Decreases in Suicidality Following Psychedelic Therapy: A Meta-Analysis of Individual Patient Data Across Clinical Trials

Suicide is a major global health problem, with roughly 800,000 deaths annually and substantially higher prevalence of suicidal thoughts and behaviours. Suicidality (including ideation, planning, and attempts) occurs across diagnostic categories, is common in major depressive disorder but also in contexts such as life-threatening illness, and predicts completed suicide. Existing interventions that reduce suicidality—such as cognitive–behavioural therapy, electroconvulsive therapy, selective serotonin reuptake inhibitors, and intravenous ketamine—have important limitations: modest or delayed effects in some cases, restricted accessibility or tolerability, brief duration of benefit (eg, ketamine commonly shows effects that wane within days), or safety concerns with repeated administration. These gaps have motivated interest in novel treatments that might rapidly and durably reduce suicidality. Weissman and colleagues set out to address the absence of a patient-level meta-analysis on psychedelic therapy and suicidality. Specifically, the investigators identified clinical trials that administered classic psychedelics (eg, psilocybin, ayahuasca, DMT, LSD, mescaline, peyote) and included a suicidality measure or suicidality item, requested participant-level suicidality data from study teams, and tested the hypothesis that psychedelic therapy would be associated with significant acute and sustained reductions in suicidality across trials. This approach aimed both to synthesise unpublished suicidality data and to examine safety signals such as acute or post-acute elevations in suicidality.

The study is an individual patient-data meta-analysis reported in line with PRISMA guidelines and registered on PROSPERO. MEDLINE, PsycINFO, and PubMed were searched from inception to 28 March 2020 and updated on 5 November 2020 using broad terms covering classic psychedelics and multiple psychiatric outcomes. Eligible records were clinical trials that administered a classic psychedelic (psilocybin, mescaline, DMT, LSD, ayahuasca, peyote) and included a suicidality measure or suicidality item at baseline and after psychedelic administration. Studies of microdosing, non-classic psychedelics (eg, MDMA), case reports, experimental non-clinical studies, and observational studies were excluded. Titles and abstracts were screened independently by two authors, with full texts similarly reviewed and disagreements resolved with a third author. The study team contacted corresponding authors of eligible trials to obtain suicidality-related participant-level data. When studies contained multiple suicidality measures/items, the investigators used a prespecified hierarchy to choose which to include (full suicidality measures preferred, then measures with more time points, items from primary outcomes, and finally the item most commonly used across included studies). Risk of bias was assessed using the Cochrane risk-of-bias tool version 2 for randomized trials and a modified Newcastle–Ottawa scale for nonrandomized open-label trials; assessments were performed and verified by different authors. For the primary analyses, only participants with baseline suicidality (measure/item score > 0) were included. Changes in suicidality were analysed as standardized mean differences (SMDs) with 95% confidence intervals, comparing grouped scores at successive time points to baseline. Time points with data from at least two studies were pooled; for variable assessment schedules, assessments within 1–2 days of a given time point were included and several extended time windows were created (eg, 1 day = 7–24 hours, 3–4 months). Heterogeneity was quantified with I2 statistics and publication bias explored via funnel plots. To examine safety, the investigators assessed acute elevations (within 6 hours of dosing) and post-acute elevations (first assessment > 6 hours) in suicidality among all participants who received a psychedelic, defining an elevation as any increase (≥ 1 point) from baseline and also reporting frequencies for ≥ 2‑ and ≥ 3‑point increases.

