Assessing expectancy and suggestibility in a trial of escitalopram v. psilocybin for depression

Depressive disorders are highly prevalent worldwide and current first-line treatments such as selective serotonin reuptake inhibitors (SSRIs) show only modest advantages over placebo. Psychedelic-assisted therapy, most prominently using psilocybin, has been proposed as an alternative approach in which one or two supervised dosing sessions interact with psychotherapeutic processes to produce therapeutic benefit. However, the conspicuous acute effects of psychedelics make effective blinding difficult, raising concerns that expectancy and related extra-pharmacological factors (for example demand characteristics or the Hawthorne effect) may bias outcomes in psychedelic trials. Szigeti and colleagues set out to examine whether pre-trial efficacy-related expectancy and baseline trait suggestibility and absorption were associated with therapeutic response in a previously conducted Phase II double-blind, randomised, parallel-group trial that compared escitalopram with investigational COMP360 psilocybin for moderate-to-severe major depressive disorder. The analysis aimed to determine whether expectancy effects differed between the two arms and whether suggestibility or absorption predicted response, with implications for the interpretation of comparative results between psilocybin therapy and standard pharmacotherapy.

The data analysed were drawn from a Phase II, investigator-initiated, double-blind, randomised, parallel-group trial in patients with moderate-to-severe major depressive disorder. The sample size reported in the extract was n = 55. Randomisation was 1:1 to two arms. In the psilocybin arm participants received two full 25 mg doses of COMP360 psilocybin given three weeks apart plus six weeks of daily placebo capsules. In the escitalopram arm participants received two 1 mg psilocybin doses (treated by the team as an active placebo for the dosing sessions) three weeks apart plus six weeks of daily oral escitalopram (10 mg/day for the first 3 weeks, 20 mg/day for the final 3 weeks). All participants received psychological support during the trial period. Trial registration and consent procedures were reported. Pre-treatment measures relevant to this analysis were efficacy-related expectancy (two visual analogue items, 0–100 scale, administered one day before each dosing session and here using the pre-first-dose values), baseline trait suggestibility measured by the Short Suggestibility Scale (SSS), and absorption measured by the Modified Tellegen Absorption Scale (MODTAS). The authors analysed ‘‘received treatment expectancy’’—that is, the expectancy score corresponding to the treatment actually received for each participant—because blinding integrity was expected to be compromised by recognisable drug effects. Primary clinical outcome for the parent trial was change in the self-rated QIDS-SR-16 score at week 6. Secondary outcomes re-analysed here included clinician-rated BDI, HAM-D, MADRS, and additional mood/wellbeing measures STAI-T and WEMWBS. Statistical analyses used linear mixed models: separate within-arm models assessed the interaction between baseline covariates (expectancy, suggestibility, absorption) and timepoint on each outcome, with patient ID as a random effect. Between-arm models examined treatment allocation, timepoint, and their interactions adjusted for expectancy or suggestibility. Baseline covariates were normalised by subtracting the median and dividing by the standard deviation so reported estimates correspond to a 1 SD change at the median. P values were adjusted for multiple comparisons using the Bonferroni method. Residual normality was checked via QQ-plots and models were implemented in R using lme4/lmerTest. The authors also used the two one-sided t tests (TOST) equivalence procedure with an equivalence bound of 0.5 standardized mean difference to test whether null results could be taken as evidence of absence of an effect. The manuscript repository reportedly contains data and analysis code to reproduce results.

