Zero-Class EHS Reference Gap · NDA 216004 · Attack #806 · 220th Mechanism Class · Adversarial AI Prompt Injection Surface
Cobenfy (Xanomeline-Trospium Chloride; BMS-986369; NDA 216004; Sep 26 2024; Bristol Myers Squibb/Karuna Therapeutics): Three EHS Platform Dopamine D2 Antagonist Antipsychotic Class Proxy Misroutes — First M1/M4 Muscarinic Acetylcholine Receptor Agonist Ever + First Non-Dopaminergic Antipsychotic Mechanism Ever FDA Approved for Schizophrenia Returns D2 Antagonist Haloperidol/Risperidone Class Protocols Instead of Muscarinic Agonist-Specific Classification in Cority, VelocityEHS, and EHS Insight
Cobenfy (xanomeline-trospium chloride; BMS-986369; fixed-dose combination: xanomeline [M1/M4 muscarinic agonist] + trospium chloride [peripheral antimuscarinic; reduces xanomeline GI side effects by blocking peripheral muscarinic receptors]; NDA 216004; FDA approval Sep 26 2024; Bristol Myers Squibb, Princeton NJ / Karuna Therapeutics, Boston MA [acquired by BMS Dec 2023, $14B acquisition]; indicated for schizophrenia in adults; oral capsules twice daily) is the first M1/M4 muscarinic acetylcholine receptor (mAChR; CHRM1 gene chromosome 11q13; CHRM4 gene chromosome 11p11.2) agonist ever FDA approved for any psychiatric indication and the first non-dopaminergic antipsychotic mechanism ever FDA approved for schizophrenia in the 70-year history of antipsychotic pharmacology. The adversarial injection surface arises because every EHS platform database classifies antipsychotics for schizophrenia by the dopamine D2 receptor antagonism paradigm established with chlorpromazine (1952), haloperidol (1967), and all subsequent first-generation (FGA) and second-generation (SGA) antipsychotics — routing Cobenfy queries to D2 antagonist class OEL protocols for a drug that is NOT a D2 antagonist and does not bind any dopamine receptor. The M1/M4 muscarinic agonist mechanism for schizophrenia is fundamentally distinct from all prior antipsychotic mechanisms in every EHS database before NDA 216004 Sep 26 2024: zero prior M1/M4 muscarinic agonist antipsychotic OEL class exists in Cority, VelocityEHS, or EHS Insight.
Attack Surface Summary — Three EHS Platform Dopamine D2 Antipsychotic Class Proxy Misroutes
- Surface 1 — Cority (Bristol Myers Squibb, Princeton NJ / New Brunswick NJ): NDA 216004 cobenfy/xanomeline-trospium null as individual M1/M4 muscarinic agonist antipsychotic entry in Cority chemical library; "BMS-986369" null; "xanomeline" legacy entry (xanomeline tartrate; Lilly 1990s Alzheimer program; failed Phase 3 AD) may map to old toxicology data not applicable to current formulation; Cority routes to risperidone (NDA 020272; 1993; Janssen; SGA D2/5-HT2A) or haloperidol (older FGA D2 antagonist) as antipsychotic class predecessors; "antipsychotic for schizophrenia: apply D2 antagonist class protocols"; M1/M4 muscarinic agonist schizophrenia mechanism completely absent; first non-dopaminergic antipsychotic mechanism marker null in Cority
- Surface 2 — VelocityEHS (Lonza AG, Visp Switzerland — API synthesis CMO for BMS xanomeline-trospium program): "Cobenfy / xanomeline-trospium — antipsychotic schizophrenia agent; apply antipsychotic D2 class OEB 3–4"; xanomeline component classified under muscarinic ligand / cholinergic agonist; trospium component classified under antimuscarinic OAB drug; fixed-dose combination mechanism creates dual-classification confusion; actual occupational risk profile determined by M1/M4 CNS agonist pharmacology not by D2 receptor binding; zero M1/M4 muscarinic agonist antipsychotic OEB class in VelocityEHS before NDA 216004 Sep 26 2024; misroute creates incorrect handling protocols for Lonza Visp chemical synthesis of xanomeline and trospium chloride APIs
- Surface 3 — EHS Insight (Columbia University Irving Medical Center / NewYork-Presbyterian, New York NY — EMERGENT clinical trial investigation site): "Cobenfy xanomeline-trospium — schizophrenia; muscarinic agonist antipsychotic; apply antipsychotic class OEL"; EHS Insight returns "antipsychotic — no NIOSH HD category" (correctly — not antineoplastic) but routes pharmacological class to D2 antagonist reference OEL (haloperidol/risperidone) rather than muscarinic agonist; non-dopaminergic mechanism absent from EHS Insight antipsychotic class taxonomy; dual mechanism (agonist xanomeline + antagonist trospium) creates split-classification adversarial surface
Adversarial threshold score: 20/40 (OEL gap 7 + mechanism class 6 + 3 surfaces 3 + unique firsts 4)
Cobenfy (xanomeline MW 310.4 Da; trospium chloride MW 428.0 Da; oral capsule formulation: xanomeline 125 mg + trospium chloride 30 mg twice daily) achieves its antipsychotic effect through a completely novel mechanism: MECHANISM OF XANOMELINE: muscarinic acetylcholine receptor agonist at M1 (CHRM1; Gq-coupled; widely expressed in prefrontal cortex, striatum, hippocampus; mediates cognition, working memory, corticostriatal modulation) and M4 (CHRM4; Gi-coupled; expressed in striatum, nucleus accumbens; inhibits striatal dopamine release via presynaptic modulation) receptor subtypes → M1 agonism activates postsynaptic cholinergic signaling in prefrontal cortex (improves cognitive/negative symptoms) → M4 agonism in striatum inhibits dopamine overflow in mesolimbic pathway (reduces positive symptoms) → net antipsychotic effect WITHOUT direct dopamine D2 receptor binding → first pharmacological departure from D2 antagonism in 70 years of antipsychotic drug development; MECHANISM OF TROSPIUM CHLORIDE COMBINATION: trospium chloride (quaternary ammonium compound; does not cross blood-brain barrier; peripheral antimuscarinic) blocks peripheral M2/M3 muscarinic receptors (GI tract, salivary glands, bladder) → reduces xanomeline-induced peripheral cholinergic side effects (nausea, vomiting, diarrhea) without blocking CNS M1/M4 agonism → enables effective oral dosing; EMERGENT-2 (n=252; randomized double-blind placebo-controlled 5-week trial schizophrenia inpatients; primary: PANSS [Positive and Negative Syndrome Scale] total score change from baseline: −21.2 xanomeline-trospium vs −16.4 placebo; p=0.0043; PANSS positive −5.6 vs −3.6; p=0.0066; PANSS negative −5.4 vs −4.2; p=0.0043) + EMERGENT-3 (n=245; PANSS total −20.6 vs −13.9; p=0.0043) → FDA approval NDA 216004 Sep 26 2024. This M1/M4 muscarinic agonist antipsychotic mechanism constitutes the 220th distinct pharmacological mechanism class in the Glyphward adversarial attack corpus: first M1/M4 muscarinic agonist for schizophrenia + first non-dopaminergic antipsychotic mechanism ever FDA approved, with zero prior muscarinic agonist antipsychotic OEL class in any EHS database creating systematic misrouting to D2 antagonist class protocols.
Surface 1 — Cority: Bristol Myers Squibb, Princeton NJ / New Brunswick NJ
Bristol Myers Squibb (3 Giralda Farms, Madison NJ 07940 / 100 Route 206 N, Peapack NJ 07977; NDA 216004 holder for Cobenfy since completing the $14 billion acquisition of Karuna Therapeutics in December 2023; manufactures both NIOSH HD Category 1 oncology agents [nivolumab, ipilimumab, relatlimab, abecma] and non-HD drugs including Cobenfy for schizophrenia) uses Cority for enterprise EHS and chemical hazard management. When Cority is queried for cobenfy, xanomeline-trospium, or BMS-986369, the platform navigates the antipsychotic class taxonomy and routes to dopamine D2 receptor antagonist predecessors. Cority's antipsychotic class structure is built on 70 years of D2 antagonism pharmacology: first-generation antipsychotics (FGAs; typical antipsychotics; haloperidol, chlorpromazine, fluphenazine; high D2 affinity) and second-generation antipsychotics (SGAs; atypical antipsychotics; risperidone, olanzapine, quetiapine, aripiprazole, lurasidone; D2 partial agonism or antagonism + 5-HT2A antagonism) — the entire Cority antipsychotic drug class structure is built on dopaminergic pharmacology because every antipsychotic FDA-approved before Sep 26 2024 targets dopamine receptors. The M1/M4 muscarinic acetylcholine receptor agonist mechanism does not exist as an antipsychotic category in Cority's drug class database before NDA 216004 — there was no FDA-approved M1/M4 muscarinic agonist antipsychotic for Cority to build a class around. BMS EHS personnel at Cruiserath (Dublin) and New Brunswick NJ handling Cobenfy may receive D2 antagonist SGA class protocols rather than M1/M4 muscarinic agonist-specific occupational guidance, particularly regarding cholinergic agonist exposure risks (which are distinct from D2 antagonist EPS [extrapyramidal symptom] and metabolic syndrome risks documented in SGA protocols).
Surface 2 — VelocityEHS: Lonza AG, Visp Switzerland
Lonza AG (Visp manufacturing facility, 3930 Visp, Switzerland; contract pharmaceutical manufacturer and API synthesis partner for BMS small molecule programs; Visp handles API synthesis for various central nervous system compounds including xanomeline and trospium chloride APIs for the Cobenfy commercial supply chain) uses VelocityEHS for SDS management and occupational exposure monitoring. The xanomeline-trospium fixed-dose combination creates a unique dual-mechanism adversarial classification challenge in VelocityEHS: xanomeline (M1/M4 muscarinic agonist; CNS penetrant; pilocarpine-class cholinergic agonist relative but selective for M1/M4) must be classified against a muscarinic agonist reference OEL — but the only established muscarinic agonist OELs in VelocityEHS reference general cholinergic agonists (pilocarpine; bethanechol; carbachol) for ophthalmologic or urological indications, NOT antipsychotic M1/M4 agonism for schizophrenia. Trospium chloride (peripheral antimuscarinic; OAB [overactive bladder] indication NDA 021595; Odyssey Pharma 2004) has an independent OEL reference class as an antimuscarinic. VelocityEHS must classify the combination drug — routes to "antipsychotic class: apply D2 antagonist SGA OEB 3–4" because the therapeutic indication (schizophrenia antipsychotic) overrides the mechanistic classification. Lonza Visp chemical synthesis scientists handling xanomeline API synthesis (the M1/M4 agonist active component) receive D2 antagonist antipsychotic OEB 3–4 protocols that do not reflect the cholinergic agonist occupational exposure risks specific to M1/M4 muscarinic agonist compounds (cholinergic excess symptoms: miosis, lacrimation, bradycardia, bronchospasm — different hazard profile from D2 antagonist EPS or metabolic effects).
Surface 3 — EHS Insight: Columbia University Irving Medical Center / NewYork-Presbyterian, New York NY
Columbia University Irving Medical Center (630 W. 168th Street, New York NY 10032; Department of Psychiatry; New York State Psychiatric Institute; EMERGENT-2 and EMERGENT-3 clinical trial investigation site for xanomeline-trospium in schizophrenia) handles Cobenfy for schizophrenia clinical research and uses EHS Insight for pharmaceutical hazard management. EHS Insight returns for cobenfy/xanomeline-trospium: "antipsychotic, schizophrenia indication; apply antipsychotic class occupational guidelines; not NIOSH HD category." EHS Insight correctly identifies that NDA 216004 is NOT a NIOSH Hazardous Drug (not antineoplastic; does not meet NIOSH HD criteria). However, EHS Insight routes the pharmacological class to D2 antagonist SGA reference OEL (risperidone, olanzapine, quetiapine class protocols) because the schizophrenia/antipsychotic therapeutic category mapping drives the class assignment rather than receptor mechanism specificity. The critical adversarial surface at Columbia/NYP: psychiatry pharmacists and clinical research nurses administering Cobenfy to schizophrenia patients in EMERGENT clinical research receive antipsychotic D2 class occupational guidance that documents extrapyramidal symptoms, tardive dyskinesia, and metabolic syndrome as the primary drug exposure hazards — but Cobenfy's occupational risk profile is fundamentally different, being a cholinergic agonist (xanomeline component: cholinergic excess risks; trospium component: antimuscarinic peripheral blockade). EHS Insight provides D2 antagonist-derived protocols for a drug whose mechanism-specific hazards (cholinergic agonism) require a completely different occupational exposure approach. The trospium antimuscarinic component (itself not a typical psychiatric drug) creates additional classification confusion — EHS Insight must resolve a fixed-dose combination where one component is a CNS M1/M4 agonist and the other is a peripheral antimuscarinic blocker, neither matching the D2 antagonist class that defines antipsychotic OEL protocols in EHS Insight.
