Ch. V — Clinical evidence: depression and anxiety

Chapter V of Post-2010 Psychedelics: An Expert-Panel Review. For the executive summary and full table of contents, start there.

Abstract. This chapter audits the post-2010 clinical evidence base for serotonergic psychedelics in depression and anxiety. We treat the major psilocybin trials (Carhart-Harris 2016/2018/2021, Davis 2021, Goodwin 2022, von Rotz 2023, Raison 2023, COMP005/COMP006), the LSD program at Basel and MindMed (Gasser 2014; Holze 2023; Robison et al. 2025), and the methodological scaffolding (Muthukumaraswamy 2021; Aday 2022; Wen 2024) under the Cochrane Risk of Bias 2.0 framework. We argue that the direction of effect is now reasonably secure across multiple programs, but that the magnitude of effect remains materially confounded by functional unblinding and expectancy, with consequences for regulatory interpretation that the Compass Phase 3 readouts have made unavoidable rather than resolved.

5.1 Psilocybin in treatment-resistant depression

The psilocybin-TRD program is the most mature evidence base in modern psychedelic psychiatry, and it is also the cleanest test of whether the field can produce regulatory-grade evidence.

The Imperial College programme opened with an open-label, uncontrolled feasibility study (n=12, two doses of 10 mg and 25 mg seven days apart) reporting marked symptom reductions on the QIDS-SR16 at one week and three months¹. An expanded analysis (n=20) at six-month follow-up reported Cohen’s d = 2.2 at week 1 and 1.4 at six months on QIDS-SR16, though four of twenty patients met remission criteria at week 5². Both papers are open-label, single-arm, and small; they establish feasibility and signal, not efficacy.

The first comparative trial randomized 59 patients with moderate-to-severe major depression to either psilocybin (25 mg × 2, three weeks apart) plus placebo escitalopram or to escitalopram 10–20 mg/day plus 1 mg psilocybin × 2³. The primary endpoint (QIDS-SR16 change at week 6) did not significantly differ between arms. Secondary endpoints (BDI-1A, HAM-D-17, MADRS, response and remission rates) numerically favoured psilocybin, but multiplicity was not formally controlled and the trial was powered for the primary endpoint. The trial is frequently mis-described in secondary literature as showing psilocybin “superior” to escitalopram; the primary endpoint result does not support that claim.

The Compass program — synthetic, GMP-grade psilocybin (COMP360) under a fixed psychological-support protocol — entered Phase 2b with a single-dose, three-arm dose-finding study (COMP001) in n=233 patients with TRD (failure of ≥2 antidepressants in the current episode)⁴. At the primary endpoint (MADRS change from baseline to week 3), the 25 mg arm showed a placebo-adjusted reduction of –6.6 points versus 1 mg control (95% CI –10.2 to –2.9; p<0.001). The 10 mg arm did not separate significantly from control (–2.5, p=0.18). Response and remission rates at week 3 in the 25 mg arm were 36.7% and 29.1%. The trial is the largest single-dose psilocybin RCT to that point, and is the basis on which Compass advanced to Phase 3 at the 25 mg dose with a 1 mg active-arm “low-dose” control retained as the comparator.

The two pivotal Phase 3 trials are central. COMP005 (n=258, single 25 mg dose vs matched microcrystalline cellulose placebo, 32 US sites) met its primary endpoint of MADRS change at week 6 with a placebo-adjusted –3.6-point reduction (95% CI –5.7 to –1.5; p<0.001), announced 23 June 2025⁵. COMP006 (total dosed n=581 across three arms: 25 mg n=296, 10 mg n=142, 1 mg n=143; two doses of 25 mg three weeks apart, with the primary endpoint comparing 25 mg vs 1 mg) met its primary endpoint with a placebo-adjusted –3.8-point reduction at week 6 (95% CI –5.8 to –1.8; p<0.001), announced 17 February 2026⁶. In COMP005, approximately 25% of participants in the 25 mg arm achieved a ≥25% MADRS reduction at Week 6; in COMP006, approximately 39% in the 25 mg arm did so. COMP006 reported serious adverse events in 2% of participants and a pooled suicidal-ideation event rate of <1% across both trials (one suicidal-behaviour event occurring in the 1 mg arm of COMP006).

