Psilocybin for depression and anxiety in cancer

The authors worry that their comparator was not actually a placebo, so part of the measured dose difference may be a high dose compared against a weakly active drug rather than against nothing, and they propose a dose 100 times lower.

Question

Does 1 mg/70 kg psilocybin produce measurable pharmacodynamic engagement relative to an inert capsule, and does 0.01 mg/70 kg produce none?

Hypothesis

1 mg/70 kg psilocybin produces a detectable pupillary dilation response relative to inert placebo, whereas 0.01 mg/70 kg does not differ from inert placebo on any marker.

Attempt type

wet-lab
Vocabulary note: This is a human pharmacodynamic dose-ranging study. None of the five permitted attemptType values describes a clinical drug administration study in people; wet-lab is the closest because the work is empirical measurement on human subjects rather than computation or synthesis.

Design

A four-arm, double-blind, randomised single-dose study in adults, run in the same supportive session format the 2016 trial used. Arms: inert capsule, psilocybin 0.01 mg/70 kg, psilocybin 1 mg/70 kg, and psilocybin 25 mg as an in-study activity check. The primary measure is objective and not self-reported: pupillary diameter by infrared pupillometry at fixed intervals across the session, because Levy-Cooperman and colleagues showed pupil dilation rises with dose even where subjective effects do not. Secondary measures are the Mystical Experience Questionnaire 30-item version (MEQ30), the Hallucinogen Rating Scale, plasma psilocin, and the per-session monitor dose guess on the same 10 cm line the 2016 trial used, plus a participant allocation guess with a confidence rating. Analysis is a pre-registered ordered contrast across the three low arms on peak pupillary change, with the 25 mg arm read first as a gate.

Controls

• vehicle Matched capsule containing no psilocybin, given with the same instruction used in 2016 that psilocybin would be administered at every session.
  Rules out: That pupillary or subjective change at 0.01 or 1 mg is produced by the capsule, the room, the monitors and the expectancy instruction rather than by psilocybin.
  Vocabulary note: The inert capsule is a true placebo rather than a pharmacological vehicle for a dissolved agent, but vehicle is the closest available kind.
• internal-positive A 25 mg psilocybin arm run in the same sessions on the same instruments.
  Rules out: That a flat result at 0.01 and 1 mg means inactivity, when it could instead mean the pupillometry and questionnaire assays were insensitive on the day. Without separation in this arm the low-dose nulls are uninterpretable.
• within-subject Three pre-dose pupillometry readings per participant on the session morning, plus a no-drug practice session, to fix each person's own baseline and diurnal drift.
  Rules out: That a small group-level pupillary difference is really ordinary within-person variation in pupil size across a long session.

Sensitivity

The design can be anchored on a published threshold but not yet powered to a number. Levy-Cooperman and colleagues dosed 0.5, 1.0, 1.5, 2.5, 3.5 and 4.0 mg with six participants per dose and 14 on placebo, and found subjective effects separated from placebo only at 2.5 mg and above, with 0.5 to 1.5 mg subjectively indistinguishable. That sets the subjective detection threshold, so this design must stand on pupillometry instead, and the per-dose pupillary means and standard deviations needed to size the arms are not in the abstract or results text obtained. The honest statement is that the sensitivity cannot be set from what is published: it needs the per-dose pupillometry variance from that study, or a pilot of about 10 participants on the inert arm to get the within-session standard deviation of peak pupillary change.

Would count as

Supported → positive · opposed → negative-no-effect
Inconclusive if: The 25 mg arm fails to separate from inert placebo on pupillometry and MEQ30, which would mean the assay or the session conditions, not the drug doses, produced the flat low-dose result.

