Meta-correlation of the effect of ketamine and psilocybin induced subjective effects on therapeutic outcome

Psychiatric disorders remain prevalent worldwide and current first-line antidepressant treatments have limited and delayed efficacy, prompting interest in alternative pharmacological approaches. Psychedelic and psychoplastogenic drugs such as ketamine and psilocybin have emerged as promising candidates because they promote rapid neural plasticity and can produce marked changes in consciousness, perception, mood and cognition. Despite shared neurobiological actions, these compounds generate qualitatively different subjective experiences: ketamine typically produces short-lived dissociation, whereas psilocybin commonly evokes mystical-type and oceanic-boundlessness experiences. Dahan and colleagues set out to examine whether those subjective effects are quantitatively related to therapeutic benefit. Noting prior inconsistent findings and methodological concerns (timing of measures, measurement instruments, and dose–response issues), the study performs a systematic review and meta-correlation analysis focused on ketamine (including racemic ketamine and esketamine) and psilocybin for treatment of depression and substance use disorder (SUD). The principal research question was the magnitude of the correlation between drug‑induced subjective effects and clinical improvement, and whether this relationship differs between ketamine and psilocybin or between depression and SUD.

The investigators registered a protocol on PROSPERO (CRD42024546815) and searched PubMed, EMBASE and Web of Science with no date or language restrictions. Searches were conducted on 26 May 2024 and updated on 30 June 2024. Two reviewers independently screened titles/abstracts and full texts, resolving disagreements by consensus. Inclusion criteria were human studies with full text that reported both treatment effects on depression or SUD and quantitative correlations between those clinical outcomes and drug‑induced subjective effects; RCTs and open‑label studies were eligible. Exclusions comprised case reports, surveys, reviews, animal studies, non‑target diagnoses, or treatments other than ketamine or psilocybin. Risk-of-bias assessment used the Newcastle–Ottawa Scale for non‑randomised trials and the revised Cochrane risk‑of‑bias tool for randomized trials, with emphasis on potential unblinding. For data extraction the team retrieved Pearson or Spearman correlation coefficients (r) from texts or figures, selecting the timepoint closest to treatment and the correlation for the primary efficacy endpoint when multiple endpoints were reported. In two papers that reported no correlation but did not provide r, the authors carried out a sensitivity analysis by imputing r=0 for those studies. Meta-analyses were performed in Comprehensive Meta-Analysis v3.0 using random-effects models to account for within‑ and between‑study variance. The investigators pooled results separately for ketamine and psilocybin, assessed heterogeneity with I2, and used leave-one-out sensitivity analyses to check for influential studies. Prespecified subgroup analyses examined disease state (depression versus SUD), trial type (RCT versus open‑label/reanalyses) and, for ketamine, route of administration (intravenous versus intranasal). Direct comparisons between ketamine and psilocybin were done via mixed‑effects analysis. The authors followed PRISMA 2009 guidance for reporting.