OLE denominator funnel for the COMP005 40% Week-26 remission figure. Compass has reported durability of response through Week 26 in both trials for participants achieving a ≥25% MADRS reduction at week 6; Week-26 disease-remission rate (MADRS ≤10) is reported as 40% in the redosed COMP005 Part B cohort.⁶ The underlying denominators reflect conditional sampling on early response, which the panel should see explicit. Of n=258 randomised to COMP005, approximately n=129 received the 25 mg dose; of those, approximately 25% (~32) met the ≥25%-MADRS-reduction response threshold at Week 6 and were eligible for the OLE re-dose protocol (Part B); of the 25-mg-arm non-remitters from Week 6 who entered Part B (a separate subset), Compass reports “over 40%” achieved remission after a second dose, a re-dose-conditional remission rate. Compass has not, as of the freeze date, published the full per-arm OLE entry funnel (the precise number of Week-6 responders entering Part B, the precise number completing Week 26 with outcome data, and arm-by-arm imputation method for OLE non-completers). The 40% figure is therefore response-conditional, not ITT, and the ITT-equivalent (Week-26 remission as a proportion of all originally-randomised 25 mg arm participants) is plausibly closer to 10–20% pending publication of the funnel. The same caveat applies to CYB003’s headline 71% 12-month remission rate: this is reported in n=12 evaluable patients out of an original Phase 2 trial population of n=22 (one parallel cohort) and represents a 12-month completer-conditional figure, not an ITT durability number; the 100% response rate among that n=12 is similarly conditional on retention.

The 25 mg arm’s –3.6 to –3.8 placebo-adjusted MADRS effect is statistically robust and is sustained at six months for early responders. It is also clinically modest on a 60-point scale, and is substantially smaller than the –6.6 effect reported in COMP001. Three plausible reasons: (i) regression to the mean toward truth under a larger and more representative sample (COMP001 was Phase 2b with stricter enrichment); (ii) re-calibration of the “1 mg active control” effect, which appears larger in Phase 3 than in COMP001 — possibly because expectancy in the 1 mg arm grew as the public salience of psilocybin grew; (iii) site-effect attenuation as the trials moved to more standard outpatient psychiatric settings. The panel should not assume the Phase 3 effect is “the true effect” any more than COMP001 was — both are point estimates under specific design and recruitment conditions.

Under RoB 2.0, the Compass Phase 3 program is at low risk for randomization, allocation concealment, and missing-data handling (Compass disclosed ITT analyses and pre-specified hierarchical testing), and at high risk on the deviations-from-intended-interventions and measurement-of-outcome domains because of functional unblinding. The 1 mg active-control design partially mitigates but does not solve this: in published psilocybin work the rate at which participants in active-dose arms correctly guess their assignment exceeds 85%, and the rate in the 1 mg arm is materially lower (see §5.5)⁷.

5.2 Psilocybin in non-resistant major depressive disorder

The MDD evidence base is smaller and more heterogeneous than the TRD program, but it bears on a regulatory question that the TRD program does not — whether psilocybin works on garden-variety MDD or only in the population that has already failed multiple antidepressants.

The Johns Hopkins waiting-list trial (Davis 2021, n=27, randomized to immediate vs delayed psilocybin-assisted therapy with two sessions at 20 mg/70 kg and 30 mg/70 kg) reported a between-group GRID-HAMD effect of approximately –13 points at week 5, with 71% response and 54% remission in the immediate-treatment arm⁸. The trial is significantly underpowered for a between-group estimate of that magnitude, and the comparator (waiting list, not active or low-dose control) is not blinded — all participants in the immediate arm knew they were receiving active treatment. The effect size is therefore an upper bound. A 12-month follow-up reported sustained remission in 58% of the cohort⁹.

The Zurich single-dose, niacin-controlled trial (von Rotz 2023, n=52, randomized 1:1, 0.215 mg/kg psilocybin vs placebo, 14-day primary endpoint) reported a placebo-adjusted MADRS reduction of approximately –13 points and a BDI reduction of approximately –12 points¹⁰. Like Davis 2021, the effect size is large; unlike Davis 2021, the trial is double-blind with an inert placebo and the functional-unblinding rate is reported.

The Usona Institute Phase 2 trial (PSIL201, Raison 2023, n=104, 25 mg psilocybin vs 100 mg niacin active placebo, 11 US sites) is the largest randomized MDD trial published as of the freeze date¹¹. At the primary endpoint (MADRS change from baseline to day 43), psilocybin produced a placebo-adjusted reduction of –12.3 points (95% CI –17.5 to –7.2; p<0.001). Response (≥50% reduction) was 41.7% vs 11.4%; sustained remission at day 43 was 25.0% vs 9.1%. The effect is large and is closer to Davis 2021 than to COMP005/006. The Phase 3 follow-up (uAspire / PSIL301) is ongoing.