Already done

• In 56 healthy adults given single doses of 0.5, 1.0, 1.5, 2.5, 3.5 or 4.0 mg psilocybin or placebo, subjective effects became distinguishable from placebo only at 2.5 mg and above, while pupil dilation increased with dose across the range, indicating pharmacological engagement below the subjective threshold. (Levy-Cooperman et al. 2026, Pharmaceuticals, doi:10.3390/ph19091496)
  So: This is the strongest reason to make pupillometry rather than any questionnaire the primary measure, and it means a subjective null at 1 mg would prove nothing about pharmacological inactivity.
• The largest psilocybin randomised trial used 1 mg as its control arm in 233 participants with treatment-resistant depression, and that arm moved the Montgomery-Asberg Depression Rating Scale by -5.4 points (95% confidence interval -8.1 to -2.7) at week 3. (Goodwin et al. 2022, New England Journal of Medicine, doi:10.1056/NEJMoa2206443)
  So: The field adopted 1 mg as a control rather than resolving the authors' doubt about it, so a clean answer now has direct bearing on how a decade of control arms should be read.
• A systematic review of low-dose psilocybin found that in the 1 mg arm 18% of participants showed a significant response and 8% achieved remission, with only 10% maintaining the improvement by week 12. (Meshkat et al. 2026, Psychedelic Medicine, doi:10.1089/psymed.2024.0039)
  So: The non-trivial but non-durable 1 mg response pattern is what the design must discriminate between a weak drug effect and an expectancy effect, which is why an inert arm under identical instruction is the load-bearing control.
• A two-centre trial in treatment-resistant depression ran psilocybin 5 mg and nicotinamide 100 mg as two separate comparators and found week-6 response of 12.5% versus 10.6% and a Hamilton Rating Scale for Depression 17-item change difference of -1.50 (95% confidence interval -3.91 to 0.90) between them. (Mertens et al. 2026, JAMA Psychiatry (EPISODE trial), doi:10.1001/jamapsychiatry.2026.0132)
  So: This is the nearest thing to the proposed comparison already run, and it found no clinical separation between a low psilocybin dose and a near-inert comparator, which is evidence that the 2016 concern may not matter clinically even if it matters pharmacologically.
• Of 112 psychedelic randomised controlled trials, only 29.5% assessed blinding integrity at all, while 57.1% cited blinding as a limitation, and functional unblinding was substantial for psilocybin. (Orsini et al. 2026, JAMA Psychiatry, doi:10.1001/jamapsychiatry.2026.0255)
  So: The 2016 trial's monitor dose-guess measurement is still better than most of the field, so this design should keep that exact 10 cm line measure to stay comparable rather than invent a new blinding index.

Searched: Searched OpenAlex for psilocybin blinding, placebo and expectancy from November 2016, Europe PMC for low-dose and subperceptual psilocybin dose-finding, and the web for any trial using a 0.01 mg psilocybin dose; the 0.01 mg dose the authors proposed appears never to have been administered, and I found no study that measured pharmacodynamic markers at 1 mg against a non-psilocybin inert capsule under matched expectancy instruction.

Feasibility

A phase 1 clinical pharmacology unit with a Schedule 1 licence and a psilocybin supply, or any of the existing academic psychedelic research centres with an established session-monitor team · 9 to 15 months · moderate, because each arm needs a full supervised session day with two monitors even though the pharmacodynamic measures themselves are cheap
Blockers: Compounding and verifying a 0.01 mg/70 kg dose, roughly 0.7 micrograms, to acceptable content uniformity may be at or past the limit of ordinary capsule manufacture; Plasma psilocin at 0.01 mg/70 kg will almost certainly sit below the lower limit of quantification of standard assays, so the pharmacokinetic confirmation that the dose was delivered may be unobtainable; Schedule 1 licensing and ethics review for administering a drug dose whose entire purpose is to be inactive

Strongest objection

A clinical pharmacologist would say the question is not worth answering as posed. Whether 1 mg has detectable receptor engagement is almost beside the point: what matters is whether it has a clinical effect, and the EPISODE trial already compared a low psilocybin dose against a near-inert comparator in 95 patients and found no clinical separation. Finding a few tenths of a millimetre of pupillary dilation at 1 mg would be a pharmacology result with no bearing on how the 2016 effect sizes should be read, so the design risks answering the literal question while leaving the real one untouched.

Not derivable from what is published

• The per-dose pupillary diameter means and standard deviations from Levy-Cooperman et al. 2026, which are needed to size the arms
• Whether a 0.01 mg/70 kg capsule can be manufactured with verifiable content uniformity, and at what cost
• The lower limit of quantification for plasma psilocin on current assays, which decides whether dose delivery can be confirmed at all at 0.01 mg/70 kg
• Whether the 2016 trial's monitor dose-guess ratings are available at session level for a direct comparison against new data

The authors note two separate holes they left: after everyone crossed over at 5 weeks there was no longer an unexposed group to compare against, so the 6-month durability numbers rest on a single pooled arm, and two of the instruments carrying the durability claim had never been psychometrically validated.

