Reconsidering evidence for psychedelic-induced psychosis: an overview of reviews, a systematic review, and meta-analysis of human studies

Interest in the therapeutic potential of serotonergic psychedelics is rapidly expanding, driven by promising results in treatment-resistant conditions such as depression, substance use disorders, anxiety and PTSD. However, patients with schizophrenia and bipolar disorder have generally been excluded from psychedelic research because of concerns that these drugs might trigger acute or long-lasting psychotic symptoms. Earlier theories, such as the serotonin-2A overactivity hypothesis, and historical clinical use of LSD in the 1950s–60s contributed to debate about whether serotonergic psychedelics can precipitate or worsen psychosis; at the same time, more recent work points to other substances (for example cannabis and stimulants) as stronger drivers of psychosis risk than classic psychedelics. Sabé and colleagues set out to clarify the evidence linking serotonergic psychedelics (LSD, psilocybin, mescaline, DMT, and MDMA) to incident psychosis, exacerbation of pre-existing psychotic disorders, and subsequent transition to schizophrenia. The study combined an overview of existing reviews with a systematic review and meta-analysis, aiming to estimate the incidence of long-lasting (>48 hours) psychotic symptoms after psychedelic use across several populations (general/indigenous populations, healthy volunteers, patients with depression, and people with schizophrenia) and to quantify conversion rates from substance-induced psychotic disorder (SIPD) to schizophrenia.

The researchers conducted a pre-registered (PROSPERO CRD42023399591) systematic review and meta-analysis following PRISMA guidance, with an updated search through November 2023. Electronic searches were run in PubMed, Embase and PsycINFO, supplemented by trial registries, title/abstract screening using Rayyan, independent dual screening and extraction, and snowball searches of reference lists. Only human studies of the specified serotonergic psychedelics were eligible; case reports and case series were excluded. Cannabis, salvia divinorum and PCP were excluded because of distinct pharmacology and high established psychosis risk. A broad range of study types was considered to capture the heterogeneous evidence base: systematic reviews, nonsystematic reviews, guidelines, meta-analyses, randomized controlled trials (RCTs), clinical trials, case-control studies, uncontrolled trials (UCTs), and prospective/retrospective cohorts and population surveys. The investigators applied a hierarchy of evidence to present findings (UCTs, cohort, case-control, RCTs) and used standard tools for quality assessment: AMSTAR-2 for systematic reviews, SANRA for narrative reviews, RoB 2 for RCTs and ROBINS-I for non-randomized studies. The Oxford Centre for Evidence-Based Medicine (OCEBM) framework was used to contextualise evidence levels. For meta-analysis of incidence proportions the team used a logit transformation to stabilise variance and a random-effects model to allow for between-study heterogeneity. Heterogeneity was quantified with the I2 statistic and the Q test, with predefined thresholds for low to considerable heterogeneity. Analyses were implemented in R (version 4.3.1) using the metafor package. Where possible the authors separated analyses by population subgroup (healthy, depressed, schizophrenia) and examined sensitivity analyses restricted to low risk-of-bias studies.

