseline. How long after a session prolactin can be measured reliably has not been studied. If a patient has undergone psychedelic therapy and presents with "elevated prolactin", it is reasonable to establish the timing of the draw before referring them for pituitary MRI.

The thyroid: almost a blank space

Here something unpleasant has to be said. TSH after psilocybin has been measured in humans exactly once — in that same 2004 study, in eight participants, and it rose only at the maximum dose. Free T3 and T4 have never been measured.

There is not a single study in patients with thyroid disease. Not a single study in patients on levothyroxine. Data on interaction with thyroid therapy do not exist at all — neither pharmacokinetic nor pharmacodynamic.

The only signal is preclinical: in Wistar-Kyoto rats psilocybin raised TSH, reversing stress-induced suppression, and did so without changes in the adrenal axis (Sci Rep 2025. PMID 40419666). That proves nothing for humans, but it does show that the thyroid axis is not neutral in this story and would be worth studying.

Blood pressure and heart rate: psilocybin is a pressor, not a chronotrope

The best quantitative data come from a pooled analysis of 85 healthy participants and 113 administrations at 15/20/25/30 mg (Becker AM, Liechti ME, et al. Neurosci Appl 2024;10:104060. PMID 40656108). Peak systolic pressure: 131±10 mm Hg on placebo versus 146±14 on 30 mg; peak diastolic 81±7 versus 93±8. The highest values recorded were 180 and 115 mm Hg respectively. Systolic exceeded 140 in half of all administrations, 160 in 6%, and 180 in none.

Heart rate behaved differently: 74±10 bpm on placebo versus 78–82 on psilocybin, and the difference from placebo was not significant; tachycardia above 100 bpm was recorded in only 7% of administrations. The QTc interval was not prolonged (maximum 481 ms). Temperature above 38 °C arose dose-dependently, in 7% to 32% of administrations.

This is a clinically meaningful distinction: psilocybin raises blood pressure but barely accelerates the pulse. LSD has the opposite profile — a larger rise in heart rate and less in pressure (PMID 35217796). A meta-analysis of 6 RCTs (n=528) confirms elevated blood pressure as a class effect: RR 2.29 (95% CI 1.15–4.53) (JAMA Netw Open 2024. PMID 38598236).

And again a boundary of knowledge that has to be stated plainly: everything above was measured in selected healthy volunteers. Cardiovascular disease is a standard exclusion criterion in every trial (Pharmacol Rep 2023. PMID 37874530). A review of 214 studies with 3,504 participants recorded no cardiovascular serious adverse events, but systematic adverse-event assessment was applied in only 23.5% of the studies published after 2005 (JAMA Psychiatry 2024. PMID 39230883). What happens in a hypertensive patient is unknown — because hypertensive patients were not enrolled.

5-HT2B and cardiac valves: what nobody has done

In the time since the review of the theoretical risk of fibrosis and valvulopathy from regular 5-HT2B stimulation (PMID 38214279), no new human data have appeared at all. Two clarifications have.

First: an echocardiographic endpoint appeared in this literature for the first time — in mice. Under chronic microdosing, serotonin caused ventricular wall thickening and d-fenfluramine caused aortic regurgitation, while LSD caused neither (ACS Pharmacol Transl Sci 2025. PMID 41789300).

Second, and this is the main point: echocardiography as an endpoint has never been used in humans for any classical psychedelic (PMID 37572027). A 2026 review concludes outright that the risk with intermittent use is "insufficiently characterised", and that a confirmed valvular signal exists only for fenfluramine, ergoline dopamine agonists and, in part, MDMA (Clin Pharmacol Ther 2026. PMID 42209433).

A curiosity an endocrinologist will appreciate: in mice a low-dose non-psychedelic psilocybin analogue improved insulin sensitivity and reduced hepatic steatosis — through the very same 5-HT2B receptor that features in the safety section as a source of risk (Pharmacol Res 2026. PMID 41475502). One receptor, two opposite stories, and both still in animals.

Oxytocin, sex hormones, growth hormone

Oxytocin is the one measure on which two papers from the same group contradict each other. One reports a significant rise at 15 and 30 mg (PMID 35217796); the other, in 32 participants, finds a rise after mescaline and LSD but not after psilocybin (PMID 37231080). The correct way to present this is as an unresolved question.

