EPIsoDE Phase IIb: what the largest psilocybin TRD trial actually shows.

What treatment-resistant depression actually is

Treatment-resistant depression (TRD) is the clinical category for a major depressive episode that has failed to respond to at least two adequate trials of antidepressant medication at therapeutic dose and duration. It is not a personality trait, not a failure of effort, and not synonymous with "hard depression" colloquially. Roughly one in three patients with major depressive disorder meet TRD criteria within their first three years of treatment.

That matters for context. SSRIs (selective serotonin reuptake inhibitors), SNRIs (serotonin-noradrenaline reuptake inhibitors), bupropion, and the augmentation strategies built on top of them work — when they work — by gradually shifting baseline neurotransmitter tone over weeks. They do not produce rapid, single-event remission. The two existing approved exceptions are ketamine (esketamine / Spravato, FDA-approved for TRD in 2019) and electroconvulsive therapy. Both can produce rapid effect; both have access, cost, and stigma problems that have kept them under-utilised relative to the burden of the condition.

Psilocybin entered this conversation because its acute mechanism — a partial agonist at the 5-HT_2A serotonin receptor — drives a transient surge in cortical neuroplasticity that is a single-event biology, and early-phase data suggested durable symptomatic effects out of a single guided session. That is the bet. EPIsoDE is the first large, rigorously controlled test of whether the bet pays at Phase IIb scale.

EPIsoDE: what the trial did

EPIsoDE was a triple-blinded (participant, investigator, rater), three-arm, active placebo-controlled Phase IIb randomised clinical trial conducted across the Central Institute of Mental Health (CIMH) in Mannheim and Charité — Universitätsmedizin Berlin, with the MIND Foundation as collaborator [Gründer 2026 EPIsoDE]. Lead investigator: Gerhard Gründer, Heidelberg University. Patients recruited between 2021 and 2024. Total enrolment: 144 adults with treatment-resistant major depression (mean age 42.6 years, 59% male).

Three arms:

• Psilocybin 25 mg — the dose used in the COMP-360 and Imperial College trials.
• Psilocybin 5 mg — a low active dose.
• Nicotinamide 100 mg — an active placebo chosen because it produces no perceptible psychoactive effect but matches the dosing-day ritual.

Patients received two dosing sessions spaced six weeks apart. The primary endpoint was a ≥50% reduction on the Hamilton Rating Scale for Depression (HAMD17) at week 6 after the first dose. Each session was supported with non-directive psychological preparation and integration — the now-standard psychedelic-assisted-psychotherapy wrapper.

Of the 144 randomised, 143 received at least one dose and 142 completed the primary endpoint analysis. By the dropout and completion math alone, this was a tightly run trial.

Why the primary endpoint missed

At week 6 — the pre-registered primary timepoint — response rates were 17.0% in the 25 mg arm, 12.5% in the 5 mg arm, and 10.6% in the nicotinamide arm. The between-group differences did not reach statistical significance. By any honest reading of the primary endpoint, EPIsoDE failed.

That is the headline. It is also incomplete.

At week 1, the 25 mg arm reported a 34.0% response rate vs. 6.4–10.4% in the comparator arms. The effect was real, was large, and was visible early — and then decayed sharply by the prespecified week-6 endpoint. The trial caught the canonical psychedelic signal (rapid onset, single-event remission for a substantial subset of responders), and then watched it fade.

The implication is uncomfortable for both camps. For psilocybin sceptics, the rapid onset is not a placebo artefact — placebo arms do not produce response rates that high that fast in TRD. The signal is real. For psilocybin advocates, the durability claim — single dose, persistent remission — is on shakier ground than the early marketing suggested. By week 6, the effect had narrowed substantially. That is consistent with what we know about 5-HT_2A-mediated plasticity: the acute window is finite, and the lasting effect depends heavily on what the patient and the therapeutic frame do with the open window.

The signal psilocybin pulls is rapid, real, and short. The clinical question is no longer "does it work" but "how do you hold the opening it creates."

The secondary endpoints — and why they matter

Pre-specified secondary endpoints — mean change in HAMD17 and Beck Depression Inventory II (BDI-II) — showed clinically meaningful differences favouring the 25 mg arm at multiple timepoints, including the week-6 timepoint where the binary responder analysis missed [Gründer 2026 EPIsoDE]. Average change from baseline at week 12 across all groups was roughly −7.5 HAMD17 points, with 27–34% response rates by that follow-up window.

Why does the secondary readout matter when the primary missed? Two reasons.

First, depression is a continuous condition. A pre-specified binary endpoint (≥50% improvement, yes/no) discards information that a continuous-measure analysis preserves. A patient whose HAMD17 fell from 22 to 13 is doing materially better. That patient does not clear the binary bar but is captured by mean-change analysis. The secondary outcomes told us the 25 mg arm produced a real antidepressant signal that the primary endpoint partially masked.

Second, the regulatory and clinical questions are different. From a strict FDA-style trial-design perspective, missing the pre-registered primary endpoint is the only thing that matters. From a clinical question of "is this intervention worth carrying forward into Phase III with a refined design," the secondary readout is exactly the signal you would look for. The published expert reaction was consistent on this: EPIsoDE did not prove efficacy at its pre-specified bar, but it did not disprove a meaningful signal either. The trial design — particularly the choice of the strict 50% response threshold, the comparator, and the six-week endpoint — is what is now under discussion.

