Magnitude of Response in Treatment and Control Groups within Psychedelic Trials for Psychiatric Disorders: A Meta-Analysis

Mental disorders cause a large global health and economic burden, and many people do not achieve adequate symptom relief with current first-line treatments such as psychotherapy and pharmacotherapy. Previous meta-analyses of conventional treatments often report small-to-moderate effects and substantial publication bias. Early clinical research suggests psychedelic-assisted psychotherapy may reduce symptoms of depression, PTSD and anxiety with generally mild short‑term adverse events, but trials to date are small, relatively homogeneous and face particular methodological challenges. S. and colleagues set out to quantify how much symptom change occurs within control arms of randomized psychedelic trials and to compare that with change in active treatment arms. Specifically, the authors aimed to (1) estimate between-group treatment effects using change scores, (2) describe within-group pre–post symptom changes separately for treatment and control arms, and (3) explore whether within-control change and between-group effects differ by control type (inactive placebo versus active/low‑dose placebo). The motivation was to understand the contribution of non‑drug contextual factors (expectancy, psychotherapy, assessment effects) to observed outcomes in psychedelic trials and to inform trial design and interpretation.

The study is a systematic review and meta-analysis prepared in line with PRISMA and preregistered in PROSPERO (CRD420251111853). The authors conducted a comprehensive search on 1 July 2025 across OVID (MEDLINE, Embase, APA PsycInfo) and PubMed for randomized controlled trials of classic psychedelics (psilocybin, LSD, DMT, MDMA, ayahuasca) for psychiatric disorders. Ketamine and other NMDA‑targeting agents were excluded. The extracted text contains a discrepancy: the search statement reports no language restrictions, but the inclusion/exclusion criteria later list non‑English publications as excluded; this inconsistency is present in the source text and is reported as such. Eligible studies were peer‑reviewed RCTs involving participants with a diagnosed mental disorder comparing a psychedelic or psychedelic‑assisted therapy with a placebo control (inactive or active/low‑dose). Studies were excluded if there was concurrent pharmacotherapy, non‑human work, or (per the extraction) non‑English publication. A two‑stage screening process was used, and screening and data extraction were performed by two independent reviewers; conflicts were resolved by discussion and a third reviewer when necessary. Extracted items included trial and participant characteristics, psychedelic type/dose/route/duration, control type, concurrent psychotherapy, and validated outcome scores at reported time points. Risk of bias was assessed with the Cochrane Risk of Bias tool. All quantitative analyses were conducted in R (version 4.4.0) using the metafor package. Primary between‑group efficacy analyses used standardized mean differences (SMDs) of change scores (treatment vs control). When change statistics were not reported, change means were computed as endpoint minus baseline and SDs of change were derived assuming a pre–post correlation r = 0.5; sensitivity analyses varied r (0.3, 0.7, 0.9). To characterise within‑arm changes the authors computed standardized mean change with change‑score standardisation (SMCC), which accounts for the paired pre–post structure. Random‑effects models were used throughout given expected heterogeneity. Control conditions were classified as inactive placebo versus active/low‑dose placebo and subgroup analyses by placebo type were performed for depressive and PTSD outcomes where data allowed. Heterogeneity was quantified with I² and τ², and small‑study effects/publication bias were assessed using funnel plots and Egger’s test when at least 10 arms were available.