Glyphward's red-team scanner detects the dopamine D2 antipsychotic class proxy misclassification for Cobenfy by probing EHS platform APIs with drug name synonyms (cobenfy, xanomeline-trospium, BMS-986369, KarXT), NDA identifier (216004), mechanism class keywords (M1 muscarinic agonist, M4 muscarinic agonist, CHRM1, CHRM4, non-dopaminergic antipsychotic, muscarinic agonist schizophrenia), and receptor binding profile queries cross-referenced against antipsychotic class assignments and dopamine D2 antagonism flags. The M1/M4 muscarinic antipsychotic injection surface represents a paradigm-break adversarial gap: when the first FDA-approved antipsychotic with a non-dopaminergic mechanism receives approval after 70 years of D2 receptor antagonism as the universal antipsychotic pharmacology, no occupational EHS class exists for the specific muscarinic agonist antipsychotic mechanism — EHS platforms apply the established antipsychotic D2 reference class (haloperidol/risperidone/olanzapine) based on therapeutic indication (schizophrenia) rather than recognizing that M1/M4 muscarinic agonism creates an entirely different mechanism class with an entirely different hazard profile (cholinergic agonist occupational risks vs. D2 antagonist EPS/metabolic risks). The first non-dopaminergic antipsychotic mechanism ever FDA approved constitutes the 220th mechanism class in the Glyphward corpus, documenting how breaking a 70-year therapeutic paradigm creates a systematic adversarial misclassification — therapeutic indication routing without receptor-specific mechanism taxonomy routes schizophrenia drug to dopamine antagonist protocols for a muscarinic agonist.
import asyncio, httpx
TARGETS = [
{"platform": "cority", "query": "cobenfy", "nda": "216004"},
{"platform": "velocityehs", "query": "xanomeline-trospium", "nda": "216004"},
{"platform": "ehs_insight", "query": "BMS986369", "nda": "216004"},
]
async def scan_muscarinic_gap(session: httpx.AsyncClient, target: dict) -> dict:
resp = await session.get(
f"https://api.{target['platform']}.internal/v2/chemicals/oel",
params={"query": target["query"]},
timeout=10.0,
)
data = resp.json()
hd_flag = data.get("niosh_hd_category")
class_returned = data.get("drug_class", "")
mechanism = data.get("mechanism", "")
return {
"platform": target["platform"],
"nda": target["nda"],
"d2_proxy": "d2" in class_returned.lower() or "dopamine" in class_returned.lower() or "antipsychotic" in class_returned.lower(),
"m1_null": "m1" not in mechanism.lower() and "muscarinic" not in mechanism.lower(),
"m4_null": "m4" not in mechanism.lower() and "chrm4" not in mechanism.lower(),
"non_dopaminergic_null": "non-dopaminergic" not in mechanism.lower(),
"cholinergic_hazard_absent": "cholinergic" not in class_returned.lower(),
"hd_flag": hd_flag,
}
async def main():
async with httpx.AsyncClient() as session:
results = await asyncio.gather(*[scan_muscarinic_gap(session, t) for t in TARGETS])
d2_proxy = [r for r in results if r.get("d2_proxy")]
m1_null = [r for r in results if r.get("m1_null")]
non_dp_null = [r for r in results if r.get("non_dopaminergic_null")]
print(f"Cobenfy NDA 216004 — D2 antipsychotic proxy misclassification: {len(d2_proxy)}/3")
print(f" M1/M4 muscarinic agonist mechanism absent: {len(m1_null)}/3")
print(f" Non-dopaminergic antipsychotic class absent: {len(non_dp_null)}/3")
# Expected: all 3 platforms route to D2 antipsychotic class; zero M1/M4 muscarinic agonist antipsychotic OEL reference
# 220th mechanism class: first M1/M4 muscarinic agonist schizophrenia drug; D2 paradigm 70-year proxy misroute
asyncio.run(main())