Adverse Events in Studies of Classic Psychedelics: A Systematic Review and Meta-Analysis

Hinkle and colleagues frame the importance of understanding harms as clinical research on classic serotonergic psychedelics (psilocybin, LSD, DMT, 5-MeO-DMT, and mescaline) expands. Earlier estimates of adverse outcomes derive from disparate sources (therapist surveys, inpatient experimental regimens, emergency-department reports) that are not readily generalisable to contemporary, safety-guided research settings, leaving uncertainty about the frequency and nature of medically or psychiatrically significant adverse events (AEs). This systematic review and meta-analysis aimed to aggregate peer-reviewed clinical and research studies that administered psilocybin, LSD, DMT, or 5-MeO-DMT, and to quantify AEs with particular attention to ICH-defined serious adverse events (SAEs) and nonserious AEs (NSAEs) that required medical or psychiatric intervention. The investigators focused on events across acute, early (acute to 48 hours), and delayed (>48 hours) timescales, and stratified findings by participant population (healthy, outpatient with neuropsychiatric disorders, and inpatient) to provide benchmark estimates useful for informed consent and future trial design.

The review protocol was preregistered on PROSPERO and followed PRISMA guidance. Hinkle and colleagues searched Scopus, MEDLINE, PsycINFO, and Web of Science from inception through 8 February 2024, and reviewed ClinicalTrials.gov AE data where available. Two independent reviewers screened abstracts and full texts; disagreements were resolved by discussion and adjudication by additional authors. Inclusion was limited to peer-reviewed human studies involving monitored administration of LSD, psilocybin, DMT, or 5-MeO-DMT. Full inclusion/exclusion details were provided in supplementary materials (not reproduced here). Outcomes combined an exploratory capture of all reported AEs with confirmatory capture of prespecified events of interest. The team used ICH terminology to distinguish seriousness from severity, and enumerated SAEs (eg, death, life‑threatening events, hospitalisation) separately from NSAEs that nevertheless required medical or psychiatric intervention (eg, physician evaluation, medication changes, intravenous therapy). Prespecified events included psychotic disorder, manic symptoms, suicidal ideation or behaviour, cardiovascular events, and hallucinogen persisting perception disorder. AEs were stratified by population category (healthy participants, outpatients with neuropsychiatric disorders, and inpatients) and by timing (acute, early/subacute [pooled], and delayed). Dose restrictions were applied to focus on psychedelic dosing ranges (examples: LSD ≥30 μg; oral psilocybin >3 mg/70 kg or any intravenous psilocybin; intravenous DMT ≥0.2 mg/kg). For meta-analysis, the authors included only AEs reported in at least three independent studies. Statistical analysis used random-intercept logistic regression models (R package 'metawes' was cited), producing pooled proportions with 95% confidence intervals and I2 heterogeneity estimates; three-level models accounted for within-study clustering when studies reported multiple participant groups. Risk-of-bias assessment adapted the Newcastle-Ottawa Scale to focus on ascertainment of AEs, including whether all events were reported and whether a severity framework was described.