Question

Does the high-dose psilocybin advantage over an active placebo persist to 6 months when the blind and the parallel groups are both maintained, rather than dissolved at 5 weeks?

Hypothesis

With no crossover and the blind held to 6 months, high-dose psilocybin retains a between-group advantage on the GRID Hamilton Depression Rating Scale 17-item version of at least a moderate size at 6 months.

Attempt type

wet-lab
Vocabulary note: Again a human randomised clinical trial with an embedded measure-validation sub-study. The five permitted values have no entry for a clinical trial or for psychometric validation, so wet-lab is used as the closest fit for empirical work on human subjects.

Design

A two-arm parallel-group double-blind trial in distressed patients with life-threatening cancer, with no crossover, randomising to psilocybin 25 mg or an active placebo, each with identical preparatory and integration support. Blinded assessments at baseline, 5 weeks, 3 months and 6 months, using an independent rater pool that never attends a session and never meets the monitors. Primary endpoint is the between-group GRID Hamilton Depression Rating Scale 17-item score at 6 months; the 5-week comparison is pre-specified as a gate that must reproduce the 2016 result before the 6-month result is interpreted. Embedded validation sub-study: the Community Observer Questionnaire and the Persisting Effects Questionnaire are administered alongside the validated clinician measures, with test-retest in the placebo arm, and are calibrated against the newly published Persisting Effects Questionnaire short form. Pre-registered stopping rule for futility at the 3-month interim.

Controls

• vehicle An active placebo such as niacin 100 mg or nicotinamide 100 mg, delivered with the same session structure and the same number of therapist contact hours.
  Rules out: That the 6-month benefit comes from the many hours of professional psychological support, the attention and the ceremony surrounding dosing, rather than from psilocybin.
  Vocabulary note: The comparator is an active placebo chosen to mimic somatic sensation, not an inert pharmacological vehicle, so vehicle is an imperfect fit.
• internal-positive The pre-specified 5-week between-group comparison, which must reproduce the 2016 between-group effect before the 6-month endpoint is read.
  Rules out: That a flat 6-month result means the benefit genuinely faded, when it could mean this sample, this therapist team or this site never produced the acute effect in the first place. This is the control that makes a 6-month null count as evidence of absence.
• designated The already-validated GRID Hamilton Depression Rating Scale and Hospital Anxiety and Depression Scale, scored in parallel with the two unvalidated instruments on the same participants at the same visits.
  Rules out: That agreement or disagreement of the Community Observer Questionnaire with clinical change is an artefact of the instrument, by giving it a validated yardstick to be calibrated against.
• within-subject Two baseline assessments one to two weeks apart before randomisation on every measure, including the observer questionnaire.
  Rules out: That apparent change is regression from a single unusually bad baseline day, and supplies the test-retest reliability the observer questionnaire has never had.

Sensitivity

This one is derivable from the record. The 2016 trial reports a mean between-group Cohen's d of 0.82 at Post-session 1 across the 11 measures meeting its conservative criteria. For a two-sample comparison at two-sided alpha 0.05 and 80% power, detecting d = 0.82 needs 24 per arm, which is almost exactly the 25 and 26 per group the 2016 trial had, so that trial was powered for the effect it found and for nothing smaller. A durable 6-month between-group effect is expected to be smaller than the 5-week one, and the design should be sized for the smaller target: d = 0.60 needs 44 per arm (88 total) and d = 0.50 needs 63 per arm (126 total). At 50 per arm the design detects d = 0.56. The attrition rate to add on top cannot be taken from the record, because the 2016 trial's losses came from a crossover design and a 6-month window, not a 6-month parallel one.

Would count as

Supported → positive · opposed → negative-no-effect
Inconclusive if: The blind is functionally broken, with participants and raters guessing allocation well above chance, or attrition from disease progression exceeds the planned allowance so that the 6-month groups are no longer the randomised groups.

