Psilocybin’s acute and persistent brain effects: a precision imaging drug trial

Psilocybin (PSIL) is a serotonin 2A (5-HT2A) agonist that has shown rapid and sometimes persistent benefits in clinical trials for conditions such as depression, end-of-life anxiety, obsessive–compulsive disorder, eating disorders, and alcohol use disorder. Previous human neuroimaging studies of psychedelics have reported acute decreases in functional connectivity within the default mode network and reduced fMRI signal power across cortex, but they have faced important limitations including inter-individual variability, head motion, autonomic changes, and potential vascular/neurovascular confounds that complicate interpretation of acute drug effects. Subramanian and colleagues aimed to address these limitations by applying a "precision imaging drug trial" (PIDT) approach, which builds on precision functional mapping (PFM) through dense repeated resting-state fMRI sampling, individualized network mapping, physiological monitoring, and advanced multi-echo acquisition and denoising. The study used a randomised cross-over design in healthy adults to compare acute and persistent brain effects of a single 25 mg oral dose of psilocybin versus an active control (40 mg methylphenidate, MTP), and included longitudinal follow-up and a replication protocol in a subset of participants. The dataset combines multi-echo resting and task fMRI, structural MRI, diffusion basis spectrum imaging, and detailed subjective and physiological measures, and is released as a resource to study individual differences and common network-level effects of psilocybin.

Design and participants. The study used a randomised, double-blind cross-over PIDT design in seven healthy adults (mean age 34.1 years, SD 9.8; 3 females; 6 Caucasian) recruited between March 2021 and March 2023. Eligibility required age 18–45 and at least one prior lifetime psychedelic exposure more than six months before enrolment; exclusions included MRI contraindications, major medical or psychiatric conditions, current psychotropic medications, pregnancy, prior adverse psychedelic reactions, and first-degree family history of psychosis. Four participants later returned for an open-label replication PSIL dose 6–12 months after the initial protocol. Interventions, blinding and facilitation. Each participant completed two drug sessions one to two weeks apart, receiving either PSIL 25 mg or MTP 40 mg per session. MTP was chosen as an active control to approximate cardiovascular and arousal effects. Both participants and facilitators were blinded during the cross-over. Dosing sessions lasted 7–8 hours and involved preparatory and post-session integration meetings; two trained facilitators (a clinician and a trainee) attended each dosing day. Participants completed mindfulness, wore eye shades and headphones with a curated music sequence, and were moved to the MRI suite about 60 minutes after ingestion for imaging during peak drug concentration. Imaging acquisition. MRI was performed on a Siemens Prisma. Sessions included high-resolution structural scans (T1/T2 at 0.9 mm isotropic), multi-echo EPI resting state scans (two or more 15-minute runs, 513 frames each) and two task fMRI runs (auditory–visual matching, 410 seconds each). Multi-echo parameters comprised five echo times (approximately 14.2, 38.9, 63.7, 88.4, 113.1 ms) with TR 1761 ms and 72 axial slices. Diffusion acquisitions optimised for diffusion basis spectrum imaging (DBSI) included b = 1500 and 3000 with 102 directions each; diffusion scans during PSIL were omitted when needed for participant comfort or time constraints. FIRMM head-motion monitoring and eye tracking were used during scanning; a subset of sessions included pulse oximetry and respiratory belt recordings for physiological regression. Behavioural and physiological measures. Subjective measures included the MEQ30 (mystical experience), Mini-IPIP (Big Five personality), Challenging Effects Questionnaire, and Emotional Breakthrough Inventory, administered at baseline and post-drug time points. Physiological signals were recorded at 400 Hz and preprocessed with PhysIO; instantaneous pulse and respiratory rates were calculated and used to generate regressors for nuisance removal in fMRI analyses. Preprocessing and analysis pipeline. Resting fMRI preprocessing included NORDIC thermal noise removal, slice timing correction, affine motion correction and registration, de-banding, spin-echo fieldmap correction (FSL top-up), gain field correction, FreeSurfer segmentation for tissue regression, temporal filtering (0.009–0.08 Hz), and frame censoring at FD > 0.30 mm. Multi-echo echoes were combined after registration using a weighted summation approach. Processed data were sampled into a 385-region parcellation (324 cortical, 61 subcortical) for functional connectivity analyses. Key analyses reported for the dataset included parcel-level FC similarity matrices, infra-slow spectral power (0.009–0.06 Hz), and network modularity (Newman's Q). A linear mixed-effects modelling framework with random effects for participant and MRI session was used to test drug-condition effects on modularity and infra-slow power; similarity matrices were compared via Pearson correlation of vectorised FC matrices.

Participants and data availability. Seven participants completed the main cross-over protocol and consented to public data sharing; four participants completed a replication PSIL session 6–12 months later. All raw and processed imaging and behavioural data are publicly available on OpenNeuro in BIDS format, organised by session with accompanying session notes and behavioural files. Subjective effects and adverse events. Mystical experience scores (MEQ30) were significantly higher on PSIL than on MTP across the four MEQ domains (p < 0.05). Mean domain scores (out of 5) for PSIL versus MTP were: mystical 3.4 versus 0.61, positive mood 3.8 versus 1.2, transcendence 3.8 versus 0.33, and ineffability 3.7 versus 0.86. One participant (P3) reported higher MEQ30 subscale scores on MTP than PSIL during the cross-over, but on return for the replication PSIL session they reported higher mystical scores consistent with other participants. Side effects 4–6 hours after PSIL ingestion included headache in one participant (14%), nausea in two participants (28%), and anxiety in two participants (28%). No serious adverse events occurred and no participants required rescue medication. Participant blinding was generally preserved; only P3 reported an incorrect treatment guess. Imaging data quantity and motion. The PIDT strategy produced a comparatively large dataset per participant: across protocol and replication visits participants provided an average of 39.4 (SD 12.2) off-drug 15-minute resting-state scans. On average there were 4.7 (SD 2.0) usable 15-minute resting scans per participant on PSIL and 2.7 (SD 0.8) on MTP. Using a stringent cutoff for usable scans (average FD ≤ 0.2 mm), 86% (6/7) of participants had at least one usable 15-minute PSIL resting scan; participant P2 lacked usable PSIL resting data. The authors note improved data quality relative to earlier psychedelic fMRI datasets. Physiological measures. Baseline session means were a pulse rate of 72 bpm (95% CI 68–77) and a respiratory rate of 11 respirations per minute (95% CI 9–13). Drug sessions produced changes in physiological signals relative to baseline; the text reports that MTP was associated with an average 16.7 bpm increase in pulse rate, although the accompanying confidence interval is not clearly reported in the extracted text. Non-drug sessions showed no consistent changes. Connectomics and network metrics. Whole-brain FC similarity matrices revealed two principal sources of similarity: within-participant across sessions and within-condition (for example, PSIL-to-PSIL). The PSIL condition was consistently less like other conditions within-subjects, indicating a distinct acute network signature, while PSIL–PSIL comparisons showed similarity within and across participants suggesting shared effects of PSIL. Network modularity (a measure of network segregation) decreased substantially from baseline during PSIL (P = 1.8754 × 10^-9) but did not remain significantly altered in the weeks following PSIL (P = 0.68). MTP produced no significant change in modularity (P = 0.85). Intra-participant reliability. Among the four participants who returned for the replication PSIL dose, whole-brain network changes were similar between the initial and replication PSIL sessions: similarity to replication r = 0.78, similarity across participants r = 0.45, P < 0.001. One participant (P2) was excluded from some analyses because of excessive motion leading to unusable data.