The search yielded 2,049 unique records; 116 full texts were assessed and 23 papers met inclusion/exclusion criteria (published 2000–2024). Of these, 15 papers concerned ketamine (13 depression, 2 SUD) and 8 concerned psilocybin (6 depression, 2 SUD). Three ketamine reports contained overlapping datasets and one of these overlapping reports was retained; two ketamine papers explicitly reported no correlation and were excluded from the primary analysis but included in a secondary analysis with imputed r=0. Twelve of the included studies were randomized controlled trials (10 ketamine, 2 psilocybin); the remainder were open‑label or reanalyses. The authors judged most studies to have moderate or high risk of bias when considering expectation/unblinding. Ketamine: Eleven studies contributed to the primary ketamine meta‑analysis (443 treated patients). The pooled correlation coefficient (r) between subjective effects and treatment efficacy was -0.310 (95% CI -0.475 to -0.124, p=0.001), with I2=55%. Leave‑one‑out testing did not identify a single dominant study. Restricting to intravenous ketamine (n=10) produced r=-0.355 (95% CI -0.548 to -0.126, p=0.003) with substantial heterogeneity (I2=99%), whereas two intranasal/esketamine studies pooled to r=-0.164 (95% CI -0.341 to 0.025, p=0.089, I2=0%). For depression only (n=9) the pooled r was -0.201 (95% CI -0.344 to -0.050, p=0.009, I2=29%); for SUD (n=2) the pooled r was -0.738 (95% CI -0.917 to -0.310, p=0.003, I2=39%). RCTs (n=8) had r=-0.357 (95% CI -0.556 to -0.118, p=0.004, I2=57%), while open‑label/reanalysis studies (n=3) showed a non‑significant pooled r=-0.221 (95% CI -0.527 to 0.136, p=0.223, I2=40%). Including the two studies with imputed r=0 increased the sample to 13 studies (471 patients) and yielded an overall r=-0.268 (95% CI -0.421 to -0.100, p=0.002, I2=48%). Psilocybin: Eight studies (183 patients) entered the psilocybin meta‑analysis. The overall pooled r was -0.495 (95% CI -0.624 to -0.341, p<0.001), with I2=28%. Leave‑one‑out analyses did not identify a dominating study. In depression‑only studies (n=6) pooled r=-0.426 (95% CI -0.550 to -0.284, p<0.001, I2=0%); for SUD (n=2) pooled r=-0.776 (95% CI -0.930 to -0.391, p=0.001, I2=40%). RCTs (n=3) produced r=-0.355 (95% CI -0.517 to -0.169, p<0.001, I2=0%), and open‑label trials pooled to r=-0.620 (95% CI -0.757 to -0.430, p<0.001, I2=17%). Comparisons and subgroup findings: Direct comparison of ketamine versus psilocybin showed no statistically significant difference when excluding the two imputed studies (ketamine n=11 r=-0.310 versus psilocybin n=8 r=-0.495; p=0.108). Including the imputed ketamine studies made the difference statistically significant (ketamine n=13 r=-0.268 versus psilocybin n=8 r=-0.495; p=0.040). Comparing depression studies without imputation showed a significant greater correlation for psilocybin than ketamine (ketamine n=9 r=-0.201 versus psilocybin n=6 r=-0.426; p=0.0228). Across treatments, correlations were larger in SUD versus depression: ketamine in SUD r=-0.738 versus depression r=-0.201 (ratio 3.7); psilocybin in SUD r=-0.776 versus depression r=-0.425 (ratio 1.8). Pooling the four SUD studies (two ketamine, two psilocybin; total n=54) produced a combined r=-0.740 (95% CI -0.862 to -0.539, I2=12%, p<0.001).

Dahan and colleagues interpret their findings as evidence of a modest but measurable relationship between drug‑induced subjective effects and therapeutic improvement for both ketamine and psilocybin in depression and SUD. The pooled correlations were moderate in size (r≈-0.31 for ketamine and r≈-0.50 for psilocybin). Converting these correlations to coefficients of determination (R2), the authors estimate that subjective effects account for roughly 10% of ketamine's therapeutic variance (5% when the two imputed studies are included) and about 24% for psilocybin. Results varied substantially across individual studies: some reported R2 values exceeding 50%, while others reported no association. Several possible explanations for the differences are offered. Measurement tools may be inappropriate or insensitive: most ketamine studies used CADSS or BPRS, scales not originally designed to capture the full spectrum of ketamine‑induced changes in consciousness, whereas psilocybin studies often use instruments targeting mystical or oceanic‑boundlessness experiences. The investigators suggest that instruments tailored to each drug's phenomenology (for example the Bowdle Visual Analog Scale for ketamine or mystical‑experience scales for psilocybin) might yield stronger correlations. Trial design and expectation/unblinding are also important: subjective effects make functional blinding difficult, which could introduce expectation bias. Nonetheless, the authors argue unblinding alone does not fully explain the observed correlations because blinded experimental manipulations that reduced subjective effects (e.g., ketamine during general anaesthesia or co‑administration of sodium nitroprusside in an experimental pain model) were accompanied by attenuated therapeutic or symptomatic effects. The discussion emphasises caution in interpretation. Heterogeneity across studies, small numbers of SUD trials (only four studies total), variable use of concomitant psychotherapy, and potential mismatches between correlation measures (Pearson versus Spearman) limit causal inference. The authors note that correlations could be non‑linear or specific to particular subjective domains (e.g., oceanic boundlessness) that were not separable in their pooled analysis. Ultimately, they state the analyses cannot confirm causality; subjective experiences may mediate therapeutic benefit in part, or they could be epiphenomenal markers of underlying neurobiological changes. The paper closes by calling for larger, better‑designed studies that use appropriate phenomenological instruments, consider psychotherapy interactions, and apply analytic choices (such as rank correlations) that match data distributions to clarify the role of subjective effects in psychedelic therapy.