Psilocybin-assisted therapy for depression: A systematic review and meta-analysis

The review followed PRISMA guidance and a prospectively registered protocol on PROSPERO (CRD 42022220190). Multiple electronic databases (including Embase, Cochrane Database of Systematic Reviews and PsycInfo) were searched from inception to February 2023. Two reviewers independently screened records using Rayyan, resolving disagreements by consensus and full-text review. Eligible studies were double-blind randomized or cross-over RCTs with at least one week duration, conducted in adults aged 18–65, comparing fixed doses of orally administered psilocybin (in capsules) against placebo in patients with primary depression (major depressive disorder, MDD) or secondary depression (depressive symptoms associated with a preceding incapacitating medical illness, for example cancer). Trials in healthy volunteers, uncontrolled studies, case reports and reviews were excluded. Diagnostic criteria were based on DSM or SCID versions where reported. Most trials provided non-directive supportive psychotherapy in an individual psychedelic-assisted psychotherapy setting. The primary outcome was change in depressive symptoms from baseline to endpoint, extracted as mean ± SD on validated scales (e.g. HAM-D, MADRS). Missing SDs were imputed from p-values or using the mean SD from other included studies. Anxiety change (Beck Anxiety Inventory) and adverse events were extracted as secondary outcomes. Risk of bias was assessed independently with the Cochrane RoB 2 tool across standard domains. For analysis, dose was standardised to total mg/70 kg. A one-stage dose-response meta-analysis was implemented using the doresmeta package in R (v4.2.2) with random-effects models and restricted cubic spline (nonlinear) functions, using knots at the 25th, 50th and 75th percentiles. Standardised mean difference (Cohen's d) was the effect measure for continuous outcomes. ED50 (dose producing half the maximal estimated effect) and ED95 (near-maximal effective dose) were estimated from the curves. For binary adverse events, the authors modelled the dose–log risk ratio relationship using a one-stage random-effects approach. Between-study heterogeneity was quantified with the variance partition coefficient (VPC), a multivariate extension of I2. Sensitivity and leave-one-out analyses were planned where appropriate.

The search identified 5,196 references; after screening 260 trials for eligibility the authors included 7 RCTs totaling 489 participants. Across studies the mean age was 44.1 years (±12.8), 53.2% were female, and mean ages for subgroup analyses were 39.2 (±11.6) years for primary depression and 56.3 (±5.3) years for secondary depression. Included trials dated from 2011–2023, follow-up durations varied up to 33 weeks (mean follow-up reported as 14.14 weeks for several studies), and the mean interval between first and second psilocybin doses was 3.54 weeks (range 2–8). Doses administered ranged from 1.5 to 50 mg/70 kg; studies delivering a single dose had a mean of 16.5 mg/70 kg and studies with two doses averaged 31.5 mg/70 kg. Prior lifetime hallucinogen use was reported in 26.3% of participants, though details such as time since last use were rarely provided. Primary outcome — depressive symptoms: Pooling all 7 RCTs (N = 489) produced a significant dose-response relationship (p < 0.0001) with substantial heterogeneity (I2 = 85%). The estimated ED50 was 10.13 mg/70 kg (95% CI 6.6–14.5) and the ED95 was 41.14 mg/70 kg (95% CI 26.37–47.80), with the curve visually ascending and beginning to plateau at higher doses. Sensitivity analysis excluding studies judged at non-low overall risk yielded similar results (ED50 10.13 mg/70 kg; ED95 36.08 mg/70 kg). Excluding the single trial of treatment-resistant depression (Goodwin et al.) lowered estimates (ED50 8.23 mg/70 kg; ED95 24.05 mg/70 kg). Leave-one-out analyses produced ED50 values in the 8–14 mg/70 kg range and ED95 values between 24–45 mg/70 kg. Subgroup analyses: For primary depression (n = 4 trials; N = 366) there was a significant dose-response relationship (p < 0.0001) with considerable heterogeneity (I2 = 80%), and the curve plateaued at an ED95 of 24.68 mg/70 kg. For secondary depression associated with cancer (n = 3 trials; N = 123) the dose-response association did not reach statistical significance (p = 0.07) and showed substantial heterogeneity (I2 = 80%); visual inspection suggested a bell-shaped curve with a near-maximal dose reported as ED95 = 8.92 mg/70 kg (the extracted text did not clearly report the associated confidence interval). Anxiety outcomes: Six RCTs (N = 258) reporting anxiety change showed a significant overall dose-response association (p < 0.0001) with plateauing at higher doses and notable heterogeneity (I2 = 85%). The pooled ED50 for anxiety was 7.58 mg/70 kg and the ED95 22.78 mg/70 kg (95% CI 15.82–47.2). Subgroup estimates differed: for primary depression the ED95 for anxiety reduction was reported as 24.68 mg/70 kg with ED50 8.86 mg/70 kg; for secondary depression the ED50 was lower (reported as 4.08 mg/70 kg), reflecting differing dose–response profiles by population. Adverse events: Dose-dependent associations were found for several somatic and psychological adverse events. Physical discomfort increased with dose (p = 0.023), modelled as a +2.35% relative risk per additional 1 mg/70 kg (i.e. RR multiplier ≈1.0235 per mg). Blood pressure increases also showed a small but significant dose-association (p = 0.042; RR ≈ +1.04% per mg). Tachycardia showed a non-significant trend (p = 0.09; RR ≈ +2.02% per mg) with large uncertainty. Nausea/vomiting and headache/migraine followed bell-shaped dose-response curves (significant at p < 0.001 and p < 0.05 respectively) with reported relative risk increments of +1.25% and +1.42% per mg on the exponential scale at certain ranges. For combined prolonged psychosis and hallucinogen persisting perception disorder (HPPD) there was a significant dose-response association (p = 0.001) with an estimated RR increment of +2.51% per 1 mg/70 kg, but these events were uncommon (12 cases among 465 subjects) and outcome definitions varied across trials. The authors noted that thresholds and definitions for adverse events (for example what constituted a blood pressure increase) were inconsistent between studies, limiting comparability.