Already done

• The Persisting Effects Questionnaire has now been formally validated for the first time, and its original factor structure was not quantitatively supported; factor analysis instead produced a two-factor positive and negative structure, and a validated 14-item short form was published. (Ehrenkranz et al. 2026, Journal of Psychopharmacology, doi:10.1177/02698811261473405)
  So: Half of this issue is now answered, and answered against the 2016 instrument: the questionnaire that carried the persisting-effects claim has a structure that does not hold up, so any new trial should use the short form and should not reuse the original subscales.
• A parallel-group double-blind phase 2b trial in adults with life-threatening illness compared psilocybin 25 mg with niacin 100 mg as active placebo, held the double blind for 6 to 7 weeks, then opened to a label extension with follow-up to 26 weeks, at a single palliative care site. (Ross et al. 2025, General Hospital Psychiatry, doi:10.1016/j.genhosppsych.2025.08.001)
  So: This resolves the parallel-group design question but not the durability one: by opening the label at 6 to 7 weeks it reproduces the 2016 limitation almost exactly, which is the gap this design is built to close.
• A parallel-group trial of 233 participants with treatment-resistant depression held the blind to week 12 and reported sustained response, defined as meeting response criteria at week 3 and every subsequent visit, as a secondary endpoint. (Goodwin et al. 2022, New England Journal of Medicine, doi:10.1056/NEJMoa2206443)
  So: It shows a 12-week blinded parallel comparison is practically achievable, so 6 months is an extension of an existing practice rather than a new idea, and its sustained-response definition is the one to borrow.
• Two-year follow-up of 30 patients with cancer and major depressive disorder given a single 25 mg dose found 15 of 28 (53.6%) with significant depression reduction, averaging -15.0 points on the Montgomery-Asberg Depression Rating Scale, and 14 (50%) with sustained reduction. (Agrawal et al. 2025, Cancer, doi:10.1002/cncr.35889)
  So: Long-term durability data in cancer patients now exist but are uncontrolled, which is precisely why a blinded parallel comparator at 6 months is still the missing piece rather than more long follow-up.

Searched: Searched OpenAlex and Europe PMC for parallel-group and blinded long-follow-up psilocybin trials, for Persisting Effects Questionnaire validation, and for any psychometric work on the Community Observer Questionnaire; the Persisting Effects Questionnaire validation exists as of 2026, but I searched Europe PMC for the Community Observer Questionnaire and found 11 hits none of which is a validation study, so that half of the issue appears still untouched.

Feasibility

A palliative care or psycho-oncology department with an established psilocybin programme, such as the groups that ran the existing cancer trials, ideally two or three of them together to reach the sample size · 3 to 4 years from first enrolment to the 6-month endpoint of the last participant · high, because every participant needs the full preparation, session and integration therapist time and the active placebo arm costs the same in staff hours as the drug arm
Blockers: Ethical pressure to open the label early, which is what collapsed the blind in both the 2016 trial and the 2025 phase 2b trial, and which an independent monitoring board may impose regardless of the protocol; Mortality and disease progression in a life-threatening-cancer sample over 6 months, which creates informative censoring that no analysis fully repairs; Keeping an independent rater pool genuinely separated from the session monitors at a single site

Strongest objection

A palliative care clinician would object that holding a dying patient on an active placebo for 6 months is not defensible. The reason both earlier trials crossed over or opened the label is not methodological carelessness, it is that in a population with limited life expectancy the ethical case for withholding an apparently effective treatment for half a year is weak, and an ethics committee may simply refuse. The design may therefore be the right experiment and still be unrunnable in this population, which would push it into a life-threatening-illness group with longer prognosis and thereby change the question.

Not derivable from what is published

• The expected 6-month between-group effect size under a maintained blind, since every existing 6-month figure in this population comes from pooled or open-label data
• The attrition and mortality rate to plan for over 6 months in a parallel design, which the 2016 crossover cannot supply
• Any psychometric property of the Community Observer Questionnaire, including its test-retest reliability, since no validation study appears to exist
• Whether an ethics committee would approve a 6-month blinded placebo arm in this population at all

This is the authors' own stated next step, in the Discussion and again in the Conclusions: run the same trial bigger and across several sites, in patients who are not overwhelmingly White and highly educated, to find out whether the effect generalises.

