Real-World Psilocybin Therapy for Treatment-Resistant Depression: a Retrospective Observational Study

Depressive disorders are highly prevalent and a major contributor to global disease burden; a substantial proportion of people with major depressive disorder (MDD) fail to achieve remission after two or more antidepressant trials and are classified as having treatment-resistant depression (TRD). Psilocybin, a 5-HT2A receptor agonist that produces dose-dependent alterations in cognition, perception and affect, has shown antidepressant effects in several randomized controlled trials (RCTs). However, RCTs often use strict eligibility criteria, have difficulties with blinding because of pronounced subjective effects, and typically test single-dose paradigms, which limits generalisability to more complex clinical populations and leaves unanswered questions about optimal dosing schedules and durability of effect. This retrospective study by J. and colleagues set out to examine the effectiveness and tolerability of psilocybin therapy delivered in a real-world clinical setting in Switzerland, where limited medical use exemptions have allowed flexible administration. The authors aimed to describe changes in depressive symptoms and adverse events in TRD patients receiving between one and four oral psilocybin dosing sessions (20–35 mg), and to test the hypothesis that psilocybin would reduce depressive symptoms and that additional dosing sessions would produce further symptom reduction. The focus is on naturalistic outcomes that may complement findings from tightly controlled trials and inform future clinical practice and research on multidose paradigms.

Design and setting: This was a retrospective observational study of electronic health records from the University Hospital of Psychiatry Zurich, covering treatments between February 2023 and May 2025. Ethical approval was obtained from the Cantonal Ethics Committee Zurich and patients provided written informed consent for the use of anonymised clinical data. Participants: The sample comprised 19 outpatients with ICD-10–diagnosed treatment-resistant depressive episodes who attended a specialised Second Opinion Consultation Service for Depression. Inclusion required failure of at least two prior antidepressant trials. Patients were aged 27–67 and were screened for physical health and contraindications using medical history, physical exam, laboratory tests, cranial MRI, ECG and EEG. Exclusion criteria included a personal or first-degree family history of psychotic or bipolar disorders, uncontrolled hypertension, recent cardiovascular events, certain cerebrovascular conditions, increased intracranial pressure, inadequately treated hyperthyroidism, and pregnancy or breastfeeding. Intervention and clinical procedures: Treatment was delivered in single in-/outpatient or group inpatient formats; patients received between one and four dosing sessions. Inpatient group courses were typically two sessions two weeks apart within a four-week stay, whereas single-setting sessions were individually scheduled. Psilocybin (synthetic, 5 mg capsules) was administered orally at individualized doses of 20–35 mg formulated under Good Manufacturing Practice. Preparation included at least two preparatory sessions to build therapeutic rapport and set expectations. Antidepressant medications were generally continued but paused on dosing days; medications with 5-HT2 receptor–blocking properties were paused for at least three half-lives. Dosing days were supervised by trained therapists (at least one in single sessions, at least two in group sessions). Integration meetings followed dosing to support reflection and transfer of insights. Adverse events were assessed via a 16-item questionnaire given immediately after dosing and via routine clinical documentation of serious adverse events. Outcomes and timing: Primary outcomes were depressive symptom scores on the clinician-rated Montgomery-Åsberg Depression Rating Scale (MADRS) and the self-report Beck Depression Inventory II (BDI). Assessments were taken at baseline (closest measurement prior to the first dosing), between dosing sessions and at the first follow-up within 42 days after the final dosing session (used as the post-treatment measurement). Secondary outcomes included categorical indicators: response (≥ 50% reduction from baseline), remission (BDI ≤ 13; MADRS ≤ 12) and reliable change assessed by the Reliable Change Index (RCI). Adverse events (AEs) and serious AEs were secondary outcomes. Statistical analysis: Paired t-tests and Wilcoxon signed-rank tests were used to compare baseline to post-treatment scores depending on normality; Hedges' g and rank-biserial correlation were reported as effect sizes. A multiple linear regression was used to examine potential bias from variable timing of post-treatment assessments (details reported in the supplement). To examine change across dosing sessions, linear mixed-effects models were fitted with session number as a fixed effect (0–4) and participant random intercepts; estimated marginal means and Tukey-adjusted pairwise comparisons were reported. AEs were summarised by number of participants affected and frequency across dosing sessions. Analyses were conducted in R; two-tailed p < 0.05 was the threshold for statistical significance.