Perez and colleagues interpret their findings as the first dose-response meta-analysis of psilocybin for depression, showing a statistically significant relationship between psilocybin dose and reduction in depressive symptoms overall, with different dose–response shapes across populations. For primary MDD the curve tended to plateau (near-maximal effect at lower doses than the pooled estimate), whereas for secondary depression related to cancer a bell-shaped curve was observed with a much lower near-maximal dose. The authors highlight that treatment-resistant patients in the single trial available appeared to require higher doses, and excluding that trial lowered the pooled ED95 substantially, illustrating how population mix affects dose estimates. The paper emphasises that anxiety outcomes also followed dose-dependent patterns and that several somatic and psychological adverse events increased with dose, though most were transient, mild-to-moderate, and manageable with psychological support during sessions. The authors discuss potential mechanisms and clinical implications, noting that 5-HT2A receptor stimulation may produce both therapeutic and paradoxical adverse psychological effects (for example anxiety or transient psychosis-like states), and that subjective factors — such as novelty of the psychedelic experience, set and setting, prior hallucinogen exposure and age — could influence response and the optimal dosing strategy. Large standardised effect sizes (reported SMDs of about 1.62 for primary depression and 1.0 for secondary depression) are noted and contrasted with typical antidepressant effect sizes (~0.3), but the investigators acknowledge that therapeutic context and the intensity of the psychedelic experience may contribute to these large effects. Key limitations acknowledged by the authors include substantial heterogeneity across included trials, especially within secondary-depression studies that combined diverse diagnoses (e.g. adjustment disorder, generalized anxiety disorder, dysthymia), inability to isolate effects for each diagnostic subtype, sparse data in treatment-resistant depression (one trial only), variable and sometimes ambiguous adverse-event definitions, uncertainty about long-term efficacy beyond one month in many trials, and concerns about blinding adequacy in psychedelic trials. The authors call for additional double-blind RCTs in specific populations (including treatment-resistant depression) that more consistently report subjective experience measures and standardised adverse-event definitions, and that examine how non-pharmacological factors (set, setting, preparation and therapeutic model) interact with dose to determine clinical response and safety.

The authors conclude that psilocybin administered with psychotherapeutic support shows a dose-response relationship for antidepressant efficacy, with a curve that reaches a plateau in primary depression and a distinct, lower near-maximal dose pattern in some secondary depression populations. Optimal dose appears to depend on patient characteristics (for example treatment resistance, comorbid anxiety, age, and prior psychedelic exposure). Short-term safety in clinical settings was described as reasonable, with mostly transient and benign adverse events, though some dose-dependent risks were identified. The dose-response estimates provided are intended to inform the design of future clinical trials, particularly those targeting treatment-resistant depression and populations with comorbid diagnoses, while emphasising the need for further RCTs to refine optimal dosing that balances efficacy and tolerability.