Sex steroids have never been measured in humans after psilocybin. The only signal from the reproductive system is a series of three women: resumption of menses after amenorrhoea, earlier onset when taken in the second half of the luteal phase, and cycle normalisation in a woman later diagnosed with polycystic ovary syndrome (J Psychoactive Drugs 2024;56:1. PMID 36682064). The participants were recruited through social media — this generates a hypothesis and proves nothing.

Three things that simply do not exist

Reduced to its most honest statement, the endocrine part reads as follows: free T3 and T4 after psilocybin have never been measured in humans, and TSH once, in eight participants, twenty-two years ago; there is not one study in patients with endocrine disease or on replacement therapy; echocardiography as an endpoint has been used in no trial of any classical psychedelic.

This is not nitpicking but a description of the boundary. A compound discussed as the future of psychiatry has still not been studied in people with the most common endocrine pathology.

What I check as an endocrinologist before any conversation about psychedelics

This section appears in no popular review, and for clinical practice it matters more than the rest. Treatment-resistant depression is a diagnosis of exclusion. Before calling depression resistant, one has to be sure it is not being sustained by an unrecognised organic cause. In my practice that regularly turns out to be endocrine or nutritional.

The minimum panel before experimental therapy is even discussed:

▸ TSH, free T4, TPO antibodies — not TSH alone. Depression and anxiety are statistically associated with Hashimoto thyroiditis even with normal TSH and T4: in euthyroid Hashimoto, lifetime odds of depression run about 6.6 times higher and of anxiety disorder about 4.9 times higher.
▸ Vitamin D, vitamin B12, ferritin — iron deficiency without anaemia and low-normal B12 produce fatigue and low mood indistinguishable from depressive symptoms.
▸ Fasting glucose and HbA1c — insulin resistance is associated with depressive symptoms and fatigue.
▸ Morning cortisol where clinically suspected — both hyper- and hypocortisolism masquerade as affective disorders.

The point is not to declare all depression hormonal — that would be as much of an overreach as promising miracles from psilocybin. The point is sequence: first exclude the causes that have cheap, safe and well-proven treatment, and only then discuss a therapy whose odds ratio is 1.73 with a confidence interval crossing unity.

The endocrine side is covered in detail in a separate article: hypothyroidism, Hashimoto and the link with depression and anxiety.

The leading axis: how I work through a case before prescribing anything

The panel above is only the first layer. What actually determines the decision in my practice comes next: assessment by axes. Instead of listing complaints linearly, I grade several systems by severity of dysfunction and identify the leading axis — the one dragging the others behind it.

The axes assessed are: immune-inflammatory (including age-related low-grade inflammation and the aftermath of past viral infection), hormonal-regulatory (thyroid, adrenals, sex hormones), metabolic (glycaemia, insulin resistance, body composition), gastrointestinal and microbiotic, the stress–sleep–neuroendocrine axis, and nutritional — the deficiencies that constrain every other system.

The purpose is not to find "all the abnormalities" but to establish a hierarchy. The axes interact: gastrointestinal dysfunction may be feeding the immune-inflammatory axis rather than being a standalone problem. Treating a secondary axis in isolation from the leading one means acting on the consequence while the cause keeps running.

Why this matters so much in a conversation about depression and anxiety. A configuration I encounter regularly is one where low mood, anxiety and disrupted sleep are secondary to the immune-inflammatory or hormonal axis. Some of these states develop through what the literature calls sickness behaviour: an inflammatory signal alters behaviour and affect, and subjectively it is indistinguishable from "real" depression.

The practical implication is uncompromising: if an affective symptom is secondary, acting on it in isolation is a dead end. That holds equally for anxiolytics, for adaptogens and — all the more — for experimental psychedelic therapy with an odds ratio of 1.73 and a confidence interval crossing unity. Each of those scenarios reproduces the same error: treating a downstream symptom while the leading axis remains untouched. Only the cost of the error differs.

Hence my answer to "should I consider psilocybin for long-standing depression": first establish which axis is leading. If the affective picture turns out to be the summit of an immune-inflammatory or thyroid cascade, discussing psychedelic therapy is premature — regardless of what the trials showed.

Does psilocybin itself act on the inflammatory axis?

Since the immune-inflammatory axis stands first in this assessment, the question is fair. And the answer is unexpected: it does — and this is the one place where the human data on psilocybin look respectable.

