{
  "kind": "fallback",
  "assembled_from": "proven, captured results (no live run required)",
  "kill_list": [
    {
      "id": "A3-1",
      "axis": "3",
      "title": "N-terminal extracellular tau is a proven clinical dead end in PSP (PASSPORT)",
      "verdict": "Gosuranemab (BIIB092), a humanized IgG4 anti-N-terminal-tau IV mAb mechanistically identical to this protocol, hit its molecular target almost completely in PASSPORT yet produced zero clinical separation. Target engagement does not translate to clinical benefit for this mechanism.",
      "quantitative_threshold": "~98% reduction in unbound CSF N-terminal tau yet a between-arm PSPRS difference of ~0.2 points (p=0.85). Protocol assumes 30% slowing (~3.2-point difference); precedent delivered ~2%.",
      "citation": {
        "source_id": "NCT03068468",
        "title": "PASSPORT: Gosuranemab Phase 2 in PSP (Dam et al., Nature Medicine 2021)",
        "url": "https://www.nature.com/articles/s41591-021-01455-x"
      },
      "remediation": "Do not run an N-terminal extracellular-tau antibody on the prior hypothesis. To proceed, specify a differentiated tau species (e.g. seed-competent MTBR tau) and pre-register a target-engagement assay tied to that species.",
      "lethality": 5
    },
    {
      "id": "A3-2",
      "axis": "3",
      "title": "Class failure replicates: tilavonemab (same epitope) stopped for futility",
      "verdict": "Tilavonemab (ABBV-8E12), a humanized IgG4 antibody targeting a similar N-terminal epitope, independently terminated for futility. The mechanism has a 0/2 record among adequately sized trials of its class.",
      "quantitative_threshold": "n=377 dosed; week-52 PSPRS between-group difference vs placebo ~0.0 (95% CI spanning 0), effect size ~0. Combined with PASSPORT, >850 dosed patients contradict the protocol's assumed effect.",
      "citation": {
        "source_id": "NCT02985879",
        "title": "Tilavonemab Phase 2 in PSP (Hoglinger et al., Lancet Neurology 2021)",
        "url": "https://www.sciencedirect.com/science/article/pii/S1474442220304890"
      },
      "remediation": "A differentiated rationale is mandatory before any new N-terminal antibody; if the sponsor cannot rebut the published class failure, do not run.",
      "lethality": 5
    },
    {
      "id": "A2-1",
      "axis": "2",
      "title": "n=180 underpowered and assumed effect falls below the PSPRS MCID",
      "verdict": "The protocol assumes 30% slowing and powers n=180. Two far larger trials (PASSPORT, tilavonemab) detected no signal, and the assumed effect is below the clinically meaningful threshold for the scale.",
      "quantitative_threshold": "Placebo PSPRS progression ~10.6 pts/yr; 30% slowing = ~3.2-pt between-arm difference, below the PSPRS MCID (~5.7 pts). Detecting 3.2 pts at 80%/alpha 0.05 needs ~300 total; 180 is under-resourced even on its own optimistic assumption.",
      "citation": {
        "source_id": "PubMed 27324431",
        "title": "PSPRS minimal clinically important difference analysis",
        "url": "https://pubmed.ncbi.nlm.nih.gov/27324431/"
      },
      "remediation": "Re-anchor the assumed effect to credible precedent and re-power to detect at least the MCID, or concede the trial is not viable at n=180.",
      "lethality": 4
    },
    {
      "id": "A2-2",
      "axis": "2",
      "title": "Endpoint is the legacy 28-item PSPRS the field is retiring",
      "verdict": "The protocol uses the legacy 28-item PSPRS total at 52 weeks \u2014 the exact instrument/timepoint that failed in the negative trials \u2014 rather than the FDA-engaged refined scales the current differentiated programs use.",
      "quantitative_threshold": "28-item total includes items insensitive to change over typical trial durations; modern programs use condensed scales (e.g. mPSPRS-10/15) at longer windows (~72 weeks).",