Sample and treatment exposure: Nineteen patients entered treatment and completed the first dosing session. Thirteen completed two dosing sessions, five completed three, and three completed four. For primary endpoint analyses, one participant lacked a valid MADRS endpoint and a different participant lacked a valid BDI endpoint, yielding n = 18 for analyses on both instruments. Primary outcomes — clinician-rated MADRS: MADRS scores decreased from baseline (mean = 30.78, SD = 9.02) to the post-treatment assessment after the last dosing session (mean = 19.89, SD = 9.65). The mean reduction was 10.89 points; the paired t-test was t(17) = 6.067, p < .001. The effect size was very large (Hedges' g = 1.37). On categorical outcomes, 6 of 18 patients (33.3%) met the response criterion (≥ 50% reduction) on the MADRS, and 4 of 18 (22.2%) met the remission threshold (MADRS ≤ 12). The mean percentage reduction on the MADRS was 35.6%. Using the RCI (threshold = 8.21 points), 10 patients (55.6%) showed reliable improvement, eight (44.4%) showed no reliable change, and none showed reliable deterioration. Primary outcomes — self-reported BDI: BDI scores decreased from baseline (mean = 32.33, SD = 12.25) to post-treatment (mean = 23.28, SD = 13.88). A Wilcoxon signed-rank test indicated a significant reduction in BDI scores (V = 153.5, p = .003). The mean reduction was 9.06 points and the effect size was large (r = .80). Five of 18 patients (27.8%) met response criteria on the BDI, and five (27.8%) met remission criteria (BDI ≤ 13). The mean percentage reduction on the BDI was 26.8%. Using the RCI (threshold = 9.5 points), eight patients (44.4%) showed reliable improvement, ten (55.6%) showed no reliable change, and none showed reliable deterioration. Change across dosing sessions: Analyses examining depressive symptoms across multiple dosing sessions indicated that the largest improvement occurred after the first dosing session, and subsequent sessions did not show consistent additive benefits at the group level. MADRS scores remained significantly decreased versus baseline across sessions, but pairwise significance compared with baseline disappeared from the third session onwards. On the BDI, significant improvement over baseline was observed after the first two sessions, with only trend-level improvements thereafter. The authors note that only a small number of patients received more than two sessions, increasing uncertainty around these session-level comparisons. Adverse events and tolerability: Across 30 recorded dosing sessions, at least one adverse event was reported in 24 sessions (80%). Patients reporting any AE tended to report multiple events (mean number of AEs per affected patient = 3.53, SD = 2.11, range 1–8). The most frequent AEs were fatigue (n = 20 occurrences), headache (n = 16) and tearfulness (n = 15). All reported AEs were transient, and no serious adverse events were recorded during the observation period.

J. and colleagues interpret their findings as evidence that psilocybin administered flexibly in a real-world clinical setting was associated with statistically significant and clinically meaningful reductions in depressive symptoms among TRD patients on both clinician-rated and self-report measures. The magnitude of mean symptom reductions (MADRS reduction = 10.89 points; BDI reduction ≈ 9 points) and large effect sizes were noted, although the authors state these improvements were somewhat smaller than pooled effects reported in prior clinical trials of MDD and TRD. Response and remission rates observed here (MADRS: response 33.3%, remission 22.2%; BDI: response 27.8%, remission 27.8%) were lower than those in some earlier RCTs that reported higher short-term response and remission, but fall within the lower range of more recent large trials. Regarding multiple dosing, the authors found no clear additive benefit at the group level: the largest symptom change occurred after the first dosing session, with limited incremental improvement from subsequent sessions. They propose several explanations: the small number of participants receiving more than two sessions (limiting power), continuation of treatment even when early non-response occurred (so later sessions were more often administered to poorer responders), heterogeneity in inter-session intervals (a possible optimal timing window is unknown), and the possibility that additional sessions simply do not confer added benefit for many patients. They also note that some individual patients did respond only after a second or third session, suggesting potential individual-level benefits. The authors situate their study as among the first to examine a multidose psilocybin paradigm in a naturalistic TRD population and argue that the sample reflects the complexity of routine clinical care—patients with multiple comorbidities, numerous prior treatment failures and chronic illness—thereby addressing a limitation of highly selective RCT samples. Nevertheless, they acknowledge important limitations: small sample size and attendant limited statistical power; underrepresentation of female and non-European patients limiting generalisability; retrospective design and reliance on clinical records that preclude causal inference; heterogeneity in treatment procedures, concomitant pharmacotherapy and psychotherapy, dosing schedules and measurement timepoints; absence of a control group; and the possibility that some patients received further dosing after the measurement window. These factors make it difficult to attribute observed changes solely to psilocybin rather than to adjunct treatments or nonspecific clinical effects. The authors recommend that future research employ larger prospective observational designs, ideally with comparator treatments (for example esketamine or electroconvulsive therapy), and control for concomitant medication, psychotherapy type and setting. They also highlight the need to identify biomarkers or predictors of treatment response to improve patient stratification and cost-effectiveness, given the resource intensity of current psilocybin treatment models. In sum, the authors conclude that their study provides initial real-world evidence of clinically meaningful symptom reduction and acceptable tolerability in TRD patients treated with one to four psilocybin sessions, while emphasising that larger, prospective studies are required to draw more definitive conclusions.