A placebo-controlled study in 60 healthy volunteers (0.17 mg/kg versus placebo, Brain Behav Immun 2023, Mason NL et al.) showed an acute fall in TNF-α and, seven days later, sustained reductions in IL-6 and C-reactive protein. And crucially: the greater the fall in IL-6 and CRP by day seven, the more durable the positive changes in mood and social behaviour. The fall in TNF-α also correlated with reduced hippocampal glutamate on 7-tesla MR spectroscopy.

Before building a theory on this — three objections. First, another human study gives the opposite result: analysis of sera from 91 participants across three RCTs (Psychedelic Medicine 2024, DiRenzo D et al.) found a transient increase in IL-8 during the first week and no change at all by week four. Second, the preclinical work contradicts it directly: in rats a single dose caused a rise in nine mediators at once, including IL-1β and TNF-α, at 24 hours and more so at one week (J Psychedelic Studies 2022).

Third — and this looks like the resolution — a systematic review of 40 preclinical studies (Prog Neuropsychopharmacol Biol Psychiatry 2024, Low ZXB et al.) formulates the pattern: psychedelics reduce pre-existing inflammation but increase it against a normal baseline. If so, the direction of the effect depends on the state of the system — and then data obtained in healthy volunteers cannot be transferred to a patient with an active inflammatory axis.

And separately: in that same Mason study, psilocybin did not alter the response to a psychosocial stressor. There is an anti-inflammatory trace, then, but no restructuring of stress reactivity.

The conclusion in the logic of this article: the signal is interesting and, for once, human, but reproducibility is not established, the direction of the effect depends on baseline state, and the work was done in healthy volunteers. That is enough to make the hypothesis worth testing, and nowhere near enough to offer psilocybin as a way of acting on the inflammatory axis.

The same standard, applied to my own data

An article that grades other people's research by level of evidence is obliged to hold the same ruler up to itself. Data from my own practice is kept to a formalised scheme, and here is what it consists of.

▸ Prospective structured records. Each case is recorded against a single set of fields — baseline measures, change at defined checkpoints, therapy adjustments — rather than reconstructed from memory after the fact. Identification is anonymised: working records use a coded identifier, never a name.
▸ Accounting by participant-flow (CONSORT-style): how many enquired, how many met criteria, how many received the protocol, how many reached analysis. Withdrawals and those lost to follow-up are counted separately rather than vanishing from the denominator — which is precisely what separates an honest percentage from a marketing one.
▸ An adverse-event log with date, grade and resolution (resolved / resolved with sequelae / ongoing).
▸ STROBE reporting — the standard for describing observational research: inclusion criteria, data sources, measures against bias, limitations.
▸ An append-only corrections log. If a previously published figure turns out to be wrong, it is not silently overwritten — the correction is recorded as its own entry.
▸ A single source of truth for numbers. Figures live in one place and flow to the pages from there, instead of being duplicated across dozens of files and quietly drifting apart. The rule inside that scheme is simple: an unknown value stays marked unknown rather than being filled in with a plausible-sounding number.

And now the honest conclusion from all of the above. However carefully such work is conducted, it is observational data: no randomisation, no control group, no blinding. By the very hierarchy I use throughout this article, a case series from practice is level 2b–4. It describes practice and generates hypotheses. It does not prove efficacy and cannot serve as a counterargument to a randomised trial.

That is exactly why I do not set my own percentages against the EPISODE data, and do not present them here as evidence. Clinical experience is valuable for something else: it indicates where to look for the cause and in what order to exclude possibilities. Proof that an intervention works comes from a different type of study — and in the case of psilocybin, that type of study is so far returning an inconclusive result.

Safety: what is known about the risks

Under controlled conditions the safety profile is broadly favourable but not flawless. A meta-analysis and systematic review of psychedelic psychotherapy safety summarises the adverse-event data (Psychiatry Res 2024;335:115880. DOI 10.1016/j.psychres.2024.115880, PMID 38579460).

Specific signals from the most rigorous trial (EPISODE, 2026):

▸ Two serious adverse reactions after 25 mg, including one case of hallucinogen persisting perception disorder (HPPD) — a condition in which visual disturbances persist after the drug has worn off.
▸ Reports of suicidal ideation on dosing days: 4% versus 1–2% in comparator groups. This is precisely the patient population in which such a signal demands maximum caution.
▸ Most adverse events were acute, occurring during the session itself.