Question

Is the between-group effect of high-dose psilocybin on cancer-related depression as large in racially and ethnically minoritised and less-educated patients as in the White, highly educated sample of 2016?

Hypothesis

In a multisite sample with enrolment stratified by race, ethnicity and education, the treatment effect on depression is not smaller in minoritised or less-educated strata than in the stratum matching the 2016 sample.

Attempt type

wet-lab
Vocabulary note: A multisite human randomised trial. No permitted attemptType value covers a clinical trial, so wet-lab is used for the same reason as in the other designs.

Design

A multisite randomised double-blind trial of psilocybin 25 mg against an active placebo in patients with cancer and clinically significant depression, with enrolment stratified at each site so that at least 40% of participants are from racially or ethnically minoritised groups and at least 40% have no four-year degree, enforced by capping the over-represented stratum rather than by hoping. Sites are chosen for catchment rather than for prior psychedelic experience, and include safety-net and community oncology settings. Therapist teams are recruited to reflect the catchment and certified against audio-recorded sessions. The primary analysis is the pooled treatment effect on depression at 5 weeks; the pre-specified secondary analysis that justifies the design is the treatment-by-stratum interaction. Monitoring includes blinded central rating, site-level effect reporting and a pre-registered rule for enrolment pause if a stratum target lags.

Controls

• vehicle Active placebo arm with identical session structure and therapist contact hours, randomised within each stratum and within each site.
  Rules out: That a smaller apparent effect in a given stratum reflects a different response to psilocybin, when it could reflect a different response to the supportive therapy, the setting or the research relationship.
  Vocabulary note: An active placebo rather than a pharmacological vehicle.
• internal-positive The stratum demographically matching the 2016 sample, namely White and college-educated participants, analysed as a pre-specified subgroup.
  Rules out: That a small or absent effect in minoritised strata means psilocybin works less well there, when it could mean the whole trial failed to reproduce the original effect for reasons of site, era, therapist or comparator. If this stratum does not reproduce roughly d = 0.82, no moderation claim can be made.
• designated Central blinded raters scoring audio recordings, with a designated set of pre-scored calibration recordings each rater must match before and during the trial.
  Rules out: That between-site or between-stratum differences in measured outcome are rater drift or differential rater behaviour rather than real differences in response.

Sensitivity

Derivable, and the number is the uncomfortable part. Using the 2016 mean between-group Cohen's d of 0.82, a two-sample test at two-sided alpha 0.05 and 80% power needs 24 per arm, 48 in total, to detect the main effect. A treatment-by-stratum interaction of the same magnitude as the main effect needs roughly four times that, about 94 per arm per stratum, which is about 376 participants across the four cells. Detecting a more plausible interaction, where the effect halves in one stratum, pushes this past 1000. So a trial sized for the main effect, which is what trials of this kind are usually sized for, cannot answer the generality question the authors raised; it can only show that the pooled effect survives in a mixed sample. Either the trial is sized for the interaction or the generality claim must come from pooling individual participant data across trials.

Would count as

Supported → positive · opposed → partial
Inconclusive if: The stratum enrolment targets are not met, so the interaction test runs on cells too small to detect even a halving of the effect, which is the most likely outcome given the field's recruitment record.