The Cybin program (CYB003, a deuterated psilocybin analog) received FDA Breakthrough Therapy Designation on 13 March 2024¹² and has reported an acute Phase 2 readout (n=22, parallel arms at 12 mg) showing 79% remission at six weeks after two doses, and a 12-month follow-up in 12 evaluable patients showing 71% remission and a mean MADRS change-from-baseline of approximately –23 points after two 16 mg doses¹³. The sample is very small, the design is not placebo-controlled in the long-term follow-up, and the durability number is anchored on a subset of completers; the headline 71% remission at 12 months and the parallel 79% / 100% response framing cannot bear the regulatory weight Cybin’s commercial communication has placed on it. The funnel is, by Cybin’s own disclosures: of the original n=22 parallel-arm cohort at 12 mg / 16 mg, n=12 were evaluable at the 12-month visit, of whom 71% met remission criteria — meaning the 12-month remission rate expressed as a fraction of the original randomised cohort is approximately 12 × 0.71 / 22 ≈ 39% ITT-equivalent, materially lower than the 71% headline. Cybin’s investor communications do not consistently disclose this funnel, and the pivotal PARADIGM Phase 3 program (APPROACH, EMBRACE, EXTEND) is now enrolling at population scale where this question will be definitively settled.

The cross-trial heterogeneity in MDD is striking: Davis 2021, von Rotz 2023, and Raison 2023 all report MADRS-or-equivalent placebo-adjusted effects of roughly –12 to –13 points; COMP005 and COMP006 in TRD report –3.6 and –3.8. Two readings of that gap: (a) TRD is genuinely harder to treat and the underlying effect is smaller; (b) the smaller, single-site, less placebo-rigorous MDD trials have substantially inflated effect estimates from expectancy and selection. Both are partially true. The Phase 3 MDD readouts from Usona and Cybin will sharpen this question.

5.3 LSD in depression and anxiety

The LSD evidence base is smaller than psilocybin’s and divides into two streams: existential anxiety in life-threatening illness (Swiss-led), and primary generalized anxiety disorder (the MindMed/Liechti program).

The Gasser study (n=12, two-arm: 200 µg LSD vs 20 µg active placebo, double-blind with crossover for the placebo arm) is the foundational modern LSD trial¹⁴. It reported reductions in trait anxiety on the STAI of approximately 1.1 effect size with sustained benefit at 12-month follow-up¹⁵. The sample is too small for inferential weight but is methodologically important as the first modern controlled LSD trial of any kind. Holze and Liechti at Basel subsequently published the dose-response pharmacokinetic profile of oral LSD across 25–200 µg doses, anchoring the dose-finding literature¹⁶.

The Holze 2023 Phase II trial (n=42, two-period crossover, 200 µg vs placebo, double-blind) reported a STAI-Global reduction of –16.2 points (95% CI –27.8 to –4.5; p=0.007) sixteen weeks after the last LSD session, with 65% of patients meeting response criteria (≥30% STAI-G reduction) vs 9% on placebo¹⁷. The trial enrolled patients with anxiety, with or without life-threatening illness, so it is not a pure existential-anxiety trial. The crossover design provides within-subject control; functional unblinding is essentially complete in any 200 µg LSD vs placebo design.

The MindMed MM120 Phase 2b trial (Robison et al. 2025, n=198 randomized across five arms: 25, 50, 100, 200 µg, and placebo) was published in JAMA in October 2025¹⁸. The dose-response relationship at the primary endpoint (HAM-A change from baseline at week 4) was significant for both 100 µg (least-squares mean difference vs placebo: –5.0 points; 95% CI –9.6 to –0.4) and 200 µg (–6.0 points; 95% CI –9.8 to –2.0). The 25 µg and 50 µg arms did not separate from placebo.

The “100 µg outperforms 200 µg” framing that has circulated in secondary literature is imprecise. At the primary endpoint, the 200 µg HAM-A point estimate is numerically larger than the 100 µg estimate (–6.0 vs –5.0). The reason MindMed selected 100 µg for Phase 3 is composite: (i) faster onset of effect; (ii) longer durability through week 12; (iii) a materially lower adverse-event burden — nausea was 40% at 100 µg vs 60% at 200 µg, and acute autonomic and dissociative effects scaled with dose. The 100 µg dose was also adequate to achieve the primary endpoint without the additional safety overhead. Phase 3 (Voyage, US; Panorama, US/EU) is enrolling at 100 µg.