From 9,667 initial records (plus 278 from an update), screening led to eight trials that met the inclusion criterion of reporting a suicidality measure/item; the study team obtained participant-level suicidality data for seven trials and included those in the meta-analysis. The final dataset comprised four randomized controlled trials and three open-label trials. Across included trials, participants represented heterogeneous clinical presentations (treatment‑resistant major depressive disorder, recurrent MDD, AIDS‑related demoralization, and cancer‑related distress). The authors report that overall risk of bias was judged to be low across included trials. Meta-analytic results indicated statistically significant reductions in suicidality at all acute time points examined (80–100 minutes, 140–160 minutes, and 180–240 minutes after administration) and at most post-acute time points. Acute effect sizes were large, with SMDs ranging from −1.48 to −1.72. Post-acute reductions remained large from 1 day through 3–4 months (SMDs reported between −1.50 and −2.36). At 6 months post‑administration the pooled effect size was more modest (SMD = −0.65), described as medium. Reductions reached statistical significance at all reported time points except 7–8 weeks, a time point for which the sample size was especially small (n = 14). Heterogeneity varied by time point: I2 was 0% at 1 day, 1 week, 5 weeks, and 6 months; mild at 2 weeks (I2 = 27.6%); moderate at 80–100 minutes (I2 = 41.0%) and 140–160 minutes (I2 = 45.8%); and marked at 180–240 minutes (I2 = 68.8%), 3 weeks (I2 = 74.3%), 7–8 weeks (I2 = 75.1%), and 3–4 months (I2 = 80.1%). Fail-safe N values (a measure of robustness to unpublished null findings) ranged from 11 to 101 across time points, highest at 1 day and 1 week. Funnel plots were inspected and, despite small sample sizes, appeared grossly symmetrical, suggesting limited evidence of publication bias albeit with low power to detect it. Regarding elevations in suicidality, acute assessments (31 individuals, all from ayahuasca studies, assessed 40–240 minutes post-dose) found 2 individuals (6.5%) with a +1 increase on the HDRS suicidality item at a single acute time point; both returned to a score of 0 at subsequent acute and post-acute assessments. No acute increases of ≥ 2 or ≥ 3 points were observed. Post-acute assessments (168 individuals, 1 day to 26 weeks post-dose) identified 5 individuals (3.0%) with a +1 increase at their first post-acute assessment; most returned to baseline by the next assessment or by end of treatment, and no increases of ≥ 2 or ≥ 3 points were observed. None of the included studies reported suicide-related adverse events attributed to psychedelic administration.

Weissman and colleagues interpret their findings as preliminary evidence that psychedelic therapy is associated with large, rapid reductions in suicidality that may persist for months, and that acute or post‑acute increases in suicidality following administration are rare. The authors note that no suicide‑related adverse events were reported across included trials. They situate these results in the context of existing treatments, observing that psychedelic therapy appears rapid‑acting similar to ketamine but may exhibit longer durability of effect (their pooled results suggest reductions lasting up to 6 months), and that psychedelics in these trials were typically administered only once or twice and have a different safety profile compared with repeated ketamine. The investigators highlight that large effect sizes were observed across diverse clinical samples (treatment‑resistant and recurrent MDD, AIDS‑related demoralization, and cancer distress), suggesting potential transdiagnostic benefit; however, they caution that these findings may not generalise to populations excluded from the trials (notably those with psychotic disorders or mania). The authors discuss candidate mechanisms of change that have been proposed in the literature, including biological pathways (eg, anti‑inflammatory effects, serotonergic modulation, increased brain‑derived neurotrophic factor and neuroplasticity, altered connectivity, and glutamatergic changes) and psychological processes (eg, reduced hopelessness, decreased emotional avoidance, and increased meaning), but note that mechanisms specific to suicidality remain untested. Key limitations acknowledged by the study team include heterogeneity in measurement (multiple suicidality instruments and single‑item measures were used), the small number of studies and sample sizes, the inclusion of heterogeneous clinical presentations, and the fact that most trials were not designed or powered specifically to evaluate suicidality. Many trials excluded individuals at serious suicide risk, several used open‑label designs, and the meta-analysis could not separate effects by psychedelic type, dose number, or intensity of psychological support. Because of these constraints, the authors emphasise that the results are preliminary and call for controlled trials that specifically recruit individuals with suicidality, employ more fine‑grained or standardised suicidality measures (eg, the Columbia–Suicide Severity Rating Scale), and investigate mechanisms and optimal administration/support parameters. They also suggest that pragmatic designs and novel assessment methods (eg, ecological momentary assessment) could be valuable in future work.