Already done

• Across 18 psychedelic studies from 1993 to 2017 comprising 282 participants, 82.3% were non-Hispanic White, 2.5% African-American, 2.1% Latino and 1.8% Asian, and there were no significant differences in recruitment methods between studies with higher and lower inclusion. (Michaels et al. 2018, BMC Psychiatry, doi:10.1186/s12888-018-1824-6)
  So: It quantifies the problem the authors flagged and shows that recruitment method alone did not fix it, which is why this design enforces stratum caps rather than relying on outreach.
• Racial and ethnic diversity in psychedelic science remains lacking as of 2026 despite efforts to increase representation. (Hughes 2026, Psychiatric Services, doi:10.1176/appi.ps.20250169)
  So: Ten years after the 2016 paper the diversity half of this next step is still open, so the design cannot assume the field has drifted into representativeness on its own.
• The largest psilocybin trial to date randomised 233 participants with treatment-resistant depression across multiple international sites, and a separate phase 2 trial randomised 104 participants with major depressive disorder across 11 United States research sites. (Goodwin et al. 2022, New England Journal of Medicine; Raison et al. 2023, JAMA, doi:10.1056/NEJMoa2206443 and doi:10.1001/jama.2023.14530)
  So: The multisite half of this next step has been delivered, but in depression populations rather than in cancer, so the generality of the 2016 cancer result specifically is still untested at scale.
• Two multisite psilocybin trials in cancer-related depression are currently enrolling and larger than the 2016 trial: NCT06319378 with a target of 100 participants across 4 sites, started April 2024, and NCT05398484, a phase 2/3 trial with a target of 200 across 2 sites, started May 2023. No results are published for either. (ClinicalTrials.gov registry query, October 2026, url:https://clinicaltrials.gov/study/NCT06319378 and https://clinicaltrials.gov/study/NCT05398484)
  So: This is the decisive constraint: proposing a fresh multisite cancer trial now duplicates work already running, so the defensible contribution is the stratified enrolment and the pre-registered moderation analysis that neither registration appears to promise, plus harmonising outcome measures with them so the data can be pooled.
• A two-centre phase 2b trial of psilocybin 25 mg in treatment-resistant depression randomised 144 participants and did not meet its primary endpoint, with week-6 response of 17.0% versus 10.6% for placebo and an adjusted odds ratio of 1.73 (95% confidence interval 0.53 to 6.23). (Mertens et al. 2026, JAMA Psychiatry (EPISODE trial), doi:10.1001/jamapsychiatry.2026.0132)
  So: A larger and more rigorous trial returning an inconclusive result is the live warning that scaling up may shrink rather than confirm the 2016 effect, so the design needs the matched-stratum positive control to tell shrinkage from non-generality.

Searched: Searched OpenAlex and Europe PMC for multisite psilocybin trials and for diversity and inclusion in psychedelic trials, and queried the ClinicalTrials.gov application programming interface for all psilocybin cancer depression trials with their enrolment counts and site counts; of 21 trials returned, only NCT06319378 and NCT05398484 are both multisite and larger than the 2016 trial, and neither has published results, so the authors' next step is under way but not yet answered.

Feasibility

A cooperative group or network able to coordinate community and safety-net oncology sites, rather than a single academic psychedelic centre, plus a funder willing to pay for interaction-level power · 4 to 6 years · high, because interaction-level power means several hundred participants each needing many hours of certified therapist time, and because standing up psychedelic capability at non-academic community sites is itself a large cost
Blockers: Two multisite cancer trials are already enrolling, so funding a third may be unjustifiable and the better route may be prospective harmonisation with them; Therapist certification at community and safety-net sites, where there is no existing pool of trained session monitors; Sample size: the interaction the authors care about needs roughly four times the participants of the main effect, and no funder has paid for that in this field; Well-documented and legitimate mistrust of research among some of the communities the design needs to enrol, which caps enrolment independently of effort

Strongest objection

A trialist would say the design confuses two questions and is unaffordable as a result. Showing that psilocybin works in a diverse sample needs a few hundred participants; showing that it works equally well across strata needs over a thousand, and nobody will fund that for a question that is about external validity rather than efficacy. The pragmatic objection is that the generality question belongs in a pre-registered individual participant data meta-analysis across the trials now completing, not in a new trial, and that a new trial sized for the main effect would answer the easier question while appearing to answer the harder one.

Not derivable from what is published

• The expected effect size in minoritised or less-educated strata, since no trial has reported one, so the interaction size the design should target is a guess
• Achievable enrolment proportions for each stratum at community and safety-net oncology sites, which decides whether the stratum caps are realistic or merely aspirational
• The published outcome measures and analysis plans of NCT06319378 and NCT05398484, which decide whether harmonisation and pooling are possible
• Whether a therapist workforce reflecting the target catchment can be recruited and certified at the required scale

The authors can show a clinical effect lasting 6 months but have no account of what in the brain carries it, and they name the specific gap as the link between enduring clinical benefit and long-term changes in cortical networks or neuroplasticity.

Question

Does a persistent reduction in anterior hippocampus to default mode network connectivity measured at 6 months track the sustained antidepressant and anxiolytic response in cancer patients?

