The role of therapeutic alliance in psilocybin treatment for treatment-resistant depression: A post hoc path analysis

The paper addresses a long-running debate in psychedelic psychiatry about how much of psilocybin’s antidepressant effect comes from the drug itself versus the psychotherapeutic support surrounding it. The authors note that therapeutic alliance, meaning the quality of the working relationship between participant and support provider, is often associated with better outcomes in conventional psychotherapy, but its role in psilocybin treatment remains uncertain. They also highlight that previous findings suggesting a strong link between alliance and outcomes came from small studies, often with different dosing, support intensity, and patient populations, making it unclear how generalisable those observations are. The study aimed to examine whether pre-dosing therapeutic alliance predicted the psychedelic experience and subsequent depression outcomes in participants with treatment-resistant depression receiving psilocybin. Using data from a larger Phase II trial, the researchers explored correlations and path models linking therapeutic alliance, dimensions of the psychedelic experience, and change in depression severity after a single 25 mg dose. The broader purpose was to test whether alliance directly influenced clinical benefit or worked mainly by facilitating the psychedelic experience itself.

This was a post hoc analysis of COMP 001, an international, multicentre, Phase II, randomised, fixed-dose, parallel-group, double-blind, dose-finding trial. Adults with treatment-resistant depression were randomised to receive a single dose of 25 mg, 10 mg, or 1 mg psilocybin as COMP360, with follow-up for 12 weeks. The main text of this paper focuses on the 25 mg group because that dose showed clinical efficacy in the parent trial, although lower-dose and whole-sample analyses were also conducted and are said to be in the supplementary material. Participants were adult outpatients with a current major depressive episode without psychotic features, of at least moderate severity, and with inadequate response to two to four antidepressant treatments. A total of 233 participants were randomised overall, with 79 assigned to 25 mg, 75 to 10 mg, and 79 to 1 mg. The mean age was 39.8 years and 52% were female. The article states that baseline characteristics were comparable across groups. Participants received psilocybin alongside monitoring and support from trained providers. The support model included three preparatory sessions before dosing, two follow-up sessions after dosing, and a largely non-directive administration session in which staff minimised input apart from safety-related support. The preparation sessions were intended to build trust, provide education, discuss life events, set intentions, and establish boundaries and support strategies. Therapeutic alliance was measured using the Scale to Assess Therapeutic Relationship-Patient version (STAR-P) after the third preparation session, the day before dosing. Depression severity was assessed with the Montgomery-Åsberg Depression Rating Scale (MADRS) at Week 3, with remote raters blind to dose and study details. Psychedelic experience was assessed using the 5-Dimensional Altered States of Consciousness Questionnaire (5D-ASC), including its dimensions such as Oceanic Boundlessness and Visual Restructuralization, and the Emotional Breakthrough Inventory (EBI). The 5D-ASC was completed after the acute drug effects had subsided on Day 1, and the EBI on Day 2 before the first follow-up session. The researchers first examined Fisher z-transformed Pearson correlations between therapeutic alliance, psychedelic experience, and depression outcomes. They then used saturated path analyses to estimate direct and indirect relationships between STAR-P, one psychedelic experience measure at a time, and Week 3 MADRS scores. Maximum likelihood estimation with Satorra-Bentler adjustments was used because multivariate normality was not met. Nominal significance was set at the 5% level, and Bonferroni correction was applied where appropriate for the indirect pathways.