The “anomaly” the panel will probe is real but in a different direction than is usually claimed: the primary endpoint estimate goes the way pharmacology would predict (higher dose, larger numerical effect), while the durability and tolerability favoured the lower dose. MindMed’s Phase 3 dose-selection logic is defensible. What the trial does not establish is whether the durability gradient between 100 µg and 200 µg reflects a true pharmacological inverted-U or whether differential dropout, expectancy, and functional unblinding contribute. The Robison et al. 2025 publication reports completer-vs-ITT analysis distinctions in the supplementary appendix; for the 200 µg arm, dropout was higher than for 100 µg, plausibly leaving a non-randomised completer set in which the apparent durability gradient is conditional on tolerability. The receptor-occupancy framing (Chapter III) does not, on first-principles pharmacology, predict an inverted-U on anxiolytic effect at 100 µg vs 200 µg — 5-HT2A occupancy at 200 µg is higher than at 100 µg, and if the therapeutic mechanism is receptor-engagement-mediated plasticity, dose-response should be monotonic. The dose-finding result is therefore most parsimoniously explained as a tolerability-mediated durability differential: 200 µg participants experienced more anxiogenic acute effect, dropped out more frequently, and the completer-set durability is biased. Whether the 100 µg dose would survive an FDA Type C meeting request for a head-to-head 100 vs 200 vs placebo three-arm Phase 3 with pre-specified non-inferiority margin on durability is uncertain; MindMed’s Voyage/Panorama programmes test 100 µg only. Phase 3 will test convergence, not resolve mechanism.

Under RoB 2.0, MM120 Phase 2b is at low risk for randomization, allocation concealment, and outcome assessment (the HAM-A is rater-administered with separation of dosing from outcome-rating staff). It is at high risk on functional unblinding for the same reason as every classical psychedelic trial.

5.4 Bridging across compounds: effect sizes in context

A useful triangulation: where does psychedelic-assisted treatment sit relative to other antidepressant interventions?

Goldberg et al. (2020) meta-analyzed four early psilocybin RCTs (N=117 across the included studies) and reported within-group pre-post Hedges’ g of 1.16–1.47 and placebo-controlled pre-post effects of g ≈ 0.82–0.83; heterogeneity was substantial (I² ≈ 70–80% across the major outcomes), placing the pooled estimate in the “uninterpretable as a single population parameter” range¹⁹. Galvão-Coelho et al. (2021) pooled twelve trials of classical psychedelics including LSD, psilocybin, and ayahuasca (n=257) and reported moderate-to-large effects on depressive symptoms in mood-disorder patients at acute, medium, and longer-term endpoints with reported I² in the 40–65% range across endpoints — moderate-to-substantial heterogeneity²⁰. Metaxa and Clarke (2024) updated the psilocybin-for-depression evidence in BMJ across nine studies (n=436) and reported Hedges’ g = 0.66 (95% CI 0.46–0.86), with I² ≈ 60–70% (substantial heterogeneity) and moderate risk of bias across studies²¹. The Goldberg g ≈ 0.82 vs Metaxa g ≈ 0.66 gap is itself evidence of between-trial heterogeneity — a 25–30% pooled-effect drop as the literature grew from 117 to 436 patients and as more rigorous Phase 3 work entered the meta-analytic pool. The honest framing is that the literature exhibits substantial between-trial heterogeneity (I² consistently above 50%, often above 65%), which means the pooled point estimates are between-trial averages that may not characterise any specific patient population well; the patient-population question — TRD-vs-MDD-vs-cancer-related-depression — is one of the principal sources of the heterogeneity and is partly addressed by stratified Metaxa analyses but not resolved.

The Metaxa and Clarke estimate is the cleanest of the three and is consistent with what Phase 3 has produced: a moderate but not dramatic effect size that is, on the standardized-mean-difference scale, in the range of effective conventional antidepressants. Cipriani’s network meta-analysis of conventional antidepressants reported SMDs of roughly 0.30–0.40 for the head-to-head efficacious agents; psilocybin’s g≈0.66 is meaningfully larger but is not in the territory of “transformatively effective.”

A second comparator anchor: the MM120 100 µg arm produced a –7.6-point HAM-A reduction from baseline (week 12), vs roughly –3.6 in the placebo arm¹⁸. The placebo-adjusted effect (~4 points on a 56-point HAM-A) is comparable to standard SSRIs in GAD on the same scale at comparable timepoints, but achieved with a single dose rather than continuous administration. The route, not the magnitude, is the distinguishing feature.

A third comparator: see §5.6 for ketamine.