Hypothesis

Participants who retain clinical response at 6 months show a greater persisting reduction in anterior hippocampus to default mode network connectivity from baseline than non-responders and than active placebo participants.

Attempt type

wet-lab
Vocabulary note: A longitudinal human neuroimaging study nested in a clinical trial. It is neither wet bench work nor computation nor synthesis; wet-lab is the closest permitted value for empirical measurement on human subjects, and the misfit is worth recording.

Design

A neuroimaging sub-study nested in a parallel-group psilocybin trial in cancer patients with depression, using longitudinal precision functional mapping rather than single scans: five pre-dose baseline scans over two weeks, a scan on dosing day, then scans at 1 week, 5 weeks, 3 months and 6 months, in both the psilocybin and active placebo arms. The pre-registered primary measure is functional connectivity between the anterior hippocampus and the default mode network, the connection Siegel and colleagues found persistently reduced for weeks in healthy adults. Secondary measures are whole-brain network modularity, following Daws and colleagues, and plasma brain-derived neurotrophic factor at matched timepoints. The analysis relates each participant's 6-month connectivity change to their 6-month clinical response, with the dosing-day scan read first as a gate. Scanner, sequence and head-motion handling are fixed in advance.

Controls

• internal-positive The dosing-day scan, which must reproduce the large acute desynchronisation of functional connectivity already established for 25 mg psilocybin before any 6-month result is interpreted.
  Rules out: That a flat 6-month connectivity result means no lasting network change, when it could mean the imaging protocol, the scanner or this patient group simply could not detect psilocybin's effect on the brain at all. Without this gate a 6-month null is uninformative rather than negative.
• designated An arousal-matched active comparator arm. Siegel and colleagues used methylphenidate 40 mg for this purpose; in a patient trial the active placebo arm serves, with the designated comparator recordings used to calibrate.
  Rules out: That the connectivity change is a non-specific consequence of taking a psychoactive drug, of arousal, or of an intense emotional day in a scanner-adjacent clinical setting.
• within-subject Five pre-dose baseline scans per participant across two weeks, establishing each individual's own connectivity variability before any drug.
  Rules out: That a 6-month connectivity difference is ordinary within-person scan-to-scan variation, which in cancer patients will be inflated by fatigue, pain medication and motion. This is also the only way to set a per-participant detection threshold.

Sensitivity

This cannot be derived from what is published, and saying so is the honest answer. Siegel and colleagues report that psilocybin acutely produced more than threefold greater functional connectivity change than methylphenidate and that the anterior hippocampus to default mode network reduction persisted for weeks, but their abstract gives no effect size for the persisting change, their follow-up window was 3 weeks rather than 6 months, and their participants were healthy adults scanned about 18 times each. Three things are needed to set a sensitivity: the magnitude and variance of the persisting connectivity reduction at the longest timepoint measured, the within-participant connectivity variance in a cancer population under treatment, and the attrition-adjusted number of usable 6-month scans. The first could come from the Siegel data on request; the second needs a precision functional mapping pilot of roughly 8 to 10 cancer patients; the third needs the parent trial's retention.

Would count as

Supported → positive · opposed → negative-no-association
Inconclusive if: The dosing-day scan fails to show the established acute desynchronisation, or head motion and attrition leave too few usable 6-month scans, in which case the result says nothing about brain mechanism either way.