Trial screening is strict, and that is fundamental to interpreting the results. Exclusions typically include a personal history of psychotic disorders, psychosis or bipolar I disorder in first-degree relatives, uncontrolled hypertension and significant cardiac disease, and pregnancy. People who enter these trials are not the average patient with depression.

Drug interactions deserve separate attention: SSRIs and other serotonergic agents alter the subjective response (EPISODE deliberately withdrew participants from antidepressants), and combination with lithium has been described as increasing the risk of adverse neurological events. Stopping an antidepressant on one's own for a "cleaner experience" is a distinct and entirely real hazard.

A particular concern is microdosing. A 2024 review notes that psilocybin and LSD are structurally similar to drugs that caused cardiac fibrosis and valvulopathy with regular use (methysergide, pergolide, fenfluramine) through 5-HT2B receptor stimulation (J Psychopharmacol 2024;38:217–224. DOI 10.1177/02698811231225609, PMID 38214279). The long-term cardiac safety of a "two to four times a week for months" regimen has not been studied.

Antidepressants, lithium and other drugs

Interactions deserve a section of their own: this is the first thing a patient already on medication asks about, and it is exactly where popular reviews most often get it wrong in both directions — frightening people with what is unconfirmed and staying silent about what is confirmed.

Antidepressant withdrawal is probably a hidden confounder in EPISODE

I'll start with a finding that bears directly on the central trial of this article. In EPISODE, participants withdrew from antidepressants before enrolment. And there is work showing what that produces: a post hoc analysis of the psilocybin-versus-escitalopram trial (J Psychopharmacol 2024. PMID 38520045) found that patients who had discontinued an SSRI or SNRI before taking part had a worse treatment effect on every outcome scale than those who entered already medication-free — even though the acute intensity of the psychedelic experience itself did not differ. The authors explicitly call for a trial comparing discontinuation with continuation.

How large the contribution of withdrawal itself can be is shown by a meta-analysis of 79 studies and 21,002 patients (Lancet Psychiatry 2024. PMID 38851198): at least one withdrawal symptom occurs in 31% of people stopping an antidepressant versus 17% on placebo — placebo-adjusted, roughly one in six or seven; severe manifestations 2.8% versus 0.6%, and escitalopram is among the drugs with more frequent withdrawal. Another meta-analysis of 50 RCTs and 17,828 participants (JAMA Psychiatry 2025. PMID 40632531), however, rates the severity more modestly — about one additional symptom in the first week, below the threshold of clinical significance, and mood worsening was not related to discontinuation. The two papers disagree on severity, and that is worth stating plainly.

What this means for reading EPISODE: part of the difference between arms may have been absorbed by a withdrawal state that hit both arms equally. That does not rescue the result, but it helps explain why the absolute response figures — 17.0% and 10.6% — came out so much lower than in the early open-label work.

Can SSRIs be continued

The data are more reassuring than is generally assumed. A crossover RCT in 23 healthy volunteers (Clin Pharmacol Ther 2022. PMID 34743319) showed that a two-week course of escitalopram before 25 mg of psilocybin did not affect the positive effects but significantly reduced the negative ones — anxiety, distressing experience and cardiovascular responses. Psilocin pharmacokinetics were unchanged. What is blunted, in other words, is the unpleasant side of the experience rather than the therapeutic one.

An open-label phase 2 study in 19 patients with treatment-resistant depression who continued their SSRI (Neuropsychopharmacology 2023. PMID 37443386) produced a MADRS reduction of 14.9 points (95% CI −20.7 to −9.2) by week three, response and remission in 8 of 19 (42.1%), and no serious adverse events. The result is comparable to trials in which medication was withdrawn.

Against this are observational data on blunting: a survey of 611 people who combined psychedelics with antidepressants (J Psychopharmacol 2023. PMID 37291890) gives a probability of a weaker-than-expected effect of 0.47 for SSRIs and 0.55 for SNRIs versus 0.29 for bupropion. The blunting persisted even 3–6 months after discontinuation and was not explained by the long half-life of fluoxetine.

The honest bottom line: no controlled trial has randomised continuation versus discontinuation of an SSRI before psilocybin. Everything available is observational or post hoc.

