Psilocybin-assisted therapy for depression: A systematic review and dose-response meta-analysis of human studies

Depression is a highly prevalent and disabling disorder with substantial non‑response and non‑adherence to conventional antidepressants. Renewed interest has focused on serotonergic (classic) psychedelics such as psilocybin because of promising early clinical signals and a plausible mechanistic overlap with serotonergic antidepressants through 5‑HT2A receptor modulation. Prior randomised trials and a recent meta‑analysis reported antidepressant effects of psilocybin, but the optimal dose for treating primary major depressive disorder (MDD), treatment‑resistant cases and depression secondary to medical illness remains uncertain, and dose‑related risk of adverse events has not been fully characterised. Perez and colleagues set out to define the dose‑response relationship of psilocybin for depressive symptoms and for common adverse events by performing a systematic review and one‑stage dose‑response meta‑analysis of double‑blind, placebo‑controlled randomised trials. The primary objective was to estimate near‑maximum effective doses (ED50 and ED95) expressed as mg/70 kg for reductions in depression scores; secondary aims included dose‑response estimates for anxiety symptoms and for physical and psychological adverse events. The authors emphasise the goal of informing dosage selection for future trials and clarifying safety profiles across different clinical populations.

The investigators conducted a systematic search of Embase, the Cochrane Database of Systematic Reviews and PsycInfo up to February 2023 for double‑blind randomised controlled trials and cross‑over trials that compared fixed doses of psilocybin (any formulation) with placebo in adults aged 18–65 years diagnosed with primary or secondary depression. Two reviewers screened records independently, resolving disagreements by full‑text review. Inclusion required at least one week of follow‑up and diagnostic ascertainment using DSM or SCID instruments; uncontrolled studies, case reports and reviews were excluded. Primary depression was defined as MDD without major comorbid psychiatric disorders (except anxiety), and secondary depression as depressive symptoms occurring alongside a serious medical illness (for example, cancer‑related depression). Data extraction captured baseline and endpoint means and SDs for depressive symptom scales (HDRS, MADRS or other validated instruments); missing SDs were imputed from p‑values or pooled SDs when necessary. Anxiety change (Beck Anxiety Inventory) and adverse events (physical discomfort, blood pressure rise, tachycardia, nausea/vomiting, headache/migraine; psychological distress, prolonged psychosis/HPPD) were also extracted. Risk of bias was assessed independently with the Cochrane RoB 2 tool across randomisation, allocation concealment, blinding, attrition and reporting; studies were classified as low, moderate or high risk according to predefined criteria. For the dose‑response synthesis, doses were converted to total mg/70 kg. The standardised mean difference (Cohen’s d) was the effect metric for continuous outcomes and log risk ratios for binary adverse events. A one‑stage random‑effects dose‑response meta‑analysis using restricted cubic splines (knots at the 25th, 50th and 75th percentiles) was implemented with the 'doresmeta' and 'metafor' packages in R v4.2.2. ED50 and ED95 estimates were derived from fitted curves; heterogeneity was quantified with the variance partition coefficient (VPC), a multivariate extension of I2. Sensitivity analyses included exclusion of studies at non‑low risk of bias, leave‑one‑out analyses and subgroup analyses by primary versus secondary depression when data permitted.