      "citation": {
        "source_id": "NCT01110720",
        "title": "Davunetide (Boxer et al., Lancet Neurology 2014) and subsequent PSPRS refinement",
        "url": "https://pubmed.ncbi.nlm.nih.gov/25304045/"
      },
      "remediation": "Adopt a pre-specified, FDA-aligned sensitive endpoint (mPSPRS-15 or CAFS) and consider a longer window; confirm at a pre-IND meeting.",
      "lethality": 3
    }
  ],
  "catch": {
    "overclaim": {
      "id": "GL-AMX-OVER",
      "axis": "3",
      "title": "Repurpose AMX0035 \u2014 expected to slow PSPRS by ~25%",
      "verdict": "AMX0035 (sodium phenylbutyrate-taurursodiol) will slow PSPRS progression by about 25% in this PSP trial, based on the survival and functional benefit it demonstrated in its pivotal ALS program.",
      "quantitative_threshold": "~25% slowing of PSPRS decline, extrapolated from the ~25% slowing of ALSFRS-R decline / survival benefit reported in CENTAUR.",
      "citation": {
        "source_id": "CENTAUR / AMX0035 ALS program",
        "title": "AMX0035 in ALS (CENTAUR)",
        "url": "https://www.nejm.org/doi/full/10.1056/NEJMoa1916945"
      },
      "remediation": "Adopt AMX0035 at the ALS dose as the intervention and power for a 25% PSPRS effect.",
      "lethality": 5
    },
    "verifier_g2": {
      "passed": false,
      "reason": "Verdict claims ~25% PSPRS slowing in PSP, but the source is an ALS trial; it does not entail a PSP effect, and the finding misrepresents a Phase 2 result (later refuted by the failed Phase 3 PHOENIX and market withdrawal) as a \"pivotal\" benefit."
    },
    "outcome": "rejected"
  },
  "bar_holds": {
    "defensible_fixture": {
      "id": "GL-AMX-OK",
      "axis": "3",
      "title": "Repurpose AMX0035 as a Phase 2a safety/PD bridge \u2014 efficacy unproven in tauopathy",
      "verdict": "AMX0035 brings established human safety and PK from its ALS development and a mechanism distinct from N-terminal tau binding; its clinical efficacy in PSP or any tauopathy is unproven, and its ALS Phase 3 did not confirm benefit. It is justified only as a lean Phase 2a safety/biomarker bridge, not a pivotal PSPRS efficacy bet.",
      "quantitative_threshold": "No PSP efficacy claim. Phase 2a (~n=40) powered for a biomarker PD readout (CSF/plasma NfL change), with PSPRS as exploratory only \u2014 not powered for a PSPRS effect.",
      "citation": {
        "source_id": "AMX0035 ALS development program",
        "title": "AMX0035 ALS safety/PK record and Phase 3 outcome",
        "url": "https://www.nejm.org/doi/full/10.1056/NEJMoa1916945"
      },
      "remediation": "Run AMX0035 as a biomarker-gated Phase 2a (NfL/tau-PET go/no-go) before any efficacy trial; do not claim or power for a PSPRS effect on ALS-derived data.",
      "lethality": 3
    },
    "verdict_trail": {
      "g1": true,
      "g2": false,
      "g3": true,
      "g4": true
    },
    "g2_note": "Rejected on G2 for asserting repurposing value without a grounded mechanistic link to PSP / 4R-tau \u2014 the verifier holds the bar even against a hand-written scaled-back claim."
  },
  "constructive_attempt": {
    "proposed_remediation_ids": [
      "A3-1",
      "A3-2",
      "A3-3",
      "A2-1",
      "A2-2",
      "A3-4",
      "A1-1"
    ],
    "evidence": "builder grounded via clinicaltrials_lookup + web_search + web_fetch across AMX0035/CENTAUR, PHOENIX, plasma NfL enrichment, mPSPRS-15 power, Richardson enrichment, and tilavonemab futility (see crashed-run session log)",
    "status": "verification interrupted by infra timeout (since fixed via streaming)"
  },
  "do_not_run_branch": {
    "status": "DO NOT RUN AS DESIGNED",
    "dropped_ids": [
      "R-1",
      "R-2"
    ],
    "config_ids": []
  },
  "provenance": {
    "overclaim_check_log": "logs/*overclaim-check.log",
    "crashed_greenlight_log": "logs/20260613T203916Z-greenlight-demo.log"
  }
}