The search yielded 3739 unique records; after screening and merging duplicate datasets, 214 unique studies met inclusion criteria. Of these, 114 studies (53.3%) reported analyzable AE data, representing 155 participant groups and a total of 3504 participants. After stratification there were 95 groups of healthy participants (n = 1726), 39 outpatient groups (n = 934), and 21 inpatient groups (n = 844); all inpatient studies dated from 1951 to 1972. Two small 5-MeO-DMT studies (n = 22 healthy; n = 16 with treatment‑resistant depression) reported no SAEs or medically significant NSAEs. Serious adverse events: Early (acute or early/subacute) SAEs were not reported among healthy participants given psilocybin (n = 659), LSD (n = 372), or DMT (n = 139). One early SAE occurred among 128 outpatients given LSD (0.8%): an episode of transient anxiety and delusions requiring antipsychotic treatment and overnight observation, with resolution by morning. In a small DMT subgroup, 1 of 6 participants with depression experienced a clinically significant hypotensive event reported as an SAE (16.7%). Psychiatric inpatients receiving LSD experienced early SAEs at a rate of 11 of 517 (2.1%), including psychotic episodes terminated with electroconvulsive therapy and seizures; some events followed very large historical LSD dosing (reported up to 2000 μg in older studies). Delayed SAEs: Median follow-up across studies was about 12 weeks (range 1–60). No delayed SAEs were reported in healthy participants who had received LSD (n = 258), psilocybin (n = 319), or DMT (n = 109). Among outpatients given LSD, 8 of 155 (5.2%) experienced delayed SAEs during follow-up, though causal attribution to LSD was uncertain and several events (eg, surgeries, pregnancy loss, deaths from preexisting illness) were judged unrelated. In contemporary high‑dose psilocybin outpatient trials, 23 of 584 participants (3.9%) experienced delayed SAEs over 12 weeks; within this subset, 11 participants with treatment‑resistant depression accounted for 17 SAEs including suicidal ideation (5 events), suicidal behaviour (3 events), nonsuicidal self‑injury (4 events), worsening depression and related hospitalisation. Some cases of post‑treatment psychosis coincided with relapse to substance use. Among historical inpatient LSD studies, 9 of 611 participants (1.5%) experienced delayed SAEs, including suicides and deaths reported in long post‑discharge intervals in older datasets. Nonserious AEs requiring medical intervention and suicidality: Medically significant NSAEs were generally rare. No published NSAEs requiring intervention for suicidal ideation or behaviour were identified among healthy participants. In outpatient psilocybin studies, 3 of 553 participants (0.5%) experienced worsening suicidal ideation, with one episode beginning the day of dosing. A single suicide death occurred in a participant with life‑threatening cancer who received a very low, effectively nonpsychoactive psilocybin dose (1 mg/70 kg); that event was not included in pooled analyses because the participant did not receive a psychoactive dose. Common adverse events and heterogeneity: Meta-analytic estimates were generated for AEs reported by at least three independent studies; forest plots were produced for headache, anxiety, nausea, fatigue, and dizziness. Most common AEs showed at least moderate heterogeneity (I2 ≥50%). Stratified analyses for psilocybin and LSD were performed to explore effects of preexisting neuropsychiatric disorders. The authors presented these data descriptively and did not perform formal hypothesis tests between drugs. Risk of bias and reporting quality: AE ascertainment and reporting quality varied substantially. Only 16 of 68 studies published since 2005 (23.5%) described a systematic approach to AE assessment, a proportion not statistically different from earlier studies (17% pre‑2005). Contemporary trials did, however, more frequently document causality attribution (29% vs 0%), severity frameworks (28% vs 2.1%), duration of follow-up (84% vs 35%), and adequacy of follow-up (79% vs 22%) compared with earlier studies. Overall, 114 of 214 included studies (53.3%) reported any AE data, and many contemporary studies reported only suspected adverse drug reactions rather than all AEs, suggesting potential underdetection or selective reporting.

Hinkle and colleagues conclude that, in monitored clinical and research environments, medically or psychiatrically significant AEs related to classic psychedelic administration are relatively uncommon but can be serious when they occur. No sustained psychotic disorders or cases meeting criteria for hallucinogen persisting perception disorder were identified among participants without preexisting psychotic illness, and deaths by suicide following high‑dose psychedelic administration were not observed in the analysed trials. The investigators note that suicidality requiring intervention occurred only among participants with preexisting depressive disorders in the reviewed literature. The authors position these findings against earlier, less comparable data sources and highlight that certain catastrophic events reported in recreational or nonclinical contexts have not been documented in contemporary trial participants. Nevertheless, they emphasise substantial heterogeneity in AE monitoring and reporting practices across studies, and warn that incomplete ascertainment and selective reporting (for example, reporting only suspected adverse drug reactions) limit confidence in incidence estimates. To address this, the paper aligns with recent regulatory guidance encouraging validated scales, dedicated AE monitoring, narrative descriptions of events, and standardised reporting of both expected psychedelic phenomena and medically significant outcomes. Key limitations acknowledged by the investigators include the inability to estimate AE risks relative to placebo because many included studies were uncontrolled, limited insight into dose–response relationships for common AEs, and methodological heterogeneity that complicates population comparisons. They also underscore selection biases in contemporary trials (stringent screening, exclusion of those with personal or family histories of psychosis or mania, and recruitment of participants with prior psychedelic experience), as well as underrepresentation of older adults, all of which constrain generalisability. The authors recommend more systematic, transparent pharmacovigilance and standardised templates for AE detection and characterisation in future psychedelic research to provide more precise risk estimates and better inform informed‑consent discussions.

The review and meta-analysis indicate that medically or psychiatrically significant adverse events following high‑dose classic psychedelic administration in monitored research settings are relatively rare but can be serious. Hinkle and colleagues stress that the accuracy of these estimates depends on study-level AE detection and reporting practices, which were frequently incomplete or inconsistent. They conclude that improved, standardised adverse‑event monitoring and transparent reporting are required to refine risk estimates and to inform participants and the public; the results should not be generalised to unsupervised, nonclinical use.