Cognitive outcomes following psilocybin-assisted therapy in treatment-resistant depression: A post-hoc analysis of a randomized, waitlist-controlled trial

Depression, including unipolar and bipolar forms, frequently involves persistent cognitive difficulties across domains such as attention, processing speed, executive function, language, and episodic memory. These cognitive impairments are common in treatment-resistant depression (TRD), are associated with poorer functioning and quality of life, and may reflect disrupted neuroplasticity in fronto-limbic circuits (for example medial prefrontal cortex, anterior cingulate cortex and hippocampus). Conventional antidepressants often fail to remediate cognitive symptoms, prompting interest in treatments that more directly engage neuroplastic mechanisms. Psilocybin is a serotonergic 5‑HT2A agonist proposed to act as a ‘‘psychoplastogen’’ that can rapidly induce neuroplastic changes in cortical regions implicated in executive function and processing speed. Clinical evidence for psilocybin’s cognitive effects in depression is limited. E. and colleagues therefore conducted a retrospective post‑hoc analysis of data from an investigator‑initiated randomized, waitlist‑controlled, open‑label trial to evaluate whether a single 25 mg dose of psilocybin administered with brief psychotherapeutic support was associated with changes in processing speed and executive function. The primary cognitive measures were the Digit Symbol Substitution Test (DSST) and Trail Making Tests A and B (TMT‑A/B), assessed at baseline, one day and two weeks post‑treatment. The researchers hypothesised that performance would improve at two weeks and that mood improvements would be associated with but not fully account for any cognitive changes; they also planned reliable change index (RCI) analyses to estimate whether observed gains exceeded expected practice effects.

This analysis used data from a randomized, waitlist‑controlled, open‑label clinical trial run at a community clinic in Mississauga, Ontario. The parent trial enrolled adults aged 18–75 with a primary diagnosis of major depressive disorder (MDD) or bipolar II disorder (BD‑II) experiencing a major depressive episode ≥3 months and meeting criteria for treatment resistance (nonresponse to ≥2 adequate antidepressant trials). Participants discontinued antidepressants, antipsychotics, ketamine/esketamine and augmenting agents for ≥5 half‑lives prior to screening; BD‑II participants could continue mood stabilisers at clinical discretion. Participants were randomised to immediate treatment or a two‑week delayed (waitlist) condition. Although the parent trial allowed up to three psilocybin doses over six months, the present post‑hoc analysis focused exclusively on cognitive change following the first 25 mg oral dose of synthetic psilocybin given in a supportive setting. Preparatory and integration psychotherapy sessions were provided (one preparatory session and two integration sessions, each 1–2 hours), and therapist dyads offered non‑specific support during the 6–8 hour dosing session. Because the parent trial was not powered to detect between‑group efficacy differences, the immediate and waitlist arms were pooled for a single open‑label cohort for this analysis. Of the 30 participants who received at least one 25 mg dose, three lacked baseline cognitive data and one withdrew before follow‑up; 26 participants had baseline cognitive assessments and 25 were retained in the longitudinal models (one participant had incomplete post‑dose data but was retained under maximum likelihood estimation). The RCI analyses used a further reduced sample depending on availability of paired pre/post data. Cognitive outcomes were the DSST (processing speed, sustained attention and some executive demands) and TMT‑A/B (visual attention/psychomotor speed and set‑shifting/executive function; TMTB‑A difference score was used to isolate executive demands from motor/processing speed). Tests were administered at baseline, one‑day and two‑weeks post‑dose; waitlist arm assessments were time‑shifted and harmonised to these labels for pooled analysis. An administration error gave participants 90 s on the DSST rather than the normative 120 s; baseline raw DSST scores were prorated to 120‑s equivalents only for z‑score normative comparisons (unadjusted raw scores were used in mixed models). Baseline performance was converted to age‑adjusted z‑scores using published norms and impairment was defined as ≤1 SD below the normative mean. Statistical analyses comprised separate linear mixed models (LMMs) for each cognitive outcome using raw scores, with participant random intercepts and categorical timepoint (pre‑dose, one‑day, two‑weeks) as the main fixed effect. Models adjusted for age, sex and treatment group (immediate vs waitlist), and included a timepoint-by‑group interaction to test differential change by assignment. Two complementary LMMs were run per outcome: Model 1 estimated total change over time; Model 2 added time‑varying MADRS total score to estimate change independent of concurrent depressive symptom change (MADRS is a clinician‑rated depression scale). Model diagnostics informed a log transformation of the TMTB‑A difference (ln[TMTB‑A + 1]) to improve residual normality; other outcomes were analysed untransformed. Missing data were handled via maximum likelihood under an assumption of missing at random. To assess clinically meaningful individual change, RCIs were computed using the Iverson method, incorporating both a clinical reference (the waitlist participants who completed pre‑ and two‑week pre‑psilocybin tests) and published normative test‑retest parameters. An RCI threshold of ±1.645 (90% CI) classified reliable improvement or decline; one‑tailed binomial tests assessed whether observed proportions exceeded chance (5%). One‑sample t‑tests on continuous RCI scores were also conducted.