That is also where EPIsoDE diverges from the earlier COMP-360 readout from COMPASS Pathways, which reported a larger effect at three weeks in a similarly-designed trial published in the New England Journal of Medicine in 2022 [Goodwin 2022 COMP-360]. The two reads, side by side, suggest the psilocybin TRD effect is genuine but smaller and more variable than the early-trial narrative implied.

The 2026 OCD parallel: a much cleaner signal

While EPIsoDE was reading out in Germany, the parallel evidence base in obsessive-compulsive disorder (OCD) was producing arguably cleaner results. A 2026 Phase II randomised, double-blind, niacin-controlled trial conducted at Yale randomised 28 adults with treatment-resistant OCD to a single dose of psilocybin (0.25 mg/kg) or niacin placebo [Ching 2026 OCD].

At 48 hours post-dose, A-YBOCS (Acute Yale-Brown Obsessive-Compulsive Scale) scores in the psilocybin group dropped from 24.07 (±6.02) to 14.31 (±8.83) — a between-group difference of 9.83 points (p < 0.001, Cohen's d = 1.64). At one week, 69.2% (9/13) of psilocybin participants met the response criterion (≥35% Y-BOCS reduction) vs. 0% (0/14) on niacin (p < 0.001; number needed to treat = 1.4). The effect persisted out to the 12-week follow-up window.

That is a much cleaner readout than EPIsoDE produced, in a smaller trial, in a different condition. The OCD literature has historically been the second psilocybin signal after depression — Moreno et al. published the foundational small-cohort OCD case series in the early 2000s — and the 2026 readout strengthens it materially.

Reading EPIsoDE and the OCD trial in parallel is the honest 2026 view: psilocybin produces a real, rapid, durable signal in OCD, and a real but variable and harder-to-hold signal in TRD. Both findings are interesting. They do not say the same thing.

Where psilocybin sits on the regulatory clock in 2026

FDA breakthrough therapy designation for psilocybin in TRD was granted in 2018, accelerating the early-phase pathway. Phase III readouts from COMPASS Pathways' COMP-360 program have been anticipated through 2025–2026; the EPIsoDE primary-endpoint miss will be read against the COMP005 and COMP006 Phase III data as it emerges. As of this writing, no psilocybin formulation is FDA-approved for any indication; the program is in advanced clinical development.

Outside the FDA pathway:

• Australia permits authorised psychiatrists to prescribe psilocybin for TRD since July 2023 — the first country to permit clinical prescription outside trial settings.
• Oregon (Measure 109) operates supervised psilocybin services since 2023, regulated as a service-provider framework rather than a medical model.
• Colorado (Proposition 122) is in the rollout phase of a similar service-provider framework.
• Canada's Special Access Program permits case-by-case psilocybin access for end-of-life distress and select TRD cases since 2022.

These pathways are not the same as FDA approval. They produce access without producing the kind of structured efficacy and safety data the Phase III program is designed to produce. Patients accessing psilocybin through state or program pathways are not in a trial and are not being measured.

What "set and setting" means — and the risks the wellness press skips

"Set and setting" — the patient's psychological state going into the session and the physical, interpersonal, and therapeutic environment of the session itself — is not a romantic phrase. It is the operational variable that the structured-trial literature has repeatedly identified as the strongest moderator of outcome [Carhart-Harris 2018]. Translated to clinical practice: the preparation, dosing-day support, and integration sessions are not optional adjuncts. They are the intervention.

That is the case the wellness press skips, in both directions. The "psilocybin is a miracle" framing skips it by implying the molecule alone produces the effect. The "psilocybin is just a placebo with better marketing" framing skips it by treating the therapy frame as confounding rather than as the active ingredient.

Risks, taken seriously:

• Acute cardiovascular load — psilocybin transiently raises blood pressure and heart rate. Patients with uncontrolled hypertension, recent cardiac events, or significant structural heart disease are not candidates outside specialised settings.
• Suicidal ideation — in EPIsoDE, 4% of participants in the 25 mg arm reported suicidal ideation post-dose, vs. 1–2% in comparator arms. This is consistent with prior trial reports and is a reason the integration window matters.
• Hallucinogen Persisting Perception Disorder (HPPD) — a rare but documented persistent perceptual disturbance after psychedelic use. EPIsoDE reported one case among 144 participants.
• Psychiatric contraindications — personal or first-degree-family-history of psychotic-spectrum illness (schizophrenia, bipolar I) is a standard exclusion in the clinical literature. The risk is precipitating a first episode or relapse.
• Drug-drug interactions — SSRIs and SNRIs attenuate the psilocybin effect by occupying 5-HT_2A receptors. MAO inhibitors substantially amplify it. Lithium increases seizure risk. Trial protocols require structured tapers, not abrupt discontinuation, for any patient transitioning from chronic antidepressant therapy.

These are not abstract concerns. They are the reason "set and setting" has structural meaning: the screening, preparation, and clinical monitoring is what converts a high-risk psychoactive experience into a low-risk therapeutic one.

A tiered framework

We do not write protocols. We write frameworks that you take to a clinician. With that established: