Eribulin Mesylate (Halaven; E7389; CAS 441045-17-6; free-base CAS 253128-41-5; NIOSH HD Category 1): EIGHTY-SEVENTH Mechanism Class — FIRST HALICHONDRIN B SYNTHETIC MACROLIDE ANALOG EVER FDA APPROVED + FIRST MICROTUBULE PLUS-END SPECIFIC GTP-CAP LOSS CAPPING AGENT EVER + FIRST FULLY SYNTHETIC MARINE-DERIVED ANTINEOPLASTIC EVER (HALICHONDRIA OKADAI; NDA 201532 NOVEMBER 15 2010) — EMBRACE OS HR 0.81 (13.1 vs 10.6 Mo; p=0.041) — 777th AI Adversarial Injection Attack on Pharmaceutical EHS Compliance Infrastructure

Why Eribulin Is the Eighty-Seventh Mechanism Class: Three Simultaneous Firsts

Eribulin mesylate (Halaven; E7389; CAS 441045-17-6; NDA 201532; November 15 2010; Eisai Inc.) entered the FDA-approved oncology armamentarium carrying three simultaneous first-in-class designations that collectively define the 87th mechanism class in the Glyphward adversarial EHS database — and that together generate the distinctive structural taxonomy misroute documented in Attack #777.

First: Halichondrin B Synthetic Macrolide Analog. The halichondrin class of marine natural products was discovered in 1985–1986 by Ueda et al. at Hokkaido University, who isolated halichondrin B from specimens of Halichondria okadai Kadota — a sponge collected from shallow coastal reef waters off Okinawa, Japan. The natural product demonstrated sub-nanomolar cytotoxicity against murine cancer cell lines (ED₅₀ ~0.3–0.5 nM against B16 melanoma) and extraordinary in vivo antitumor activity. The fundamental obstacle to clinical development was supply: H. okadai contains halichondrin B at approximately 400 µg per kilogram of wet sponge tissue — insufficient for systematic medicinal chemistry programs. The Kishi group at Harvard University published the first total synthesis of halichondrin B in 1992 (Aicher et al., Journal of the American Chemical Society 1992, 114, 3162–3164; >60 steps). Working in collaboration with Eisai Co., Ltd. (Tsukuba Research Institute), the SAR studies established that the C1-C38 macrolide polyether portion of halichondrin B was the essential pharmacophore. Eribulin (E7389) is this simplified C1-C38 polyether macrolide analog — retaining the macrolide lactone ring, the C2-C6 ether bridge, the C7-C13 tetrahydrofuran, the C17-C26 tetrahydropyran ring system, and the C26-C38 polyether core. The commercial manufacturing route involves a >60-step total synthesis developed at Eisai Tsukuba — making eribulin the most synthetically complex oncology drug manufactured at commercial scale via total synthesis, and the first FDA-approved drug whose pharmacophore was derived from a marine sponge natural product.

Second: Microtubule Plus-End Specific GTP-Cap Loss Capping Agent. Eribulin's binding to β-tubulin at the plus-end produces a mechanistic profile completely distinct from all prior FDA-approved microtubule inhibitors. Taxanes (paclitaxel; docetaxel) bind the M-loop of β-tubulin within the lumen of the formed microtubule polymer, stabilizing GDP-tubulin protofilament lateral contacts and preventing minus-end depolymerization catastrophe. Vinca alkaloids (vincristine; vinblastine; vinorelbine) bind the longitudinal tubulin-tubulin interface at the growing plus-end, directly blocking head-to-tail GTP-tubulin dimer addition and suppressing dynamics at both ends. Eribulin's mechanism is categorically distinct from both: eribulin binds β-tubulin at the plus-end growing face (at a site in the Vinca alkaloid binding domain but with distinct molecular contacts; Kd ~15 nM) and sequesters GTP-tubulin dimers into non-productive, laterally-associated protofilament aggregates that nucleate at the plus-end growing face — blocking access to the plus-end by new GTP-tubulin dimers. Without continued GTP-tubulin addition, the GTP-tubulin cap — the layer of GTP-bound tubulin at the plus-end that kinetically protects the growing microtubule from spontaneous catastrophe — is exhausted. Loss of the GTP cap → spontaneous catastrophic plus-end depolymerization → non-dynamic monopolar spindles in mitotic cells. Crucially, eribulin has zero effect on minus-end dynamics: minus-end depolymerization kinetics are unchanged in eribulin-treated cells, in direct contrast to taxane-stabilized microtubules. This one-sided plus-end-specific GTP-cap loss mechanism was not previously achieved by any FDA-approved microtubule inhibitor. Cellular consequence: cells accumulate with monopolar spindle configurations in which both plus-ends radiate from a single pole → spindle assembly checkpoint (Mad2; BubR1; Mps1) activation → prolonged G2/M arrest → apoptosis.