The search yielded 3,295 records (320 duplicates removed); after screening 32 full texts were assessed and 14 RCTs (total n = 643) met inclusion criteria. Risk of bias across the 14 trials was predominantly low: ten studies were rated low risk overall, three were judged to have some concerns (primarily in outcome measurement), and one study was assessed as high risk overall owing to deviations from intended interventions and additional concerns about outcome measurement and selective reporting. Between‑group meta‑analyses of change scores favoured the psychedelic treatment arms across outcomes. For depressive symptoms (k = 13 comparisons) the pooled SMD was −0.82 (95% CI −1.17 to −0.47), with I² = 60.1% indicating moderate–substantial heterogeneity. For PTSD symptoms (k = 10) the SMD was −0.89 (95% CI −1.14 to −0.65), with I² = 0%. For anxiety symptoms (k = 5) the SMD was −0.66 (95% CI −0.94 to −0.38), with I² = 0%. Sensitivity analyses excluding studies with high risk or some concerns produced similar between‑group estimates for depression (k = 10; SMD = −0.88; 95% CI −1.31 to −0.44; I² = 68.3%) and PTSD (k = 9; SMD = −0.89; 95% CI −1.14 to −0.64; I² = 0%). Funnel plots and Egger’s tests provided limited evidence of small‑study effects for the between‑group analyses (Egger p = 0.24 for depression; p = 0.15 for PTSD). Within‑group analyses demonstrated clinically meaningful pre–post symptom reductions in treatment arms and also measurable reductions in control arms, though pooled control‑arm SMCC estimates are not clearly reported in the extracted text. Reported pooled within‑treatment arm effects were large: depression (k = 13; SMCC = −1.30; 95% CI −1.69 to −0.92; I² = 84.1%), PTSD (k = 10; SMCC = −1.47; 95% CI −1.86 to −1.08; I² = 61.7%), and anxiety (k = 5; SMCC = −1.16; 95% CI −1.40 to −0.92; I² = 0%). Funnel inspection suggested asymmetry for treatment‑group within‑group analyses and Egger’s test indicated evidence of small‑study effects (p = 0.005). Sensitivity analyses varying the assumed within‑study correlation (r = 0.3–0.9) altered the magnitude of pooled effects as expected (larger absolute values at higher r) but did not change the direction of effects. One control arm was omitted from the SMCC meta‑analysis because its sample size (n = 2) produced a non‑estimable effect size. Exploratory subgroup analyses by control type found no clear differences in between‑group effects by placebo type for depressive or PTSD symptoms, but descriptive within‑control analyses suggested larger PTSD symptom reductions in trials using inactive placebo compared with active/low‑dose controls; the authors caution these subgroup findings are exploratory given small numbers of studies per subgroup.

S. and colleagues interpret their findings as showing that psychedelic‑assisted interventions produce greater symptom reductions than control conditions for depression, PTSD and anxiety in randomized trials, with larger treatment effects for depression and PTSD and more moderate effects for anxiety. At the same time, control arms in these trials exhibited clinically meaningful within‑group improvements, indicating that non‑pharmacological trial factors—such as expectancy, therapeutic contact and structured psychotherapy delivered to both arms—likely contribute to observed change under trial conditions. The authors note that control‑arm improvements in their synthesis appear less extreme than some reports of large placebo responses in broader mental health literature, but direct comparisons are limited by differences in populations, concomitant psychotherapy, follow‑up intervals and effect‑size computations. They highlight methodological challenges specific to psychedelic trials: intense experiential effects can lead to functional unblinding and strong expectancies; control conditions often include substantial preparatory and integration psychotherapy, creating a rich therapeutic context even in placebo arms; and low‑dose or active placebos may not be pharmacologically inert and may still affect expectancy or produce subtle subjective effects. The within‑control finding of larger PTSD improvements under inactive placebo (versus active/low‑dose controls) is discussed as a potentially important signal that comparator choice can influence outcomes, though the authors emphasise limited power and inconsistent measurement of blinding and expectancy across studies. Key limitations acknowledged by the authors include inconsistent reporting of expectancy and blinding integrity (preventing quantitative evaluation of these factors), the inclusion of psychotherapy in placebo arms (which makes it impossible to disentangle drug‑specific from therapy effects within the available trials), modest sample sizes and limited power to detect subgroup differences, and sensitivity of within‑group estimates to assumptions about pre–post correlations. The authors recommend that future trials report expectancy and masking integrity more systematically, consider designs that can separate psychotherapy from drug effects (for example therapy‑alone arms or 2×2 designs), increase sample sizes, and run mechanistic and head‑to‑head comparator studies to better isolate drug‑specific effects. They conclude that careful design, justification and transparent reporting of control conditions—as well as routine assessment of blinding and expectancy—are necessary so that any incremental benefit attributable to classic psychedelics can be interpreted in the context of the broader therapeutic milieu.