The search returned 5,196 records; after screening and applying eligibility criteria, seven RCTs (total N = 489) were included. Four trials (N = 366) enrolled patients with primary depression (MDD), including one trial that specifically recruited treatment‑resistant depression; three trials (N = 123) enrolled participants with depression or anxiety secondary to cancer. Doses in the included studies ranged from 1.5 to 50 mg/70 kg; studies administering a single fixed dose had a mean of 16.5 mg/70 kg, while studies with two doses averaged 31.5 mg/70 kg. Participants were majority female (53.2%); mean ages reported were 44.1 ± 12.8 overall, 39.2 ± 11.6 for the primary subgroup and 56.3 ± 5.3 for the secondary subgroup. Prior lifetime hallucinogen use was reported in 26.3% of participants in pooled data. Trials delivered psilocybin orally in capsules in an individual psychedelic‑assisted psychotherapy setting with non‑directive supportive psychological support. For depressive symptoms across all studies (n = 7; N = 489) there was a significant dose‑response relationship (p < 0.0001) with substantial heterogeneity (I2 ≈ 85%). The pooled dose‑response curve yielded an ED50 of 10.13 mg/70 kg and an ED95 of 41.14 mg/70 kg (95% CI 26.37–47.80), indicating half‑maximal and near‑maximal effect levels respectively. Sensitivity analyses excluding studies at non‑low overall risk produced a similar association (p < 0.0001) with an ED95 of 36.08 mg/70 kg. Excluding the trial that enrolled treatment‑resistant patients lowered the ED50 and ED95 to 8.23 and 24.05 mg/70 kg respectively. Leave‑one‑out analyses showed ED50s ranging from 8 to 14 mg/70 kg and ED95s from 24 to 45 mg/70 kg. Subgroup analyses showed a significant dose‑response for primary depression (n = 4; N = 366; p < 0.0001; I2 ≈ 80%), with the curve plateauing at an ED95 of 24.68 mg/70 kg. The secondary depression subgroup (n = 3; N = 123) did not reach statistical significance for dose‑response (p = 0.07; I2 ≈ 80%); visual inspection suggested a bell‑shaped relationship with a lower near‑maximal point (reported ED95 ≈ 8.92 mg/70 kg), implying higher doses delivered no additional benefit and possibly reduced effect for that population. Anxiety outcomes (n = 6 RCTs; N = 258) also demonstrated a significant dose‑response (p < 0.0001) with heterogeneity (I2 ≈ 85%). For combined populations the ED50 for anxiety was 7.58 mg/70 kg and the ED95 22.78 mg/70 kg (95% CI 15.82–47.2). Subgroup results indicated differing ED50/ED95 values for primary versus secondary depression (primary ED95 24.68 mg/70 kg, ED50 8.86 mg/70 kg; secondary ED50 4.08 mg/70 kg), again with considerable heterogeneity. For adverse events, dose‑associations were identified for several outcomes. Physical discomfort showed a significant positive association with dose (p = 0.023), estimated as a 2.35% relative increase in risk per additional 1 mg/70 kg of psilocybin. Blood pressure elevation was also dose‑associated (p = 0.042) with a ~1.04% increased risk per 1 mg/70 kg. Tachycardia showed a non‑significant positive trend (p = 0.09). Nausea/vomiting and headache/migraine followed bell‑shaped dose patterns with significant associations (p < 0.001 and p < 0.05) and small relative risk increases per mg. A combined psychiatric adverse‑event outcome (prolonged psychosis >24 h and HPPD) was significantly associated with dose (p = 0.001) with an estimated 2.51% relative increase per 1 mg/70 kg, but such events were rare (12 cases among ~465 subjects) and outcome definitions varied across trials. The authors emphasised variable outcome thresholds and inconsistent adverse‑event definitions between studies, which constrain interpretation.

Perez and colleagues interpret their findings as evidence for a dose‑response relationship between psilocybin and reduction in depressive and anxiety symptoms, with differing shapes across clinical populations. They report a plateauing curve for primary MDD and a bell‑shaped curve for depression secondary to medical illness, concluding that optimal dosing depends on the target population. In pooled analyses the ED50 approximated 8–10 mg/70 kg and ED95 values varied widely depending on sensitivity and subgroup analyses (reported between ~24 and 41 mg/70 kg), with treatment‑resistant patients tending to require higher doses. The authors note that anxiety‑prone or comorbid populations may reach maximal therapeutic benefit at lower doses and that higher doses can exacerbate acute anxiety or psychosis‑like reactions related to ego‑dissolution phenomena. They situate the results relative to previous syntheses that suggested a putative optimal dose near 30–35 mg/70 kg, noting methodological differences such as use of dose‑response modelling here and inclusion of more trials. The discussion highlights that large effect sizes were observed (standardised mean differences around 1.0–1.6), but acknowledges that effect magnitude may be influenced by the intensity of the psychedelic experience and supportive psychotherapeutic context rather than pharmacology alone. Tolerability findings are described as broadly favourable in the short term: most adverse events were transient and mild–moderate (headache, nausea), though dose‑related increases in somatic events and rare prolonged psychological events were observed. Key limitations are acknowledged: a small number of RCTs, heterogeneous populations (especially within the secondary‑depression group where diagnoses varied), variable follow‑up durations, inconsistent definitions and thresholds for adverse events across trials, limited data on prior hallucinogen exposure, and only one RCT in treatment‑resistant depression or comparing psilocybin with a standard antidepressant. The authors also raise the unresolved issue of adequate blinding in psychedelic trials. For future research they recommend standardised reporting of psychotherapeutic procedures (set and setting), measurement of subjective psychedelic intensity, clearer adverse‑event definitions, and trials designed to define optimal dosing patterns and the role of non‑pharmacological mediators of response.

The authors conclude that psilocybin administered with psychotherapeutic support demonstrates a dose‑response relationship in treating depressive symptoms, with the near‑maximal effective dose differing by patient group. Short‑term safety in trial settings appears acceptable with mostly transient, benign adverse events, but dose‑dependent somatic and rare psychological events were observed. These dose‑response estimates are intended to inform dose selection in future RCTs, particularly for treatment‑resistant depression and for populations with comorbid conditions, while underscoring the need for further double‑blind trials with standardised psychotherapeutic approaches and longer‑term safety assessment.