Comparator triangulation in TRD/MDD: vortioxetine, augmentation strategies, and ECT. A clinical pharmacologist will want the comparator universe extended beyond esketamine. Three reference points: (i) vortioxetine in TRD-adjacent samples (Citrome 2014 review; Mahableshwarkar 2015) produces placebo-adjusted MADRS reductions of approximately –2 to –4 points at 8 weeks, in the range of COMP005/006. (ii) Atypical-antipsychotic augmentation of failed first-line SSRI/SNRI (aripiprazole, quetiapine XR, brexpiprazole; Spielmans 2013 MA; Mohamed STAR*D follow-ons) gives placebo-adjusted MADRS reductions of –2 to –5 points, again broadly in the COMP005/006 range. (iii) ECT in TRD (UK ECT Review Group 2003 Lancet MA; Pagnin 2004) produces HAM-D reductions of –15 to –20 points relative to sham ECT in placebo-controlled bilateral trials — placing ECT in transformative territory that no current psychedelic-pharmacology programme approaches. The honest framing of psilocybin’s –3.6/–3.8 Phase 3 effect is: somewhere between augmentation pharmacotherapy and ECT (closer to augmentation), advantaged by acute durability (Week 26 in responders) if it holds at population scale through Phase 4. The pharmacoeconomic case is not “best-in-class effect size”; it is “comparable-effect-size with episodic-rather-than-chronic dosing.” That distinction is fragile against alternatives if maintenance dosing proves necessary at 12–18 months, which Phase 4 will settle.

A consistent finding across meta-analyses is that effect estimates from early-phase, single-site, smaller trials substantially exceed estimates from larger Phase 3 trials. This is what the methodology literature would predict (publication bias, expectancy, smaller-trial variance), and it is one of the most defensible reasons for the panel to discount headline 79%-remission figures from Phase 2 readouts.

5.5 Methodology critique under Cochrane RoB 2.0

Modern psychedelic trials are at low or moderate risk on three of the five RoB 2.0 domains: randomization, missing-data handling (the Compass, Usona, and MindMed programs use ITT with pre-specified imputation), and selective-reporting (pre-registered protocols are now standard in industry-sponsored work). They are at high risk on two domains, both of which compound: (1) deviations from intended interventions arising from functional unblinding; (2) measurement of outcome contaminated by expectancy bias in patient-rated and rater-administered scales alike.

Functional unblinding is the central problem. Muthukumaraswamy et al. (2021) reviewed blinding-assessment data across psychedelic RCTs and reported guess-correctness rates exceeding 85% in active-dose arms versus chance (~50%) for inert placebo arms⁷. The Compass and MindMed programs partially mitigate by using active low-dose comparators (1 mg psilocybin; 25 µg / 50 µg LSD), but these comparators were chosen precisely because they were thought to be subjectively inert at the population level — meaning they are still functionally identifiable from a 25 mg / 100 µg active dose by most participants. The Carhart-Harris 2021 psilocybin-vs-escitalopram trial reported that 86% of psilocybin-arm participants correctly guessed their assignment.

Aday et al. (2022) systematized the recommendations: (i) measure pre-treatment expectancy explicitly with validated instruments; (ii) measure post-treatment blinding success; (iii) use active comparators with similar subjective signatures where pharmacologically feasible (e.g., methylphenidate, niacin, modafinil, dextromethorphan); (iv) train sites in expectancy-neutral preparation; (v) report functional-unblinding rates in primary publications²². The Compass Phase 3 program reports blinding success but the data has not been fully published; the Robison et al. MM120 publication reports limited functional-unblinding data.

Wen et al. (2024), in a systematic review of psychedelic-trial design under Blumberger and Daskalakis at the Centre for Addiction and Mental Health, reported that across the studies that did assess blinding, 100% of subjects in active arms and 93% in placebo arms correctly identified their assignment in seven of the reviewed trials²³. The implication is that for any psychedelic-arm participant who shows a large pre-post change, it is currently impossible to determine the fraction of that change attributable to (a) pharmacology, (b) expectancy in the patient, (c) confirmation bias in the rater (where outcome is rater-assessed and the rater is also functionally unblinded), and (d) the standard psychotherapy frame in which the dosing occurs.

This is not a reason to discard the evidence base. It is a reason to (a) treat headline effect sizes as overestimates of pure pharmacological effect; (b) weight Phase 3 effects more heavily than Phase 2 effects (Phase 3 trials have larger and more diverse samples in which expectancy is more averaged across baseline-belief heterogeneity); (c) read durability data (Week 26 in COMP005, twelve months in CYB003) as the most informative endpoint, because expectancy effects in conventional pharmacology rarely persist for six months without continued dosing — though they can persist longer in the presence of dramatic acute experience.