In the 25 mg group, therapeutic alliance showed only a weak negative correlation with change from Baseline to Week 3 MADRS scores (r = -0.178), suggesting that higher alliance was associated with somewhat greater symptom reduction, but the relationship was modest. By contrast, correlations between psychedelic experience measures and clinical outcome were much stronger, with reported values for EBI, Oceanic Boundlessness, and Visual Restructuralization ranging from about -0.508 to -0.637. Alliance was more closely correlated with certain psychedelic experience dimensions than with depression outcome, especially Visual Restructuralization (0.299), Oceanic Boundlessness (0.298), and Auditory Alterations (0.258) in the 25 mg group. These correlations were weaker in the lower-dose groups. Path analysis found no nominally significant direct effect of therapeutic alliance on Week 3 MADRS scores in any dose group. However, in the 25 mg group, therapeutic alliance had nominally significant direct effects on Oceanic Boundlessness (β = 0.28, p = 0.01), Visual Restructuralization (β = 0.27, p = 0.005), and Auditory Alterations (β = 0.25, p = 0.02). In the 10 mg group, there was one nominally significant direct effect of STAR-P on Anxious Ego Dissolution (β = -0.23, p = 0.04), but the explanatory power of that model was low. The strongest direct predictors of Week 3 MADRS scores were the psychedelic experience measures themselves. EBI had the largest association with outcome (β = -0.59, p < 0.0001), followed by Visual Restructuralization (β = -0.54, p < 0.0001) and Oceanic Boundlessness (β = -0.53, p < 0.0001). Auditory Alterations was also a nominally significant predictor in the 25 mg group (β = -0.24, p = 0.03). Reduction of Vigilance and Anxious Ego Dissolution were not significant predictors in the 25 mg analysis. Indirect effects of therapeutic alliance on Week 3 MADRS through psychedelic experience were small. In the 25 mg group, indirect effects via Oceanic Boundlessness and Visual Restructuralization were nominally significant (both β = -0.15), but only the Visual Restructuralization pathway remained significant after Bonferroni correction (p bf = 0.04). No other indirect effects were nominally significant in the 25 mg group or other dose groups. The models explained the most variance when they included EBI, Visual Restructuralization, or Oceanic Boundlessness, with R2 values of 37%, 30%, and 29% respectively. Models involving Auditory Alterations, Anxious Ego Dissolution, and Reduction of Vigilance explained much less variance, at 9%, 5%, and 3% respectively.

The authors interpret the findings as showing that therapeutic alliance had only a limited role in psilocybin’s antidepressant effect in this trial. In the 25 mg group, alliance was weakly related to improvement in MADRS scores and had no direct effect on depression outcome in the path models. Instead, alliance appeared to be associated mainly with aspects of the psychedelic experience, particularly Oceanic Boundlessness and Visual Restructuralization, which in turn were strongly related to symptom improvement. The authors therefore argue that alliance may help facilitate the experience rather than drive clinical benefit directly. They contrast these findings with earlier research that reported a stronger link between therapeutic alliance and outcome. The authors suggest that differences in study design may help explain the discrepancy, including the much smaller sample in previous work, the non-treatment-resistant population, the open-label single-site setting, and the substantially longer preparation period in that study. They note that their own preparation consisted of only around 2 to 3 hours, whereas earlier studies may have included much more intensive supportive psychotherapy. The authors further argue that their data support the view that the efficacy of COMP360 in depression is largely tied to the pharmacological effect of psilocybin and the psychedelic experience it induces, rather than to psychotherapy-like mechanisms in the conventional sense. They also suggest that their findings have implications for how support providers are trained and qualified, but caution that more research is needed before firm standards are set for clinical implementation. Several limitations are acknowledged. The authors note that STAR-P may not be the ideal measure for this setting, and that alliance scores were already high with relatively little variability, which may have limited the ability to detect associations. They also emphasise that correlation and path analysis cannot establish causality, and that alliance may partly reflect other factors such as expectancy, personality traits, openness, absorption, or suggestibility. Another limitation is that the antidepressant effect was already evident by one day after dosing, so the analysis does not address the potential impact of post-dose support sessions. The authors conclude that the role of support after dosing remains an open question for future research.

The authors conclude that the preparation and support framework helped create a strong therapeutic alliance and an acceptable psychedelic experience, but that alliance itself was not a major driver of clinical efficacy in treatment-resistant depression. They state that the psychedelic experience, not the alliance, appears to be the main pathway linking psilocybin treatment to antidepressant benefit. They also argue that these findings do not support earlier claims that therapeutic alliance has a stronger effect on outcomes, and they suggest that psilocybin’s pharmacological effects may be better understood as principal agents of therapeutic change rather than merely as aids to psychotherapy.