Lithium: the one interaction with a genuinely worrying signal

The numbers here are impossible to skip. An analysis of reports of psychedelics combined with psychotropic medication (Pharmacopsychiatry 2021. PMID 34348413) found that of 62 accounts of taking a psychedelic while on lithium, seizures were mentioned in 47%, severe negative experiences in a further 18%, and 39% required medical attention. For comparison: of 34 accounts involving lamotrigine, not one involved a seizure.

There is also a published clinical case: an adolescent with three generalised seizures within half an hour of taking LSD while on a subtherapeutic lithium dose, with no seizure history, whose earlier exposures while on fluoxetine and bupropion had passed without consequence (S D Med 2024. PMID 38986146). Mechanistic support comes from an experiment: in rats given lithium, a 5-HT2A agonist provoked seizures at doses that cause none on their own (PMID 7583243).

The caveat is obligatory: the entire signal rests on internet reports and one clinical case — no cohort, no denominator, with obvious selection bias. But given an effect of that size and a cost of error that high, this is the case where waiting for better data is unreasonable.

Serotonin syndrome: a largely theoretical risk

Contrary to a widespread fear, there is not one documented case of serotonin syndrome from combining psilocybin with an SSRI in controlled trials. A pharmacology review (Psychopharmacology 2022. PMID 34251464) concludes that true serotonin toxicity requires an overdose or a combination with an MAO inhibitor, and that "non-MAOI psychedelic plus non-MAOI serotonergic drug" falls into the low-risk category. A review of 18 studies (PMID 40937732) states directly that taking antidepressants together with classical psychedelics is "generally safe and tolerable, without increased risk of serotonin syndrome, particularly for psilocybin".

The real risk zone is not SSRIs but preparations containing MAO inhibitors: ayahuasca contains β-carbolines, and that is where combination with serotonergic drugs is dangerous.

Psychiatric contraindications: what lies behind them

Excluding people with a personal or family history of psychosis goes back to a consensus safety document (J Psychopharmacol 2008. PMID 18593734) — an expert agreement rather than a quantitative risk estimate. Quantitative estimates came later. In a review of 214 studies (3,504 participants, JAMA Psychiatry 2024. PMID 39230883), serious adverse events amounted to 0% in healthy volunteers and about 4% in participants with pre-existing neuropsychiatric disorders — worsening depression, suicidal behaviour, psychosis, seizures. Psychotic reactions under controlled conditions are estimated at under 1% (PMID 40614615).

Bipolar disorder is a separate matter. A synthesis of 23 papers (Mol Psychiatry 2026. PMID 42215638) gives a spread that is telling in itself: rates of hypomania and mania of 5.8% in controlled psilocybin trials in depression and up to 30% in naturalistic samples with bipolar disorder. An open-label trial in bipolar II disorder (JAMA Psychiatry 2024. PMID 38055270, 15 participants dosed) found no mania signal, but it is far too small to exclude one.

Two gaps are worth naming plainly, so that absence of data is not mistaken for absence of risk: no controlled comparison of continuing versus discontinuing an SSRI has ever been run, and the lithium signal has not one study with a proper denominator.

What is NOT proven

▸ Microdosing as treatment. A review of controlled studies (Maastricht University, 2025) concludes that small doses produce subtle changes in mood and neural activity but no convincing clinical advantage over placebo.
▸ Alzheimer disease and neurodegeneration. Only preclinical data on dendritic growth and reduced neuroinflammation. There is no clinical evidence of efficacy in patients.
▸ Anorexia nervosa. A single phase 1 open-label study in females, n=10 (Nat Med 2023;29:1947–1953. DOI 10.1038/s41591-023-02455-9). That is a tolerability assessment, not efficacy.
▸ OCD. A 2006 pilot (n=9, DOI 10.4088/JCP.v67n1110) and a 2025 study using 10 mg (Compr Psychiatry 2025. DOI 10.1016/j.comppsych.2025.152619) where symptom reduction lasted about a week. Early stage.
▸ Durability for most patients. COMPASS data show gradual decay; some patients need repeat courses.
▸ Transfer of results to retreats and ceremonies. This is arguably the most important item. Trials involve screening for contraindications, a known dose of synthetic compound, preparation, medical supervision during the session and psychotherapeutic integration afterwards. Remove any of those components and you are studying something else. There are no safety or efficacy data outside that structure.