A further methodological issue is that the therapeutic frame in psychedelic trials is not standardised across sponsors. Compass uses a fixed but non-manualised psychological-support model; Usona uses the EMBARK model; MindMed uses a comparatively minimal therapy frame; the Imperial / Carhart-Harris work uses an in-house preparatory and integration protocol. The variation in non-pharmacological frame is itself uncontrolled in cross-trial comparisons, which means that effect estimates may reflect therapy as much as drug. This is a foundational ambiguity in the field that the regulatory pathway has not yet resolved; FDA has approved Spravato (esketamine) under REMS without specifying a therapy frame, but the psilocybin pathway is plausibly going to require therapy-as-component, which complicates standard NDA review.

A final note on multiplicity: the Phase 2 programs frequently report effect estimates on multiple correlated scales (MADRS, BDI, QIDS, HAM-D, GRID-HAMD, HAM-A, STAI, and clinician-impression scales). The Compass Phase 3 protocols used hierarchical testing for the primary endpoint and a small number of key secondary endpoints, with proper alpha control; the Phase 2 literature is much less consistent. Headline secondary-endpoint effects from Phase 2 papers should be treated as hypothesis-generating, not confirmatory.

The table below records, for each major trial, which endpoints carry FWER-controlled (alpha-protected) p-values versus nominal (hypothesis-generating) p-values:

The frequent secondary-endpoint cross-trial claims of “consistency” lack formal multiplicity adjustment. In particular: durability-of-response across Week 26 in COMP005 is alpha-protected through Week 12 but durability through Week 26 enters the “secondary durability” cluster that is hierarchically downstream of the primary endpoint; the OLE Part B 40%-remission figure is a non-hypothesis-test descriptive statistic. The panel should treat the Phase 3 primaries (MADRS Δ at Week 6 for COMP005/006; MADRS Δ Day 29 for BPL-003; HAM-A Δ Week 4 for MM120; CAPS-5 Δ for MAPP1/2) as the load-bearing claims; all other secondary endpoints discussed across this chapter should be read as hypothesis-generating with appropriate caution.

5.6 Ketamine as comparator

Ketamine and esketamine sit alongside the classical psychedelics as the only other approved or near-approved rapid-acting agents for treatment-resistant depression, and they provide the most informative comparator for effect size and durability.

Daly et al. (2018), the foundational intranasal esketamine trial (Phase 2, n=67, four-week, randomized to esketamine 28/56/84 mg twice weekly plus oral antidepressant vs placebo), reported MADRS reductions of approximately –4 to –10 points relative to placebo across dose arms, with onset within hours²⁴. The pivotal Phase 3 TRANSFORM trials produced placebo-adjusted MADRS reductions of approximately –4 points at week 4 (TRANSFORM-2, the positive of the three), which is the basis on which the FDA approved Spravato in March 2019 under REMS restrictions for in-clinic administration.

The SUSTAIN-2 trial (Wajs 2020, n=802, one-year open-label, esketamine 28/56/84 mg plus oral antidepressant) demonstrated maintenance of effect through one year with manageable safety²⁵. SUSTAIN-3, the open-label extension, has now reported safety and durability beyond two years.

The instructive comparison for the panel is this: esketamine’s pivotal placebo-adjusted MADRS effect (approximately –4 points) is essentially indistinguishable from the COMP005/006 effect (–3.6 to –3.8), and the durability requires continued dosing (typically weekly to bi-weekly indefinitely). Psilocybin’s Phase 3 durability profile, with sustained response through Week 26 after one or two doses, is the principal pharmacoeconomic argument for a different regulatory and care pathway. Whether that durability survives larger, more diverse Phase 4 samples is the open question.

Esketamine is also a useful reference for functional-unblinding: ketamine produces dissociation that is also obvious to patients, but the FDA approval framework absorbed this through the REMS structure rather than insisting on perfect blinding. The regulatory precedent has implications for psilocybin’s NDA path that are taken up in Chapter XI.

5.7 Where the depression/anxiety evidence sits in 2026

The state of the evidence base at the freeze date can be summarized as follows:

• Psilocybin-TRD: two positive Phase 3 trials with replicated primary endpoints and sustained durability through Week 26. Compass has guided to NDA filing in Q4 2026. The effect size is statistically robust and clinically modest. The regulatory question is whether FDA accepts the 1 mg active-control design as adequate, given the precedent established by the Lykos CRL where functional unblinding contributed to non-approval (see Chapter XI for regulatory framing; Chapter XII for the Lykos methodology forensics).
• Psilocybin-MDD (non-resistant): Phase 2 evidence (Davis 2021, von Rotz 2023, Raison 2023) consistently shows large effect estimates; Phase 3 (Usona uAspire, Cybin PARADIGM) is ongoing. The effect-size gap between TRD and MDD Phase 2 estimates is a key analytic question.
• LSD-anxiety: Phase 2b for MM120 in GAD published in JAMA; Phase 3 (Voyage, Panorama) topline expected H2 2026. The dose-selection logic for 100 µg is defensible. Whether MM120’s Phase 3 effect estimate matches the Phase 2b estimate is the open question.
• 5-MeO-DMT: see Chapter VII. Phase 2b for both BPL-003 (intranasal) and GH001 (inhalable) is positive; GH001’s placebo-adjusted –15.5-point MADRS effect is implausibly large and warrants Phase 3 replication before being treated as a real population-level estimate.

A cautionary frame: the Lykos pathway (Phase 3 success → AdCom rejection → CRL) demonstrates that the gap between Phase 3 endpoint achievement and FDA approval is wider for psychedelic compounds than for conventional pharmacology. The Lykos AdCom was lost not on efficacy but on functional unblinding, trial-conduct concerns, and adverse-event reporting (see Chapter XII for the methodology forensics). The Compass and MindMed Phase 3 programs differ from MAPP1/MAPP2 in important respects — synthetic compound, fixed-dose, more standardized therapy frame, no allegations of misconduct — but the underlying methodological challenge of functional unblinding has not been solved.

A second cautionary frame is what Burke, Blumberger, Wen and colleagues have called the appropriate scope of “psychedelic exceptionalism”²⁶. Wen et al. (2024) is the cleanest recent statement that psychedelic trials should not be evaluated under a different evidentiary standard from other psychiatric pharmacology, and that effect estimates currently in circulation are likely overestimates absent the methodological reforms Aday et al. recommended²³²².

5.8 Open questions

1. Durability beyond six months: Compass has reported through Week 26. Cybin has reported through 12 months in n=12. We do not have publicly disclosed durability data in TRD samples beyond six months at population scale. The cost-effectiveness of a single-dose paradigm pivots on this number; if 50% relapse by 18 months, the pharmacoeconomic case shifts toward maintenance.
2. Relapse handling and re-dosing: COMP006 used a two-dose induction; COMP005 used a single dose. The optimal induction regimen is not established. Whether a relapse responds to re-dosing as well as the first dose is not established.
3. Integration with maintenance: psilocybin / LSD will plausibly be deployed alongside continued SSRIs, psychotherapy, or both. The interaction effects are not well-studied; the Carhart-Harris 2021 trial is the only published head-to-head against an SSRI and was not designed to test combination.
4. Cost-effectiveness: an in-clinic 8-hour monitored dosing visit at oncology-grade staffing costs is approximately $5,000–10,000 per session in current US pricing models. Compass’s projected list price is in the same range. Without payer reimbursement comparable to Spravato, real-world access will be limited even on approval.
5. Identifying responders: the variance in patient-level response is large. Pre-treatment predictors — baseline severity, anxious-distress subtype, personality traits, prior psychedelic experience, suggestibility — have been investigated but no validated screening tool exists.
6. Non-responders and adverse outcomes: a non-trivial fraction of patients (≥30% in most trials) do not respond or relapse. The population that worsens under psilocybin is not well-characterized. HPPD, suicidality signals (the COMP001 25 mg arm reported three suicide-related events), and protracted destabilization are real and rare; the trial samples are too small to estimate base rates with precision (see Chapter XII).

