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

Psychiatric disorders affect millions worldwide, and conventional antidepressants often have limited and delayed efficacy, motivating the search for alternative treatments. Psychedelics (classical serotonergic agents such as psilocybin and non-classical agents such as ketamine) produce rapid psychoplastogenic effects and distinctive subjective experiences (e.g., dissociation for ketamine; mystical-type experiences for psilocybin) that some hypothesize are linked to clinical benefit. Evidence on whether these subjective effects mediate therapeutic outcomes is mixed, with critiques citing temporal mismatches between symptoms and outcomes, limited sensitivity of measurement tools, and lack of dose-response relationships. The authors conducted a systematic review and meta-correlation analysis focused on ketamine and psilocybin to quantify the correlation between drug-induced subjective effects and therapeutic improvement in depression and SUD, and to test whether correlations differ between ketamine and psilocybin.

Narrative reviews have described therapeutic potential and subjective experiences of psychedelics, including the role of mystical-type experiences. Ballard and Zarate argued that dissociation may not mediate ketamine’s antidepressant effects, citing inconsistent correlations across studies. Potential reasons for inconsistency include differences in timing of measurements, insensitive or non-specific tools to capture subjective effects (e.g., CADSS and BPRS not designed for ketamine/psilocybin-specific experiences), and absence of dose-response relationships. Multiple prior RCTs and open-label studies have evaluated ketamine and psilocybin across depression and SUD, but no prior quantitative meta-correlation focused specifically on linking subjective effects to outcomes for these two drugs had been performed.

Protocol registered in PROSPERO (CRD42024546815). Systematic searches were conducted in PubMed, EMBASE, and Web of Science (initial: May 26, 2024; update: June 30, 2024), without date or language restrictions, supplemented by reference screening. Inclusion criteria: human studies; full-text available; depression or SUD outcomes; reported subjective effects from ketamine (racemic or esketamine) or psilocybin; and quantitative correlation between subjective effects and therapeutic outcomes; RCTs or open-label designs. Exclusions: case reports, surveys, reviews, animal studies, diagnoses other than depression or SUD, or treatments other than ketamine or psilocybin. When multiple datasets were combined in a reanalysis, original overlapping studies were excluded to avoid duplication. Risk of bias: non-randomized studies via Newcastle-Ottawa Scale; RCTs via Cochrane RoB 2, with attention to risks of unblinding/expectation bias. Data extraction: Pearson or Spearman correlation coefficients (r) and sample sizes were extracted from text or figures. If multiple timepoints existed, the timepoint closest to treatment was used; if multiple efficacy endpoints were reported, the primary endpoint’s correlation was used. Random-effects meta-analyses (Comprehensive Meta-Analysis v3) pooled correlations separately for ketamine and psilocybin; heterogeneity assessed by I²; sensitivity via leave-one-out analysis; subgroup analyses by disorder (depression vs SUD), trial type (RCT vs open-label/reanalysis), and for ketamine by route (IV vs IN). Two studies that reported absence of correlation without giving r were included in a secondary analysis with r imputed as 0. Mixed-effects models compared ketamine vs psilocybin.

