A Meta-Analysis of Placebo-Controlled Trials of Psychedelic-Assisted Therapy

The authors conducted a systematic review following PRISMA guidelines, searching PsycInfo, ERIC, Medline, Academic Search Premiere and CINAHL for trials published from January 1994 to March 2019. Inclusion criteria were: peer-reviewed original research, randomized placebo-controlled clinical trials, and trials assessing MDMA, psilocybin, ayahuasca, DMT (including 5‑MeO‑DMT), or LSD for diagnosed psychiatric conditions listed in the DSM‑IV or DSM‑V. Non‑English reports were excluded. Search terms combined substance names with clinical focus terms. Study selection proceeded from 931 records to a final set of nine studies after title, abstract and full‑text screening; the team contacted corresponding authors to identify additional studies and excluded one paper during extraction for lack of usable data. For cross‑over trials, only pre‑cross‑over data were used because cross‑over removes the placebo comparison. A senior investigator checked extracted data and disagreements were resolved by discussion. Data extraction recorded diagnosis, primary outcome measure, sample size at the primary endpoint, study design, drug and dose, number of therapy‑only and dosing sessions before the primary endpoint, control type, percentage female, mean age, and percentage White participants. Risk of bias was assessed using Cochrane Handbook methods. Effect sizes were computed as standardized mean differences transformed to Hedges g to correct small‑sample bias; the authors reference conventional benchmarks (g ≈ 0.2 small, 0.5 medium, 0.8 large, 1.2 very large). A random‑effects model was chosen a priori and analyses were performed using Meta‑Essentials. Moderator tests included six continuous variables (percent men/women, percent White, mean age, number of dosing sessions, number of psychotherapy sessions, and sample size) and two categorical variables (inert placebo versus other controls; low‑dose psychedelic versus other controls).

Nine randomized, placebo‑controlled trials met inclusion, comprising a total of 211 participants. Five trials tested MDMA, two tested psilocybin, one tested ayahuasca and one tested LSD. Where race was reported (seven studies), 85% of participants identified as White; overall 53% identified as women. The mean number of dosing sessions before the primary endpoint was 1.9 (SD = 0.8), range 1–3. Control conditions varied: five trials used very low doses of the psychedelic as the control, three used inert placebo, and one used an active placebo (niacin). Three trials retained a placebo control through a longer‑term follow‑up assessment. The pooled between‑groups effect at the primary endpoint was Hedges g = 1.21, which was statistically significant (z = 9.48, p < .001). Tests indicated very low heterogeneity: Qw(8) = 5.61, p = .69, and I2 could not be calculated because of a negative numerator, consistent with minimal between‑study variance beyond sampling error. Rosenthal's failsafe N was 193, suggesting tolerance to additional null studies, and the funnel plot was reported as not indicative of publication bias. Subgroup analyses comparing classical psychedelics (n = 4 studies) with MDMA (n = 5 studies) found no significant difference: classical psychedelics Hedges g = 1.20, MDMA Hedges g = 1.22, z = .08, p = .94. Pre‑planned moderator analyses found no statistically significant moderators: sample size (p = .85), number of dosing sessions (p = .97), number of psychotherapy sessions (p = .64), percent White (p = .60), percent women (p = .61), mean age (p = .93), inert placebo versus other controls (p = .81) and low‑dose psychedelic versus other controls (p = .82). Three studies provided placebo‑controlled long‑term follow‑up data (follow‑ups at approximately seven weeks, two months and six months after the primary endpoint). The average effect size at the primary endpoint for these three studies was g = 1.47 and at follow‑up was g = 1.36, a 7.5% reduction, indicating largely sustained effects over the available follow‑up intervals. Risk‑of‑bias assessment noted that all studies were preregistered and had low attrition, but reporting of randomization and allocation concealment was often incomplete; five studies were judged high or unclear risk for blinding of patients, staff or clinicians and three studies were high risk for blinding of outcome assessment.

The investigators interpret their findings as supporting efficacy of psychedelic‑assisted therapy across four clinical conditions represented in the included trials: PTSD, anxiety/depression associated with life‑threatening illness, unipolar depression and social anxiety in autistic adults. The overall pooled effect (Hedges g = 1.21) is described as very large and, using probabilistic language, corresponds to an approximately 80% chance that a randomly selected patient receiving psychedelic‑assisted therapy will do better than a randomly selected patient receiving placebo. The authors highlight that these effects were observed after only 1–3 dosing sessions combined with psychotherapeutic support, and they consider this suggestive of a mechanism in which benefits extend beyond short‑term pharmacological action. Strengths noted include low attrition, absence of serious adverse events reported in the trials, and adherence to preregistered primary outcomes. The discussion also emphasises important limitations: small total sample size and small number of trials, limited diversity with predominance of White participants, heterogeneity of target conditions across trials, and methodological shortcomings in reporting and blinding. The authors caution that effect sizes in early, small trials can decline as larger trials are conducted and note that unpublished or later phase data (for example combined Phase II MDMA trials) have reported smaller pooled effects than found here, which could reflect differences in effect‑size calculation or publication/selection effects. Methodological challenges receive particular attention. Because psychedelics often produce recognisable subjective effects, effective blinding is difficult; several included trials used low‑dose psychedelic controls or active mimetic controls to address this, but the authors recommend routinely assessing blind success by asking staff, assessors and participants to guess allocation. They also call for clearer reporting of randomization and allocation concealment, better quantification and balancing of concurrent psychotherapeutic contact (given the presumptive importance of set and setting), and trial designs that can isolate the contributions of the pharmacological agent versus psychotherapy (for example by randomising psychotherapy amount or type while holding dosing constant). Finally, the authors urge larger, more diverse trials to permit moderator and mediator analyses, and they recommend future research on durability of effects (ideally maintaining placebo control when ethical), direct comparisons with standard pharmacological or psychotherapeutic treatments, and cost‑effectiveness evaluations.

The meta‑analytic review concludes that current randomized, placebo‑controlled trials provide promising evidence for psychedelic‑assisted therapy, with large between‑group effect sizes relative to many established pharmacological and psychotherapeutic interventions. While the reviewed trials showed strengths such as low attrition and adherence to preregistered outcomes, the authors emphasise that methodological improvements and larger, more diverse trials are needed. Overall, the findings are presented as supportive of continued research into psychedelic‑assisted therapies as a potentially important new direction in mental health treatment.