Legal status

Psilocybin is a controlled substance in most countries and legal access is limited to clinical research. Specific regulatory exceptions exist: since July 2023 authorised psychiatrists in Australia may prescribe psilocybin for treatment-resistant depression, and Oregon and Colorado operate supervised-session programmes. In the European Union and Ukraine, therapy is available only within research protocols.

This article is a review of scientific data, not a guide to action or a recommendation for use. I do not provide psilocybin-assisted therapy and do not advise on access to controlled substances.

Conclusion

Over twenty years psilocybin research moved from impossible to publishable in the leading journals, which is a scientific event in itself. The most convincing data concern tobacco dependence (40.5% versus 10%) and existential distress in cancer patients, though Cochrane grades the certainty of the latter as low.

For depression the picture is more complicated than it is usually presented. The two most rigorous trials — the escitalopram comparison (2021) and EPISODE (2026) — missed their primary endpoints. The systematic review of blinding showed the field suffers unblinding rates above 90%. That does not mean there is no effect; it means the true effect is smaller than advertised and not yet precisely established.

The practical implication for a patient with long-standing, unresponsive depression: before pursuing experimental options, it is worth excluding the causes that have cheap and proven treatment — the thyroid first of all, then B12, iron and vitamin D deficiency and disorders of carbohydrate metabolism. Less exciting than the neurobiology of mystical experience, but far ahead on the ratio of benefit, risk and evidence.

This article is informational and does not replace a medical consultation. Psilocybin is a controlled substance; all results cited were obtained under medical supervision and do not apply to unsupervised use. If you are experiencing depression, consult a physician: treatments exist with proven efficacy and a known safety profile.