Sample characteristics and baseline performance: Twenty‑six participants completed pre‑dose cognitive testing (mean age 43.1 years, SD 12.7; 65% male; 73% white). Most had MDD (n = 22, 85%) and 4 had BD‑II. Education was high (92% ≥ some college/university). Baseline depressive severity averaged MADRS 27.8 (SD 7.3) and mean duration of illness was 17.1 years; treatment resistance was pronounced (mean 11.7 failed medication trials). On age‑adjusted norms a subset showed impairment (≤1 SD below mean): DSST n = 7 (26.9%), TMT‑A n = 4 (15.4%), TMT‑B n = 4 (15.4%), TMTB‑A n = 3 (11.5%). Linear mixed model findings: In unadjusted LMMs (Model 1) there was a significant main effect of timepoint for all cognitive outcomes. DSST scores improved from baseline to one‑day (mean difference = 4.72, 95% CI [1.96, 7.49], p < .001) and baseline to two‑weeks (mean difference = 4.26, 95% CI [0.27, 8.25], p = .034), with no significant change between one‑day and two‑weeks. TMT‑A completion times decreased (improved) from baseline to one‑day (mean difference = −5.45 s, 95% CI [−9.25, −1.66], p = .003) and to two‑weeks (mean difference = −6.00 s, 95% CI [−9.42, −2.58], p < .001). TMT‑B showed a timepoint effect with significant improvement by two weeks (baseline to two‑weeks mean difference = −12.27 s, p < .001) and one‑day to two‑weeks improvement (mean difference = −8.80 s, p = .039); baseline to one‑day was not significant. The log‑transformed TMTB‑A difference score improved significantly from baseline to two‑weeks (mean difference ≈ −0.33 to −0.35 on the log scale, p between .005 and .022 across models). Treatment group and age were generally non‑significant predictors; sex effects favoured females on some TMT metrics in unadjusted models. When MADRS was included as a time‑varying covariate (Model 2) results varied by test. The DSST timepoint effect was attenuated and no longer statistically significant (F(2, 27.36) = 2.95, p = .069); higher MADRS scores were associated with worse DSST performance (β = −0.17, p = .048), suggesting DSST gains were partly mood‑related. In contrast, improvements on TMT‑A, TMT‑B and the TMTB‑A difference remained significant after adjusting for MADRS, and MADRS itself did not significantly predict these TMT outcomes, supporting a degree of mood‑independent change on measures of processing speed/executive function. Sex effects on TMT‑A were attenuated after adjusting for MADRS, but sex remained a significant predictor for TMT‑B and TMTB‑A in some models. Reliable change analyses: Using the clinical waitlist reference (pre‑psilocybin retest values), reliable improvement (RCI > 1.645) was observed in 2/24 participants (8.3%) on the DSST, 2/24 (8.3%) on the TMT‑A, and 3/24 (12.5%) on the TMT‑B. None of these proportions significantly exceeded the 5% expected by chance (binomial p‑values .116–.339). Group‑level one‑sample t‑tests on continuous RCI scores showed no significant change for the DSST (t(23) = −0.291, p = .387), a trend for TMT‑A (t(23) = 1.703, p = .051), and a significant group‑level improvement for TMT‑B (t(23) = 2.224, p = .018, d = 0.454). Using normative reference estimates produced similar results: reliable improvement was seen in 3 participants (12%) on the DSST, 2 (8.3%) on TMT‑A, and 1 (4.2%) on TMT‑B, with none exceeding chance and no significant group‑level effects. Reliable decline (RCI < −1.645) occurred for 2/24 participants (8.3%) on the DSST under both reference frameworks, but this proportion did not exceed chance; no reliable decline was observed on TMT‑A or TMT‑B. Overall, the researchers report modest, short‑term group‑level improvements on processing speed and executive function measures following a single 25 mg psilocybin dose; some TMT changes remained significant after controlling for depressive symptom change, whereas DSST gains appeared partly explained by mood improvement. However, only a minority of individuals showed RCI‑defined reliable improvement, and proportions did not significantly exceed chance.

The authors interpret their findings as preliminary evidence that psilocybin‑assisted psychotherapy (PAP) may be associated with modest, short‑term improvements in tasks tapping processing speed and executive function in people with treatment‑resistant depression. Improvements on DSST and TMT‑A appeared as early as one day and persisted to two weeks, while TMT‑B and the executive index (TMTB‑A) reached significance at two weeks. The attenuation of DSST effects when controlling for MADRS suggests that improvements on that test were at least partly mediated by concurrent mood change, whereas TMT findings remained significant after adjustment, indicating the possibility of mood‑independent cognitive effects. The findings are positioned relative to the sparse prior literature: they broadly align with few previous trials that reported modest or domain‑specific cognitive changes following psilocybin, and extend them by including an earlier (one‑day) assessment. The absence of evidence for short‑term cognitive deterioration is noted as supporting the short‑term cognitive safety of psilocybin and has practical implications for post‑dose monitoring and activity restrictions. The authors acknowledge several important limitations that constrain interpretation. Chief among these are the small sample size, lack of placebo control and blinding, limited cognitive battery (only DSST and TMTs), short follow‑up (two weeks), practice effects from repeated testing with identical forms, and a DSST administration error requiring prorating from 90 s to 120 s for normative comparisons. These factors raise the possibility that at least part of the observed improvements reflect practice or measurement variability rather than true procognitive effects. The sample’s demographic homogeneity (predominantly white and highly educated) and predominance of MDD over BD‑II further limit generalisability. The authors also note unequal test exposure across randomised groups and that several potentially relevant covariates (age at depression onset, inflammatory markers, metabolic comorbidities, detailed treatment history, childhood adversity) were not controlled for. They further discuss the inability of the design to disentangle the contributions of psychotherapeutic support versus psilocybin itself. For future work the authors recommend adequately powered randomized controlled trials with larger, more diverse samples, expanded cognitive batteries covering additional domains, longer follow‑up to assess durability, inclusion of patient‑reported outcome measures, pre‑registration of cognitive endpoints, and incorporation of neurobiological measures (for example fMRI, EEG) and analytic approaches (for example mediation or latent growth models) to clarify mechanisms and whether cognitive change is independent from, contributes to, or results from mood and functional recovery. They emphasise that cognition should be assessed as an independent treatment domain in psychedelic research rather than solely as a secondary mood‑related endpoint.

In this exploratory post‑hoc analysis E. and colleagues report that a single 25 mg psilocybin dose given within a brief psychotherapeutic framework was associated with modest, short‑term improvements on measures of executive function and processing speed in people with treatment‑resistant depression. However, given the small sample size, absence of a placebo control, short follow‑up and potential practice effects (including a DSST administration error), the authors conclude these results should be interpreted cautiously. The observed changes may reflect practice‑related gains or other non‑specific influences rather than a direct procognitive effect of psilocybin, and controlled trials with more comprehensive designs are required to determine whether cognition is a meaningful and distinct therapeutic target of psilocybin‑assisted therapy.