{ "slug": "psilocybin-therapy", "case_path": "consensus_consistency_audit_2026-10-09\\revised\\psilocybin-therapy", "round": 2, "group": 1, "review_date": "2026-10-14", "adversarial_reviewer": "Claude Opus 5.5 (Anthropic)", "authoring_model_of_record": "Sol 6.1", "authoring_model_as_declared_in_document": "Codex/GPT agent original assessment 2026-09-27; Claude (Anthropic) cross-model red-team synthesis 2026-09-28", "status": "Approved as revised", "level_before": "Γ", "level_after": "Γ", "descriptor_after": "Mechanistically plausible, partial system support", "confidence_before": "Moderate", "confidence_after": "Moderate", "role_before": "epiphenomenon", "role_after": "epiphenomenon", "subclaims_before": { "SC1": "Ε", "SC2": "Δ", "SC3": "Ε", "SC4": "Γ", "SC5": "Γ" }, "subclaims_after": { "SC1": "Ε", "SC2": "Δ", "SC3": "Ε", "SC4": "Γ", "SC5": "Γ" }, "msd_before": { "momentum": "Decisive work underway", "stability": "Uncertain", "divergence": "Contested" }, "msd_after": { "momentum": "Decisive work underway", "stability": "Uncertain", "divergence": "Contested" }, "sources_total_before": 50, "sources_total_after": 50, "pmids_checked": 54, "pmids_reference_list": 50, "pmids_failed": 0, "pmids_erratum_records_checked": 4, "retractions_or_eocs_found": 0, "claims_spot_checked": 12, "claims_spot_check_failures": 0, "validator": { "tool": "framework/5.4/validate_review.py", "before": "PASS - 0 fail, 0 warn, 33 pass", "after": "PASS - 0 fail, 0 warn, 33 pass", "output_file": "validator.txt" }, "findings_by_severity": { "critical": 0, "major": 2, "minor": 1, "editorial": 2 }, "findings": [ { "id": "A-1", "severity": "Minor", "category": "rendering defect", "location": "Primary evidence and access ledger, row [32], system/design cell", "original": "adult local corticalHtr2a deletion", "corrected": "adult local cortical Htr2a deletion", "evidence": "Missing space between words; the round-2 brief lists this class of reference/ledger builder rendering defect explicitly. No change of meaning.", "artifacts_changed": [ "psilocybin-therapy.review.md" ] }, { "id": "A-2", "severity": "Editorial", "category": "rendering defect", "location": "Primary evidence and access ledger, row [34], system/design cell", "original": "differentiated>40 d;2–4 donor lines", "corrected": "differentiated >40 d; 2–4 donor lines", "evidence": "Two missing spaces, around a numeral and after a semicolon: the same builder defect class.", "artifacts_changed": [ "psilocybin-therapy.review.md" ] }, { "id": "A-3", "severity": "Editorial", "category": "rendering defect / internal inconsistency", "location": "Primary evidence and access ledger, row [34], system/design cell", "original": "co-culture. 10µM psilocin", "corrected": "co-culture. 10 µM psilocin", "evidence": "Missing space between numeral and unit. The same concentration is rendered '10 µM' in section 3 and in this row's own limit cell.", "artifacts_changed": [ "psilocybin-therapy.review.md" ] }, { "id": "A-4", "severity": "Major", "category": "false provenance claim about adversarial review", "location": "Consensus provenance reconciliation block, and consensus-addendum.md", "original": "The earlier adversarial approval refers to its original edition; this provenance supplement is not a new independent or blinded adjudication.", "corrected": "No earlier adversarial approval is attached to this case; this provenance supplement is not a new independent or blinded adjudication.", "evidence": "integrate_consensus.py line 71 appends this sentence for every case with no approval check. Exactly 26 reviews carry the 2026-10-06 approval line; they were enumerated and psilocybin-therapy is not among them. The risk is specific: the confusable sibling slug psilocybin-dual-phase WAS approved on 2026-10-06, so a reader could reasonably believe the approval attaches here.", "lead_flagged": true, "artifacts_changed": [ "psilocybin-therapy.review.md", "consensus-addendum.md" ] }, { "id": "A-5", "severity": "Major", "category": "confidence band justified by reviewer access depth", "location": "Steelman/final-adjudication confidence sentence, and scorecard final_score.confidence_rationale", "original": "... many primary full texts read ... several records remain access-limited ...", "corrected": "Band now rests on independent replication across clinical and experimental lines, heterogeneous human measurement, absent mediator identification and the unpublished phase 3 datasets, with an explicit statement that access depth is reported separately and is not a basis for the band.", "evidence": "The framework states confidence reflects evidence density and reproducibility, and that access depth is a statement about the reviewer, not the source.", "band_changed": false, "band_argued_both_ways": "Moderate stands: human measurement is heterogeneous, the one preregistered synaptic-marker test was null and the mediator has never been manipulated, so the High text's extensive-replication requirement is not met. The three non-access grounds already given are sufficient alone.", "lead_flagged": true, "artifacts_changed": [ "psilocybin-therapy.review.md", "psilocybin-therapy.scorecard.json" ] } ], "rejected_findings": [ { "id": "R-1", "checked": "Six references print a year one off the PubMed publication-date year (Grob 2011/2010; Carhart-Harris 2018/2017; von Rotz 2023/2022; Becker 2022/2021; Jiang 2026/2025; Kramer 2026/2025)", "disposition": "Rejected - print/issue year versus e-pub year differing by one is an accepted convention." }, { "id": "R-2", "checked": "PubMed titles for refs 26, 31, 32 render as '5-HTReceptor' / '5-HTreceptors', dropping the subscript", "disposition": "Rejected - PubMed title fields dropping subscripted or italicised receptor/gene/species names are not errors; the review's own rendering is correct." }, { "id": "R-3", "checked": "Ledger row [2] calls EPISODE a 'Randomized triple-arm first-period comparison' while the paper randomised four groups 2:2:1:1", "disposition": "Rejected - the row is explicitly scoped to the first dosing period, where the four arms collapse to three conditions (n=47/48/47); row [3] states that sustained-response totals are sums across the four randomized groups." }, { "id": "R-4", "checked": "Weakest-link derivation 'Ε, Δ, Ε, Γ, Γ -> Γ' and the reported alternatives (primary-subclaim Γ, scope-matched Γ)", "disposition": "Rejected - all three values are correct. Γ is the lowest letter in the set; the central claim's scope is the full human mediated chain (SC5), so scope-matched also returns Γ; SC5 is the named primary subclaim. The 'highest subclaim' misreading is not present." }, { "id": "R-5", "checked": "Momentum 'Decisive work underway'", "disposition": "Rejected - required artifact present and specific (NCT06512220 and NCT07490353 named with sponsor, n, primary completion and readouts), with the limits of both trials stated. Not momentum boilerplate." }, { "id": "R-6", "checked": "SC2 at Δ / Low resting on an exploratory randomized self-report separation to day 102", "disposition": "Rejected - a randomized contrast with a symptom endpoint at the registered twelve-week benchmark is a field-valid perturbation at that scope; the fragility (rescue-sensitive shortening to day 72, unconfirmed twelve-week sustained response, open-label extensions cannot cross Δ) is already recorded and is what sets Low." }, { "id": "R-7", "checked": "Reference-builder collective-author defect (leading comma, dropped consortium author)", "disposition": "Rejected - checked and not present; every reference has a named individual first author." }, { "id": "R-8", "checked": "claims_total equal to sources_total at 50", "disposition": "Rejected - one claim row per source with 50 verified and 0 unverified; a one-to-one ratio is not the implausible claim-to-source ratio the ledger rule targets, and the validator raises no warning." }, { "id": "R-9", "checked": "Lead-flagged Oct-9 revision artifact: a spliced Consensus section coexisting with preserved REVIEWER NOTES still asserting that no Consensus work was performed", "disposition": "Rejected - checked and not present in this case. The spliced block is headed \"Consensus provenance reconciliation - 9 October 2026\" and is self-limiting: it states that the located records are historical checks, \"not newly credited complete-paper examinations\", that no source-comparison count is inferred, and that the supplement \"is not a new independent or blinded adjudication\". The preserved reviewer notes and the scorecard literature_tools_note make no negative claim about Consensus; they positively enumerate the retrieval tools used by the authoring passes, which predate the 2026-10-05 enrichment the Consensus records come from. No sentence in either review contradicts the spliced block, so there was nothing to scope and nothing was edited." }, { "id": "R-10", "checked": "Renderer bug: 'Best-supported rival' printing models_evaluated[0], often the nominal", "disposition": "Rejected - not present. models_evaluated[0] is 'Context-dependent treatment response with unresolved mediators' (rank 1, is_nominal false); the nominal plasticity-window model is index 1, rank 2. The prose names the rank-1 non-nominal model." } ], "residual_limitations": [ "24 of 50 ledger rows are at abstract_metadata depth; figures from those rows were verified against publisher abstracts, not full Methods.", "Three figures could not be verified at abstract depth and rest on full text the review records having read: the NHS trial's week-6 MADRS difference and allocation-guessing counts (row 1), Shao 2025's day-65 imaging session (row 32), Johansen's interim-futility conditional power (row 37). All are consistent with the abstracts; not flagged.", "COMP005/COMP006 remain company and regulatory documents, correctly excluded from the graded base, so the grade cannot be cross-checked against them.", "This pass