Already done

• Longitudinal precision functional mapping with roughly 18 magnetic resonance imaging visits per participant found that 25 mg psilocybin acutely disrupted functional connectivity more than threefold more than methylphenidate 40 mg, driven by desynchronisation across spatial scales and strongest in the default mode network, and caused a persistent decrease in connectivity between the anterior hippocampus and the default mode network lasting for weeks. (Siegel et al. 2024, Nature, doi:10.1038/s41586-024-07624-5)
  So: This supplies the specific connection to pre-register as the primary measure and the methylphenidate comparator logic, and it is the reason the design uses many scans per person rather than a before-and-after pair.
• In two clinical trials of depression, antidepressant response to psilocybin was rapid and sustained and correlated with decreases in functional magnetic resonance imaging brain network modularity, measured 1 day after a 25 mg dose in one trial and 3 weeks after the second dose in the other. (Daws et al. 2022, Nature Medicine, doi:10.1038/s41591-022-01744-z)
  So: It establishes network modularity as a measure that already correlates with clinical response, so it belongs as the secondary measure, and it sets the current state of the art at 3 weeks, which is what a 6-month measurement would extend.
• A single dose of psilocybin produced roughly 10% increases in dendritic spine size and density in mouse medial frontal cortex within 24 hours, driven by an elevated spine formation rate and still present 1 month later, alongside reduced stress-related behavioural deficit and elevated excitatory neurotransmission. (Shao et al. 2021, Neuron, doi:10.1016/j.neuron.2021.06.008)
  So: It is the strongest evidence that a structural trace persists after a single dose, which is why a plasma neuroplasticity marker is worth collecting, but it is in mice and at one month, so it cannot stand in for the human 6-month question.
• Resting-state functional magnetic resonance imaging before and after psilocybin in treatment-resistant depression identified cerebral blood flow and resting-state connectivity changes associated with antidepressant response. (Carhart-Harris et al. 2017, Scientific Reports, doi:10.1038/s41598-017-13282-7)
  So: It is the first clinical imaging mechanism study and the template this design follows, and its short follow-up is the limitation this design is built to pass.
• In the largest single-site psychedelic neuroimaging dataset to date, 62 adults scanned with functional magnetic resonance imaging and electroencephalography before and on the day of psilocybin showed reorganisation into structured context-sensitive patterns whose strength scaled with the depth of self-dissolving experience and with next-day mindset change. (Stoliker et al. 2026, Nature, doi:10.1038/s41586-026-10910-z)
  So: It shows the acute brain-to-experience link is now well characterised and ties it to subjective measures of the kind the 2016 trial's mediation analysis used, which sharpens the remaining gap to specifically the long timescale rather than the mechanism in general.

Searched: Searched OpenAlex for psilocybin neuroplasticity, brain network and connectivity work from November 2016, and Europe PMC for psilocybin imaging in cancer populations and for any imaging follow-up at 6 months; the acute and few-week mechanism literature is now large, but I found no psilocybin neuroimaging study in a cancer population and none that imaged at the 6-month clinical endpoint, so the specific gap the authors named is still open.

Feasibility

An academic centre running a psilocybin cancer trial with on-site research magnetic resonance imaging and precision functional mapping expertise, which in practice means a partnership between a psycho-oncology group and an imaging group · 4 to 5 years, set by the parent trial's enrolment and the 6-month endpoint rather than by the imaging · high, because nine or more scan sessions per participant plus the parent trial's therapist hours is the most expensive combination in this set
Blockers: Nine scan sessions over 6 months is a heavy burden for patients with advanced cancer, and non-completion will not be random with respect to disease progression; Head motion, fatigue, pain and analgesic medication degrade connectivity estimates in exactly the participants most likely to drop out; Attrition from mortality over 6 months may leave too few usable 6-month scans regardless of the enrolled number; The effect size for the persisting connectivity change is unpublished, so the sub-study cannot be formally powered at the point of funding

Strongest objection

An imaging methodologist would say the primary measure is a hostage to a single unreplicated finding. The anterior hippocampus to default mode network result comes from one small intensively-sampled healthy cohort with a few-week follow-up, and pre-registering it as the primary endpoint at 6 months in sick, moving, medicated patients is a bet that it replicates, persists far longer than shown, and survives a far worse signal-to-noise environment. If it does not, the study returns an uninterpretable null at great cost, and the methodologist would argue for a multivariate connectivity measure or a replication of the acute finding in the target population first.

Not derivable from what is published

• The magnitude and variance of the persisting anterior hippocampus to default mode network connectivity reduction, which Siegel et al. report qualitatively as lasting weeks without an effect size in the abstract
• Within-participant functional connectivity variance in cancer patients under treatment, which no published study supplies
• Whether the persisting connectivity change lasts anywhere near 6 months, since the longest published follow-up is 3 weeks
• Expected usable-scan attrition over 6 months in this population
• Whether plasma brain-derived neurotrophic factor changes measurably after psilocybin in humans, which the record does not address and which I did not confirm in the literature