References

1. Carhart-Harris RL, Bolstridge M, Rucker J, et al. Psilocybin with psychological support for treatment-resistant depression: an open-label feasibility study. Lancet Psychiatry 2016;3(7):619–627. PMID: 27210031.
2. Carhart-Harris RL, Bolstridge M, Day CMJ, et al. Psilocybin with psychological support for treatment-resistant depression: six-month follow-up. Psychopharmacology (Berl) 2018;235(2):399–408. PMID: 29119217.
3. Carhart-Harris RL, Giribaldi B, Watts R, et al. Trial of psilocybin versus escitalopram for depression. N Engl J Med 2021;384(15):1402–1411. PMID: 33852780.
4. Goodwin GM, Aaronson ST, Alvarez O, et al. Single-dose psilocybin for a treatment-resistant episode of major depression. N Engl J Med 2022;387(18):1637–1648. PMID: 36322843.
5. Compass Pathways press release, COMP005 primary-endpoint readout, 23 June 2025.
6. Compass Pathways press release, COMP006 primary-endpoint readout, 17 February 2026.
7. Davis AK, Barrett FS, May DG, et al. Effects of psilocybin-assisted therapy on major depressive disorder. JAMA Psychiatry 2021;78(5):481–489. PMID: 33146667.
8. Gukasyan N, Davis AK, Barrett FS, et al. Efficacy and safety of psilocybin-assisted treatment for major depressive disorder: prospective 12-month follow-up. J Psychopharmacol 2022;36(2):151–158. PMID: 35166158.
9. von Rotz R, Schindowski EM, Jungwirth J, et al. Single-dose psilocybin-assisted therapy in major depressive disorder. eClinicalMedicine 2023;56:101809. PMID: 36636296.
10. Raison CL, Sanacora G, Woolley J, et al. Single-dose psilocybin treatment for major depressive disorder: a randomized clinical trial. JAMA 2023;330(9):843–853. PMID: 37651119.
11. Cybin Inc. CYB003 FDA Breakthrough Therapy Designation press release, 13 March 2024. 11a. Cybin Inc. CYB003 Phase 2 12-month efficacy press release, 18 November 2024. [VERIFY — no peer-reviewed publication identified.]
12. Gasser P, Holstein D, Michel Y, et al. Safety and efficacy of lysergic acid diethylamide-assisted psychotherapy for anxiety associated with life-threatening diseases. J Nerv Ment Dis 2014;202(7):513–520. PMID: 24594678.
13. Gasser P, Kirchner K, Passie T. LSD-assisted psychotherapy for anxiety associated with a life-threatening disease: qualitative follow-up. J Psychopharmacol 2015;29(1):57–68. PMID: 25389218.
14. Holze F, Vizeli P, Ley L, et al. Pharmacokinetics and pharmacodynamics of LSD microdoses. Clin Pharmacol Ther 2021;109(3):658–666. PMID: 32975835.
15. Holze F, Gasser P, Müller F, Dolder PC, Liechti ME. LSD-assisted therapy in patients with anxiety with and without a life-threatening illness: a randomized, double-blind, placebo-controlled phase II study. Biol Psychiatry 2023;93(3):215–223. PMID: 36266118.
16. Robison R, Barrow R, Conant C, et al. Single treatment with MM120 (lysergide) in generalized anxiety disorder: a randomized clinical trial. JAMA 2025;334(15):1358–1372. PMID: 40906494.
17. Muthukumaraswamy SD, Forsyth A, Lumley T. Blinding and expectancy confounds in psychedelic randomized controlled trials. Expert Rev Clin Pharmacol 2021;14(9):1133–1152. PMID: 34038314.
18. Aday JS, Heifets BD, Pratscher SD, Bradley E, Rosen R, Woolley JD. Great Expectations: recommendations for improving the methodological rigor of psychedelic clinical trials. Psychopharmacology (Berl) 2022;239(6):1989–2010. PMID: 35359159.
19. Wen A, Singhal N, Jones BDM, Zeifman RJ, Mehta S, Shenasa MA, Blumberger DM, Daskalakis ZJ, Weissman CR. A systematic review of study design and placebo controls in psychedelic research. Psychedelic Med (New Rochelle) 2024;2(1):15–24. PMID: 40051762.
20. Goldberg SB, Pace BT, Nicholas CR, Raison CL, Hutson PR. The experimental effects of psilocybin on symptoms of anxiety and depression: a meta-analysis. Psychiatry Res 2020;284:112749. PMID: 31931272.
21. Galvão-Coelho NL, Marx W, Gonzalez M, et al. Classic serotonergic psychedelics for mood and depressive symptoms: a meta-analysis. Psychopharmacology (Berl) 2021;238(2):341–354. PMID: 33427944.
22. Metaxa A-M, Clarke M. Efficacy of psilocybin for treating symptoms of depression: systematic review and meta-analysis. BMJ 2024;385:e078084. PMID: 38692686.
23. Daly EJ, Singh JB, Fedgchin M, et al. Efficacy and safety of intranasal esketamine adjunctive to oral antidepressant therapy in treatment-resistant depression. JAMA Psychiatry 2018;75(2):139–148. PMID: 29282469.
24. Wajs E, Aluisio L, Holder R, et al. Esketamine nasal spray plus oral antidepressant in patients with TRD: long-term safety phase 3 (SUSTAIN-2). J Clin Psychiatry 2020;81(3):19m12891. PMID: 32316080.
25. Burke MJ, Blumberger DM. Caution at psychiatry’s psychedelic frontier. Nat Med 2021;27(10):1687–1688. PMID: 34635858. doi:10.1038/s41591-021-01524-1.

← Ch. IV · Overview · Ch. VI →