Naturalistic psychedelic use and changes in depressive symptoms

Growing evidence from recent randomised trials suggests that single‑dose psilocybin administered with psychotherapy can produce rapid and sustained antidepressant effects in clinical populations, and earlier studies have reported a relatively low risk of worsening depressive symptoms and other adverse events. At the same time, naturalistic use of psychedelics has increased in countries such as the United States, including among people with depressive symptoms or disorders, raising the need to understand effects outside of controlled research and clinical settings. Several longitudinal observational studies of naturalistic psychedelic use have reported reductions in depressive symptoms, but those studies often recruited self-identified advocates or planned users and lacked unexposed comparison groups, which may introduce expectancy or selection biases and limit generalisability. Simonsson and colleagues therefore used a longitudinal observational cohort design to explore associations between naturalistic psychedelic use and changes in depressive symptoms in a large sample of US residents aged 18–50. Given prior indications that contextual factors and the quality of the acute psychedelic experience may shape outcomes, the study included additional exploratory analyses examining psychedelic use in a constructed 'risk context', the severity of challenging psychedelic experiences, and how these relate to changes in depressive symptoms over a two‑month follow‑up period.

The investigators recruited US residents aged 18–50 using purposive sampling via the Prolific Academic platform during two waves (June–September 2023 and June 2024). The study description omitted mention of psychedelics to reduce selection bias. Participants provided informed consent, completed a baseline survey (T1), and were invited to a follow‑up survey approximately two months later (T2). Institutional review procedures determined the study to be exempt from review. Participants received modest payments for completing T1 and T2. Depressive symptoms were assessed at T1 and T2 using the 9‑item self‑report Montgomery Åsberg Depression Rating Scale (MADRS‑9). At T1 participants reported demographics (age, gender identity, education), degree of religiosity, political affiliation, past two‑month psychedelic use, and current or past diagnosis of depressive disorders. At T2 participants reported past two‑month use of a range of substances (including psychedelics, alcohol, nicotine, cannabis, MDMA, stimulants, opioids, benzodiazepines/barbiturates, inhalants). Those reporting past two‑month psychedelic use at T2 retrospectively described their most intense experience during the period, completed the Challenging Experiences Questionnaire, and provided dose and contextual data. The team constructed a three‑level 'risk‑context' variable based on three context indicators previously linked to adverse outcomes: a major life event prior to the experience, a negative mindset prior to the experience, and absence of psychological support during the experience (0 = no psychedelic use, 1 = psychedelic use with none of these three risk factors, 2 = psychedelic use with at least one of the three). An extended version incorporated three additional context items (insufficient preparation, disagreeable physical environment, dose too large) to form a six‑factor composite used in sensitivity analyses. Outcomes included continuous change in MADRS‑9 scores and binary indicators of minimal clinically important differences (MCIDs), defined here as a standardised mean difference ≥ 0.24 for worsening or ≤ -0.24 for improvement. Analyses used linear regressions for continuous outcomes and logistic regressions for MCIDs, controlling for age, gender identity, education, religiosity, political affiliation, past two‑month psychedelic use at T1, past two‑month use of specified substances at T2, and survey year. Models restricted to participants who used psychedelics during the study additionally controlled for dose size. Continuous predictors were standardised (z‑scored). To address missing T2 data (no missingness at T1), the researchers implemented Multivariate Imputation by Chained Equations with random forest imputations to generate 20 imputed datasets; results were pooled using Rubin’s rules. A two‑sided p < .05 threshold was used for significance. Sensitivity analyses included complete‑case analyses and models using the extended 'risk‑context' variable.

Of 21,990 participants who completed T1, 12,345 completed T2 (56.1% retention; 43.9% attrition). Among T2 completers, 505 participants (4.1%) reported psychedelic use during the study period. Users differed from non‑users on age, gender, and psychiatric history; among those who used psychedelics during the study period, 36.4% reported that their most intense experience occurred in a 'risk context'. Further sample characteristics and clinical symptom statistics were reported in tabular form. In covariate‑adjusted linear regression models using the imputed data, psychedelic use during the study period was modestly associated with increases in depressive symptoms (standardised coefficient β = 0.12, p = .019). Disaggregating by context, psychedelic use in a 'risk context' showed a moderate association with increases in depressive symptoms relative to no psychedelic use (β = 0.30, p < .001) and relative to psychedelic use not in a 'risk context' (β = 0.27, p = .004). Sensitivity analyses using complete (non‑imputed) data yielded largely unchanged coefficient directions and significance. Among participants who used psychedelics during the study period, psychedelic use in a 'risk context' was strongly associated with reporting a more challenging psychedelic experience (β = 0.59, p < .001). Severity of a challenging psychedelic experience was itself modestly associated with increases in depressive symptoms (β = 0.16, p = .007). In a model that included both risk context and challenging experience simultaneously, the challenging experience remained modestly associated with depressive symptom change (β = 0.14, p = .031) while the direct association for risk context was attenuated and no longer statistically significant (β = 0.14, p = .224), suggesting that challenging experience severity accounted for the association between risk context and symptom worsening. The authors report that analyses of MCIDs are presented in the supplement and that sensitivity analyses using the extended 'risk‑context' variable produced broadly similar results across models.

The study found that naturalistic psychedelic use was associated with increases in depressive symptoms only when use occurred in a constructed 'risk context', and that this association appeared to be explained by the occurrence of more challenging psychedelic experiences. Simonsson and colleagues interpret these findings as indicating that naturalistic psychedelic use is not uniformly therapeutic and may be associated with worsened depressive symptoms under particular contextual circumstances. These results contrast with prior longitudinal studies that reported reductions in depressive symptoms following naturalistic psychedelic use; the authors suggest that differences in sample characteristics, recruitment approaches, and the contexts in which psychedelics were used may account for discrepant findings. The positive association between severity of challenging experiences and symptom worsening also differs from some earlier reports and leads the investigators to propose that the relationship may be nonlinear or contingent on whether a challenging experience is transient and resolved versus persistent. The authors acknowledge several strengths, including the large sample size, the inclusion of an unexposed comparison group, and recruitment materials that did not mention psychedelics. They also list important limitations: purposive rather than representative sampling and recruitment from an online panel limit generalisability; relatively high attrition (43.9%) and reliance on multiple imputation may introduce bias; contextual data on psychedelic use were limited and several potentially relevant variables (such as intention for use) were not measured; individual context factors were combined rather than analysed separately, which may obscure interactions; all measures were self‑reported, introducing potential recall or social‑desirability biases; and the exploratory, observational design precludes causal inferences and no corrections for multiple comparisons were applied. The investigators therefore recommend cautious interpretation and call for further epidemiological research to clarify for whom and under what circumstances naturalistic psychedelic use may be associated with worsening depressive symptoms.