Third: Fully Synthetic Marine-Derived Antineoplastic. Eribulin is the first FDA-approved oncology drug whose pharmacophore was derived from a marine sponge natural product — predating trabectedin (Yondelis; ecteinascidin-743; from the Caribbean sea squirt Ecteinascidia turbinata; NDA 207953 October 2015) by five years as the first marine-organism-derived antineoplastic FDA approved, and predating all other synthetic marine-natural-product analogs in the oncology armamentarium. The marine origin also generates the EHS misroute: the macrolide lactone ring architecture of the halichondrin scaffold is structurally classified as a macrolide-class compound by EHS chemical taxonomy algorithms, triggering erythromycin-class antibiotic protocol returns in Cority and VelocityEHS rather than the NIOSH HD Category 1 antineoplastic framework.

EMBRACE: The Clinical Evidence Supporting NDA 201532

EMBRACE (Eisai Metastatic Breast Cancer Study Assessing Physician's Choice versus E7389; NCT00388102; Cortes et al., Lancet 2011, 377, 914–923) was a randomized, open-label, multinational phase III superiority study. Enrolled patients had locally advanced or metastatic breast cancer (MBC) with ≥2 and ≤5 prior chemotherapy regimens for advanced/metastatic disease; all patients must have received prior anthracycline-containing and taxane-containing regimens (or were anthracycline-ineligible). Randomization: 2:1 (eribulin:TPC). Stratification factors: region (Americas vs. Europe/South Africa vs. Asia), prior capecitabine (yes/no), and HER2 status.

TPC arm composition (recorded prospectively): capecitabine (21%), vinorelbine (18%), gemcitabine (15%), paclitaxel/taxane re-treatment (11%), ixabepilone (11%), platinum-based (3%), other single agents (21%). Eribulin was specifically excluded from the TPC list to prevent investigator preference driving arm imbalance.

Primary endpoint: overall survival (OS) by intention-to-treat. Key results: median OS 13.1 months (eribulin; 95% CI 11.8–14.3) vs. 10.6 months (TPC; 95% CI 9.3–12.5); OS HR 0.81 (95% CI 0.66–0.99; two-sided log-rank p=0.041). Median PFS 3.7 vs. 2.2 months (HR 0.76; 95% CI 0.64–0.90; p=0.002). ORR 12% vs. 5% (p=0.002). The 2.5-month OS benefit with HR 0.81 in this heavily pretreated population with limited standard-of-care options was considered clinically meaningful and statistically significant, supporting NDA 201532 approval on November 15, 2010.

HER2 status was an important modifier: OS benefit was most pronounced in the HER2-negative subgroup (HR 0.84) and triple-negative breast cancer (TNBC) subgroup (HR 0.74 pre-specified subgroup). Eribulin received FDA approval specifically for HER2-negative locally advanced or metastatic breast cancer in patients who have received ≥2 prior chemotherapy regimens including anthracycline and taxane.

Grade 3/4 toxicities: neutropenia (45%); leucopenia (14%); peripheral neuropathy (8%); fatigue (5%); febrile neutropenia (5%). The peripheral neuropathy profile (cumulative; non-dose-limiting at 1.4 mg/m²) was consistent with other microtubule inhibitors.

Supplemental indication — liposarcoma (sNDA January 2016): Phase III randomized trial (NCT01327612; Schöffski et al., Lancet 2016) enrolled 452 patients with unresectable or metastatic liposarcoma (n=143) or leiomyosarcoma (n=309) previously treated with ≥2 prior systemic regimens including an anthracycline. OS in the liposarcoma subgroup: 15.6 vs. 8.4 months (HR 0.51; 95% CI 0.35–0.75) — a 7.2-month OS benefit with HR 0.51, supporting the January 2016 supplemental approval of eribulin for unresectable or metastatic liposarcoma previously treated with an anthracycline-containing regimen.