References

▸ Mertens LJ, et al. Efficacy and Safety of Psilocybin in Treatment-Resistant Major Depression: The EPISODE Randomized Clinical Trial. JAMA Psychiatry 2026;83(5):448–460. DOI 10.1001/jamapsychiatry.2026.0132, PMID 41848690 — level 1b (triple-blind RCT, active placebo, n=144); primary endpoint not met.
▸ Blinding Integrity in Psychedelic Randomized Clinical Trials: A Systematic Review. JAMA Psychiatry 2026. DOI 10.1001/jamapsychiatry.2026.0255, PMID 41984443 — systematic review of 112 RCTs.
▸ Control Group Outcomes in Trials of Psilocybin, SSRIs, or Esketamine for Depression: A Meta-Analysis. JAMA Netw Open 2025;8:e2524119. DOI 10.1001/jamanetworkopen.2025.24119, PMID 40736734.
▸ Carhart-Harris RL, et al. Psilocybin with psychological support for treatment-resistant depression. Lancet Psychiatry 2016;3:619–627. DOI 10.1016/S2215-0366(16)30065-7 — level 2b (open-label, n=12).
▸ Davis AK, et al. Effects of Psilocybin-Assisted Therapy on Major Depressive Disorder. JAMA Psychiatry 2021;78:481–489. DOI 10.1001/jamapsychiatry.2020.3285 — level 1b (n=27, waiting-list control).
▸ Carhart-Harris R, et al. Trial of Psilocybin versus Escitalopram for Depression. N Engl J Med 2021;384:1402–1411. DOI 10.1056/NEJMoa2032994 — level 1b (n=59); primary endpoint not significant.
▸ Goodwin GM, et al. Single-Dose Psilocybin for a Treatment-Resistant Episode of Major Depression. N Engl J Med 2022;387:1637–1648. DOI 10.1056/NEJMoa2206443 — level 1b (n=233).
▸ Raison CL, et al. Single-Dose Psilocybin Treatment for Major Depressive Disorder. JAMA 2023;330:843–853. DOI 10.1001/jama.2023.14530 — level 1b (n=104).
▸ Gukasyan N, et al. Efficacy and safety of psilocybin-assisted treatment for major depressive disorder: prospective 12-month follow-up. J Psychopharmacol 2022;36:151–158. DOI 10.1177/02698811211073759 — level 2b (n=27).
▸ Long-term observational follow-up after a single dose of psilocybin for treatment-resistant depression. J Clin Psychiatry 2025. DOI 10.4088/JCP.24m15449 — level 2b.
▸ Psilocybin-assisted psychotherapy for treatment-resistant depression: randomized trial of repeated doses. Med 2024. DOI 10.1016/j.medj.2024.01.005, PMID 38359838 — level 1b (n=56).
▸ Johnson MW, et al. Pilot study of the 5-HT2AR agonist psilocybin in the treatment of tobacco addiction. J Psychopharmacol 2014;28:983–992. DOI 10.1177/0269881114548296 — level 2b (n=15).
▸ Johnson MW, et al. Long-term follow-up of psilocybin-facilitated smoking cessation. Am J Drug Alcohol Abuse 2017;43:55–60. DOI 10.3109/00952990.2016.1170135 — level 2b (n=15).
▸ Psilocybin or Nicotine Patch for Smoking Cessation: A Pilot Randomized Clinical Trial. JAMA Netw Open 2026. DOI 10.1001/jamanetworkopen.2026.0972 — level 1b (n=82); 40.5% versus 10%.
▸ Bogenschutz MP, et al. Percentage of Heavy Drinking Days Following Psilocybin-Assisted Psychotherapy vs Placebo in Alcohol Use Disorder. JAMA Psychiatry 2022;79:953–962. DOI 10.1001/jamapsychiatry.2022.2096 — level 1b (n=93).
▸ Psilocybin-assisted therapy for relapse prevention in alcohol use disorder: a phase 2 randomized clinical trial. eClinicalMedicine 2025. DOI 10.1016/j.eclinm.2025.103149 — level 1b (n=37); no significant difference.
▸ Ross S, et al. Rapid and sustained symptom reduction following psilocybin treatment for anxiety and depression in patients with life-threatening cancer. J Psychopharmacol 2016;30:1165–1180. DOI 10.1177/0269881116675512 — level 1b (n=29).
▸ Griffiths RR, et al. Psilocybin produces substantial and sustained decreases in depression and anxiety in patients with life-threatening cancer. J Psychopharmacol 2016;30:1181–1197. DOI 10.1177/0269881116675513 — level 1b (n=51).
▸ Agin-Liebes GI, et al. Long-term follow-up of psilocybin-assisted psychotherapy for psychiatric and existential distress in patients with life-threatening cancer. J Psychopharmacol 2020;34:155–166. DOI 10.1177/0269881119897615 — level 2b (n=16).
▸ Psilocybin-assisted group therapy in patients with cancer diagnosed with a major depressive disorder. Cancer 2024;130:1101–1112. DOI 10.1002/cncr.35010, PMID 38105655 — level 2b (n=30).
▸ Psychedelic-assisted therapy for treating anxiety, depression, and existential distress in people with life-threatening diseases. Cochrane Database Syst Rev 2024. DOI 10.1002/14651858.CD015383.pub2, PMID 39260823 — certainty of evidence low.
▸ Carhart-Harris RL, et al. Neural correlates of the psychedelic state as determined by fMRI studies with psilocybin. PNAS 2012;109:2138–2143. DOI 10.1073/pnas.1119598109.
▸ Carhart-Harris RL, et al. The entropic brain. Front Hum Neurosci 2014;8:20. DOI 10.3389/fnhum.2014.00020.
▸ Siegel JS, et al. Psilocybin desynchronizes the human brain. Nature 2024;632:131–138. DOI 10.1038/s41586-024-07624-5.
▸ Psychedelics, Mystical Experience, and Therapeutic Efficacy: A Systematic Review. Front Psychiatry 2022;13:917199. DOI 10.3389/fpsyt.2022.917199.
▸ Safety and risk assessment of psychedelic psychotherapy: a meta-analysis and systematic review. Psychiatry Res 2024;335:115880. DOI 10.1016/j.psychres.2024.115880, PMID 38579460.
▸ Rouaud A, et al. Microdosing psychedelics and the risk of cardiac fibrosis and valvulopathy. J Psychopharmacol 2024;38:217–224. DOI 10.1177/02698811231225609, PMID 38214279.
▸ Peck SK, et al. Psilocybin therapy for females with anorexia nervosa: a phase 1, open-label feasibility study. Nat Med 2023;29:1947–1953. DOI 10.1038/s41591-023-02455-9 — level 2b (n=10).
▸ Moreno FA, et al. Safety, tolerability, and efficacy of psilocybin in 9 patients with obsessive-compulsive disorder. J Clin Psychiatry 2006;67:1735–1740. DOI 10.4088/JCP.v67n1110 — level 4 (n=9).

Related article: hypothyroidism, Hashimoto and the link with depression and anxiety — what to check first when low mood persists.