The search identified 2,170 records; 223 full texts were assessed and 131 publications from 1947–2023 were included. These comprised 14 systematic reviews, 20 nonsystematic reviews and 96 individual empirical studies: 35 RCTs, 10 case-control studies, 30 uncontrolled trials and 22 cohort studies. Uncontrolled trials (UCTs): In healthy volunteers, older and recent LSD and DMT UCTs reported occasional session-confined paranoid symptoms or dose-dependent psychotic-like effects (visual hallucinations, anxiety). Open-label psilocybin work noted transient paranoia or ideas of reference during sessions in up to 22% of patients, but these were typically acute. Eleven historical UCTs administered LSD to people with schizophrenia (oral doses 30–500 μg; some IV studies), with most reporting no long-lasting psychotic reactions; nevertheless, two intravenous studies observed more marked acute reactions and one reported 3 of 65 patients with long-lasting psychotic reactions lasting up to a week. Studies in children and in autism spectrum disorder reported short-term behavioural or speech improvements in some cases, but diagnostic and methodological concerns limit interpretation. Overall, UCTs were often old and carried critical risk of bias and low OCEBM levels. Cohort studies: Population and cohort analyses produced heterogeneous findings. A Finnish hospital-discharge register analysis reported a cumulative 24% risk of schizophrenia-spectrum diagnosis after an initial hallucinogen-induced psychosis. By contrast, ethnographic/epidemiological studies among frequent ayahuasca or peyote users reported extremely low rates of persistent psychosis (on the order of 1 in tens of thousands). Some cohort evidence suggested lifetime psychedelic use associated with lower odds of recent inpatient mental health treatment (adjusted OR 0.7, p = 0.01). Many cohort studies were low quality and rated level 4 on OCEBM. Case-control studies: Historical comparisons suggested that LSD-induced psychoses differ phenomenologically from schizophrenia (for example more visual than auditory hallucinations). One case-control study of MDMA users noted less blunted affect and more hostility; rare cases with severe aggression have been reported. Randomized controlled trials (RCTs): Recent RCTs in healthy volunteers and patients with depression or PTSD—generally excluding anyone with personal or familial psychosis risk—reported no enduring psychotic reactions. Across MDMA trials in PTSD, short-term anxiety reactions were common and dose-related, but persistent psychosis was not observed. Older RCTs in schizophrenia were few and methodologically limited; one 1966 trial combining high-dose LSD with thioridazine reported improved mood and social behaviour but carried critical biases. Other findings: MRI studies of regular psychedelic users did not reveal specific brain harm in the reviewed work. Hallucinogen persisting perception disorder (HPPD) was not associated with worsening psychotic symptoms in people with schizophrenia. Guidelines on hallucinogen research were identified; most contemporary contraindications for psychotic disorders trace back to case reports. Meta-analytic results: The pooled meta-analysis incorporated studies reporting lifetime incidence or session-related incidence of psychedelic-induced psychosis. Across pooled datasets (n = 7,931 sessions from 10 experimental/cohort studies) the combined incidence of long-lasting (>48 h) psychedelic-induced psychosis was 0.4% (95% CI 0.1–1.0) with substantial heterogeneity (I2 = 75%). Sensitivity analysis restricted to two low-risk-of-bias studies that excluded participants with psychosis risk yielded a similar pooled incidence of 0.4% (95% CI 0.1–1.6) with no heterogeneity (I2 = 0%; n = 490). When limited to healthy participants the incidence fell to 0.2% (95% CI 0.1–0.3; I2 = 0%). Incidence among patients with depression was 0.4% (95% CI 0.1–2.8). Two older studies in people with schizophrenia produced a higher pooled incidence of 3.8% (95% CI 1.6–8.9; I2 = 0%), but the authors emphasise these studies are dated and of lower quality. Additional pooled estimates: Lifetime occurrence of prolonged psychosis across five studies (n = 622) was 2.0% (95% CI 0.5–7.7; I2 = 0%); excluding lower-quality schizophrenia studies reduced this to 0.6% (95% CI 0.2–1.8). Two indigenous population studies (n = 123,800) yielded a pooled incidence of 0.002% (95% CI 0–0.006). Conversion from SIPD to schizophrenia was reported in three follow-up studies (n = 353), with a pooled conversion rate of 13.1% (95% CI 9.4–17.9) over a mean follow-up of 10.5 years (I2 = 24%).

Sabé and colleagues interpret their findings as indicating that serotoninergic psychedelics rarely induce long-lasting psychotic syndromes in general populations and in contemporary controlled trials that exclude people at risk for psychosis. Population-based and indigenous studies reported very low incidence rates, and recent RCTs in selected participants found enduring psychotic effects in under 2% of cases. At the same time, the investigators note that RCTs systematically excluded individuals with personal or family histories of psychosis, so these trials do not resolve the risk in the groups typically omitted from psychedelic studies. The authors highlight that psychedelic-induced psychotic reactions were more commonly reported in uncontrolled, recreational, or illicit-use contexts, or when novel psychoactive substances were implicated. Historical uncontrolled studies in people with schizophrenia reported some short-term exacerbations but mostly no long-lasting psychotic reactions; nevertheless, those studies are old, scarce and at high risk of bias. Importantly, pooled data show that among people who develop SIPD, roughly 13% later receive a diagnosis of schizophrenia—an outcome the authors underscore as clinically meaningful. Several limitations are emphasised. The overall evidence base is heterogeneous and generally low quality; only a small subset of studies were suitable for meta-analysis. Diagnostic criteria and study methods have evolved over the 70-year literature span, complicating comparisons. Potential confounders such as prior substance exposure, drug purity, and unsupervised settings are often unreported. The definition of psychedelic-induced psychosis is not uniform across studies, and blinding in psychedelic RCTs is problematic because of the distinctive subjective effects of these drugs. These factors limit causal inference and the generalisability of pooled incidence estimates. Implications discussed by the authors include the possibility of reconsidering the categorical exclusion of all people with schizophrenia from psychedelic research, particularly for stable patients with predominant negative or depressive symptoms and ongoing antipsychotic treatment. Nonetheless, they recommend a conservative, stepwise approach: further high-quality, controlled cohort and randomised studies are required, along with development of predictive models for SIPD and transition to schizophrenia. Practical concerns noted include dose selection, potential interactions with antipsychotics (many of which antagonise 5-HT2A receptors), and the importance of supervised clinical settings and substance purity to mitigate risk.

The authors conclude that incidence of serotonergic psychedelic-induced psychosis in the general population appears to be relatively low, but that among individuals who develop SIPD there is a notable risk of later transition to schizophrenia (pooled conversion ~13%). Given the predominance of low-quality evidence and the small number of robust studies, definitive conclusions cannot yet be drawn. Sabé and colleagues call for higher-quality research to clarify risk–benefit ratios, to develop prediction models for SIPD and conversion to schizophrenia, and to determine whether the current categorical exclusion of people with schizophrenia from psychedelic trials is justified. Meanwhile, they recommend a conservative approach to including this population in research until further data are available.