Study selection yielded 23 studies for primary/secondary analyses: 15 ketamine (depression n=13; SUD n=2) and 8 psilocybin (depression n=6; SUD n=2). Primary analysis included 11 ketamine studies (9 depression, 2 SUD) and 8 psilocybin studies. Ketamine (n=11; 443 patients): pooled r = -0.310 (95% CI -0.475 to -0.124; p=0.001), I²=55%. Leave-one-out showed no dominance by any single study. Subgroups: IV ketamine (n=10) r = -0.355 (-0.548 to -0.126; p=0.003; I²=99%); IN ketamine/esketamine (n=2) r = -0.164 (-0.341 to 0.025; p=0.089; I²=0). Depression (n=9) r = -0.201 (-0.344 to -0.050; p=0.009; I²=29%); SUD (n=2) r = -0.738 (-0.917 to -0.310; p=0.003; I²=39%). RCTs (n=8) r = -0.357 (-0.556 to -0.118; p=0.004; I²=57%); open-label/reanalysis (n=3) r = -0.221 (-0.527 to 0.136; p=0.223; I²=40%). Including two studies with imputed r=0 (total n=471, 13 studies) gave r = -0.268 (-0.421 to -0.100; p=0.002; I²=48%). Psilocybin (n=8; 183 patients): pooled r = -0.495 (95% CI -0.624 to -0.341; p<0.001), I²=28%. Depression (n=6) r = -0.426 (-0.550 to -0.284; p<0.001; I²=10%); SUD (n=2) r = -0.776 (-0.930 to -0.391; p=0.001; I²=40%). RCTs (n=3) r = -0.355 (-0.517 to -0.169; p<0.001; I²=0%); open-label (n=2) r = -0.620 (-0.757 to -0.430; p<0.001; I²=17%). Between-drug comparisons: excluding imputed studies, ketamine r = -0.310 vs psilocybin r = -0.495 (p=0.108); including imputed studies, ketamine r = -0.268 vs psilocybin r = -0.495 (p=0.040). In depression-only analyses (no imputation): ketamine r = -0.201 vs psilocybin r = -0.426 (p=0.0228). SUD pooled across drugs (4 studies; 54 patients): r = -0.740 (95% CI -0.862 to -0.539; p<0.001), I²=12%. Interpretation in terms of R²: subjective effects account for an estimated ~10% of ketamine’s therapeutic variance (≈5% including imputed studies) and ~24% for psilocybin overall; within depression: ≈4% (ketamine) and ≈18% (psilocybin); for SUD: ≈54% (ketamine) and ≈60% (psilocybin), noting small SUD sample size.

The meta-analyses indicate a statistically significant, modest correlation between subjective effects and therapeutic outcomes for both ketamine and psilocybin, with stronger correlations for psilocybin, particularly in depression, and for both drugs in SUD. These results support the hypothesis that subjective experience may mediate part of the therapeutic response, though the effect size is moderate and variable across studies. Measurement instruments may influence observed relationships: ketamine studies commonly used CADSS or BPRS, which may not capture the full spectrum of ketamine’s subjective experiences, potentially attenuating correlations; psilocybin studies often used tools targeting mystical-type experiences (e.g., MEQ, oceanic boundlessness), potentially yielding stronger associations. Expectation bias and functional unblinding are concerns in psychedelic research; however, patterns differed by drug (higher correlations in psilocybin open-label vs RCTs; the reverse for ketamine), and available evidence suggests the association is not solely a byproduct of unblinding. Alternative designs reducing expectation bias (e.g., administering ketamine under general anesthesia; pharmacologically attenuating subjective effects with sodium nitroprusside) suggest that dampening subjective effects may reduce therapeutic effects, though these studies are not direct clinical replications and may involve additional confounders. Nonlinearity and methodological heterogeneity (use of Pearson vs Spearman correlations, timing of measurements, and diverse outcome measures) may contribute to variability. The stronger correlations observed in SUD compared with depression could reflect disorder-specific mechanisms, measurement differences, or small-sample effects. Overall, findings align with a model where subjective effects partially mediate outcomes, with drug- and disorder-specific nuances.

This study provides the first targeted meta-correlation of the association between subjective effects and therapeutic outcomes for ketamine and psilocybin in depression and SUD. Both drugs show modest but significant correlations, stronger for psilocybin than ketamine overall and particularly in depression, and stronger for SUD than depression for both drugs. These results suggest subjective experiences may contribute meaningfully—but not exclusively—to treatment response. Future research should: employ measurement tools tailored to each drug’s phenomenology; examine specific subjective dimensions (e.g., oceanic boundlessness) rather than aggregate scores; use designs that mitigate expectation bias (e.g., enhanced blinding strategies or pharmacologic modulation of subjective effects); test nonlinear associations (favoring Spearman’s rank); disentangle effects of concomitant psychotherapy; and expand robust RCT evidence, especially in SUD, to clarify causality and generalizability.

- Limited number of eligible studies, especially for SUD (4 studies; 54 patients), reducing precision and generalizability. - Heterogeneity across studies in timing of measurements, instruments for subjective effects (e.g., CADSS, BPRS may be suboptimal for ketamine) and outcomes. - Mix of RCTs and open-label/reanalyses with varying risk of bias; potential expectation bias and unblinding remain concerns. - Some studies lacked explicit correlation coefficients and required imputation (r=0), which may bias pooled estimates. - Possible non-linear relationships not captured by Pearson correlations; pooling different correlation types may affect results. - Inclusion of psychotherapy varied across studies and could confound associations. - The meta-correlation assesses association, not causality; subjective effects may be epiphenomena with shared pharmacodynamics rather than mediators.