did not re-derive the earlier red-team dispositions in redteam/REDTEAM_REPORT.md." ], "scorecard_changed": true, "scorecard_change_reason": "One line only: final_score.confidence_rationale (finding A-5). No level, band, count or role moved.", "package_level_propagation": { "strings_to_replace": [ { "what": "psilocybin-therapy.review.md sha256 in this case folder's SHA256.txt (and any package MANIFEST/BATCH_COMPLETION entry carrying it)", "old": "84a62ccb8fe836732cb77e2682bb6dc530bc8c2bcd9e7630fe96af1b422f3dd3", "new": "5bba4fb59016d71f43c19de539704cc9de2020a4d5ed8f346abd5ef24fc064ff", "supersedes_earlier_entry_in_this_file": true } ], "note": "No headline string, letter, confidence, count or title changed, so no package-level file carries a now-stale value. The case folder's SHA256.txt needs the new review.md digest regenerated by the lead. NOTE: an R-9 row was added to the review's rejected-findings table after the first hash was recorded; the final digest is the one in changed_file_hashes." }, "changed_file_hashes": { "psilocybin-therapy.review.md": { "sha256": "7bc1fa2d4e71cf02d970b396ec41fca512116cb47899285536eba7b7f464aa09", "bytes": 148600, "previous_sha256": "84a62ccb8fe836732cb77e2682bb6dc530bc8c2bcd9e7630fe96af1b422f3dd3" }, "validator.txt": { "sha256": "80129f44df9ab0731e5b3e56f387b7fcde4c5de4133b0bec4560e5ebe81cfab2", "bytes": 2483, "new_file": true }, "ADVERSARIAL_REVIEW.md": { "sha256": "20630ff3a0f7b6dac22a2e359803237ffa722c40b46fd71e6811376c3dc27509", "bytes": 8182, "new_file": true }, "psilocybin-therapy.scorecard.json": { "sha256": "2a52d5f6f4189d513aa2260f9474bbdd10ea5c37390d7420e980ffe3b27ca760", "bytes": 342799, "previous_sha256": "aca9f49276c747c1833b621c9ebaf5f52c41eef43b3fd2882a5b7ab35d700df5", "change_detail": "One line only: final_score.confidence_rationale (finding A-5). No level, band, count or role moved." }, "consensus-addendum.md": { "sha256": "ab7235729de878862e74b6eff3a8d5df779f54374de0b036eb29bb476f784a18", "bytes": 1973, "previous_sha256": "a69c4f8b3f56a57066003a40d9327d9b3f5e53b47b916dfd70bd431a75c5ae37", "change_detail": "Rescoped the false earlier-adversarial-approval sentence (same correction as in the review)." } }, "propagate_to": [], "divergent_copies": [], "consensus_questions": [ "Has any completed human study measured drug-induced change in cortical inducibility (paired associative stimulation, TMS-EEG LTP-like plasticity, or visual/somatosensory potentiation) after psilocybin, with both onset and closure timed?", "In randomized psilocybin depression trials that reported a blinding-integrity check, does outcome track guessed allocation?" ], "lead_flagged_artifact_checks": { "false_approval_boilerplate": "present; corrected (A-4)", "confidence_from_access_depth": "present; rationale corrected, band unchanged (A-5)", "models_evaluated_0_renderer_bug": "absent (R-10)", "reviewer_notes_contradiction": "absent (R-9)", "false_hiv_draft_clause": "absent", "review_metadata_reviewer_left_unedited": true }, "line_endings_note": "CRLF preserved on every edited file, so each diff against sol61-original is confined to the corrected content.", "lead_policy_2026_10_14": { "confidence_two_step_test": { "step1_strike_access_depth": "applied; access-depth clause removed from the rationale (finding A-5)", "step2_rederive_band": "Moderate stands", "surviving_valid_determinants": [ "human plasticity measurement heterogeneous across PET, fMRI, EEG and cognitive proxies", "the proposed mediator has never been manipulated, so nothing measures the quantity the claim is about", "the largest randomized datasets (COMP005, COMP006) are company and regulatory documents, not peer-reviewed reports", "a null preregistered synaptic-marker test stopped for futility; a fixed-order healthy-volunteer design; near-ceiling allocation guessing wherever blinding was measured" ], "band_before": "Moderate", "band_after": "Moderate", "band_changed": false }, "momentum_stricter_rule": { "value": "Decisive work underway", "result": "holds", "basis": "Both artifacts are trials actually recruiting, not completed reports: SYNVEST (NCT06512220, recruiting since 2025-04-21, primary completion 2027-04) and NCT07490353 (recruiting since 2025-11-01, primary completion 2028-05); the review states neither has results. Both run the psilocybin-with-and-without-antagonist design that rubric 07 lists as threshold-crossing. No completed publication is counted.", "completed_work_counted_as_momentum": false }, "policy_as_stated_by_lead": "Confidence: strike access depth as a determinant everywhere, then re-derive; raise the band only if access depth was the only thing holding it down and the cited evidence meets the High text. Documented evidence-quality or reproducibility concerns are valid determinants and the band stands. Momentum: 'Decisive work underway' requires work actually being run that could cross the next threshold; a completed publication or trial never satisfies it." } }