Pre-treatment expectancy and baseline traits: Across the full sample, pre-trial efficacy-related expectancy was substantially higher for psilocybin than for escitalopram. Estimated means were 54% for psilocybin versus 28.2% for escitalopram, with an estimated difference of 25.8 ± 3.5 (p < 0.001). There were no significant differences by randomised allocation in these expectancy scores, nor in baseline suggestibility or absorption measures. Within-arm associations between expectancy and outcomes: In the escitalopram arm, higher pre-treatment ‘‘received treatment expectancy’’ for escitalopram was associated with greater improvement over time on several clinician- and self-rated measures. Significant expectancy × timepoint interactions were observed for HAM-D (estimate ± S.E.: -3.91 ± 0.90, adjusted p = 0.001), BDI (-5.47 ± 1.61, adj. p = 0.013), MADRS (-4.87 ± 1.52, adj. p = 0.022), and STAI-T (-5.20 ± 1.68, adj. p = 0.028). By contrast, the escitalopram expectancy interaction was not significant for the QIDS-SR-16 (the trial primary outcome) or WEMWBS. To give an anchor, on the HAM-D each 1 SD increase in expectancy (approximately 22 points on the 0–100 scale) was associated with a 3.91-point reduction in score. In the psilocybin arm, pre-treatment psilocybin expectancy showed no significant association with change over time on any outcome (all adjusted p values = 1 after correction). Equivalence testing for the expectancy × timepoint interaction in the psilocybin arm was non-significant on all scales, indicating the data could not rule out effects as large as the pre-specified 0.5 S.M.D. bound. Suggestibility and absorption: Baseline suggestibility was not associated with therapeutic response in the escitalopram arm. In the psilocybin arm, however, baseline suggestibility showed significant interactions with timepoint across all reported outcomes, implying higher suggestibility predicted larger improvements following psilocybin therapy. For example, on HAM-D each 1 SD increase in suggestibility (roughly 10 points on the SSS) was associated with a 3.46-point greater reduction in depression score (estimate ± S.E.: -3.46 ± 0.92, adj. p = 0.005). Similar significant effects were reported for BDI, MADRS, QIDS-SR-16, STAI-T and WEMWBS in the psilocybin arm. Baseline absorption did not show significant associations with outcomes in either arm. Between-arm models adjusted for expectancy or suggestibility: When models were adjusted for pre-treatment expectancy, the treatment × timepoint interaction was not significant on any scale after multiple-comparison correction, implying no between-treatment difference once expectancy was accounted for; equivalence tests were non-significant. A treatment × timepoint × expectancy three-way interaction was significant on HAM-D and MADRS, indicating some scale-specific differences when expectancy was modelled. When adjusting for suggestibility, the between-arm treatment × timepoint interaction remained significant on most scales (HAM-D, BDI, MADRS, STAI-T, WEMWBS) favouring psilocybin, but not on QIDS-SR-16. The treatment × timepoint × suggestibility interaction was significant across all scales, consistent with the within-arm finding that suggestibility modified response more in the psilocybin arm than in the escitalopram arm. Safety and other results: The extracted text did not present additional numerical safety data or adverse-event tables within these sections. Figures and supplementary tables were referenced for regression lines and further model details.

Szigeti and colleagues interpret their findings as indicating that, although participants expressed substantially higher pre-trial positive expectancy for psilocybin than for escitalopram, only escitalopram expectancy predicted response within its arm, while psilocybin expectancy did not predict psilocybin response. The escitalopram finding aligns with prior literature on SSRIs and expectancy effects. The absence of a clear positive expectancy effect in the psilocybin arm was unexpected given widespread evidence that expectancy influences outcomes across medical contexts. The authors note that a non-significant result does not prove absence of an effect: equivalence testing was inconclusive, so effects as large as 0.5 S.M.D. could not be excluded with this sample size. The investigators discuss several possible explanations for the lack of a detectable positive expectancy effect in the psilocybin arm. They rejected a simple ceiling effect because average psilocybin expectancy was only about 51% on the 0–100 scale. Non-linear relationships between expectancy and outcome are proposed as a possibility that the present sample was too small to test. The authors also emphasise that adjusting between-arm comparisons for expectancy removed the previously reported superiority of psilocybin on most scales, suggesting that an imbalance in expectancy—specifically lower expectancy for escitalopram—may have biased unadjusted between-group results, effectively acting as a nocebo in the escitalopram arm rather than a direct positive expectancy benefit in the psilocybin arm. Baseline trait suggestibility was robustly associated with treatment response in the psilocybin arm but not the escitalopram arm. The authors suggest this could indicate that extra-pharmacological factors, including verbal suggestibility, therapist influence, or demand characteristics, may contribute to response in psychedelic therapy, or alternatively reflect an interaction between psychological traits and pharmacological effects. Absorption did not predict outcomes despite prior work linking it to the acute intensity of psychedelic experiences. The authors call for further work to determine whether the relationship between suggestibility and psilocybin response is biological, psychological, or both, and how it might interact with acute drug effects and psychotherapeutic processes. Limitations acknowledged by the study team include that this analysis was not pre-registered and therefore exploratory; causal inference is not possible without experimental manipulation of expectancy; the sample size was limited and underpowered to reject equivalence at the pre-specified bound; the expectancy measure was not a validated instrument; and no data were collected on participants' guesses about treatment allocation or the confidence of those guesses, preventing assessment of blinding integrity. The authors recommend larger, ideally pragmatic or real-world studies, and experimental manipulation of expectations to clarify causality. They also highlight the importance of investigating negative expectancies (nocebo effects) in future psychedelic trials.

In this re-analysis of a randomised trial comparing escitalopram with psilocybin therapy for depression, participants reported higher pre-trial positive expectancy for psilocybin than for escitalopram. Pre-trial expectancy for escitalopram predicted therapeutic response to escitalopram across multiple outcome measures, whereas psilocybin-specific expectancy did not predict response to psilocybin. Baseline trait suggestibility predicted response to psilocybin but not to escitalopram. The authors emphasise that these results are exploratory and that further research is needed to clarify causality and the mechanisms linking suggestibility and psychedelic therapy response.