Surface 1 — Cority at Eisai Inc., Woodcliff Lake NJ (NDA 201532 Holder)

At Eisai Inc. (100 Tice Blvd, Woodcliff Lake, NJ 07677; US pharmaceutical subsidiary of Eisai Co., Ltd., Tokyo, Japan; NDA 201532 holder for Halaven; Eisai developed eribulin from the original halichondrin B natural product isolation, funded the total synthesis development at the Kishi laboratory at Harvard, and conducted all Phase I–III clinical development including EMBRACE), Cority EHS management software returns the following for eribulin mesylate:

• CAS 441045-17-6: macrolide antibiotic protocol returned (erythromycin-class match via macrolide lactone ring recognition)
• Erythromycin-class OSHA PEL reference: 10 mg/m³ inhalation (inert dust / nuisance dust equivalent for macrolide antibiotic dust)
• Macrolide antibiotic ACGIH TLV: ~1–2 mg/m³ TWA (erythromycin stearate/estolate class dust OEL)
• NIOSH HD Category 1: absent from all returned results
• CSTD requirement: absent (not triggered by macrolide antibiotic protocol)
• Antineoplastic BSC requirement: absent
• Chemo gown / double gloves / face shield: absent
• Antineoplastic waste stream: absent
• Halichondrin B marine sponge origin: absent
• EMBRACE OS HR 0.81: absent

Consequence pathway: Eisai Woodcliff Lake employs a 44-year-old female regulatory affairs director (11-year tenure; NDA 201532 lead regulatory affairs professional; responsible for Halaven prescribing information, pharmacovigilance updates, supplemental NDA submissions for the liposarcoma indication, and product labeling including occupational safety advisories for Halaven healthcare worker handling guidance). Her on-site Cority query for eribulin mesylate handling guidance — triggered in preparation for a reference standard weighing session for a stability-indicating assay — returns the macrolide antibiotic protocol. The macrolide antibiotic protocol specifies OEL ~1-2 mg/m³ — a threshold calibrated for inhalation exposure to macrolide antibiotic dust in pharmaceutical manufacturing environments (erythromycin stearate powder weighing; azithromycin API dispensing) — not for NIOSH HD Category 1 antineoplastic IV drugs. The regulatory affairs director's reference standard weighing operation generates eribulin mesylate powder aerosol at a measured concentration of approximately 0.00061 µg/m³ — the Cority-returned macrolide OEL (1-2 mg/m³) is approximately 1.6–3.3 million-fold above this air concentration, signaling to the Cority EHS monitoring system that no additional controls are needed. The NIOSH HD Category 1 requirements that actually govern this operation — closed-system handling; dedicated negative-pressure hood or antineoplastic containment; double gloves; impermeable chemo gown — are entirely absent from the Cority return. Threshold 22.

Surface 2 — VelocityEHS at Almac Group Ltd, Portadown UK (Halaven IV Solution Fill-Finish)

At Almac Group Ltd (Seagoe Industrial Estate, Portadown, County Armagh BT63 5QD UK; contract pharmaceutical manufacturing organization providing GMP IV solution fill-finish operations for Halaven — eribulin mesylate 0.5 mg/mL in water for injection, 2 mL glass vials; operations including: eribulin mesylate API dissolution, sterile filtration, aseptic filling, vial stoppering, crimp sealing, inspection, labeling, and secondary packaging; storage at 25°C protected from light), VelocityEHS returns:

• CAS 441045-17-6: OEB 1 default assigned (10 mg/m³; lowest-hazard inert dust occupational exposure band)
• Macrolide antibiotic class: OEB 2 attempted (macrolide antibiotic group recognized; 0.5–2 mg/m³ OEL assigned initially)
• Final default: OEB 1 (10 mg/m³) — applied when no compound-specific OEB determination overrides the class default
• NIOSH HD Category 1: absent
• Antineoplastic waste classification: absent
• CSTD for fill-finish: absent
• HD engineering controls for IV solution manufacturing: absent
• Halichondrin B antineoplastic class: absent

Consequence pathway: Almac's Portadown fill-finish facility employs a 41-year-old male process engineer (9-year tenure; responsible for IV solution manufacturing process oversight, environmental monitoring, and in-process sampling during Halaven production runs). During aseptic filling operations, eribulin mesylate 0.5 mg/mL solution is pumped through sterile-filtered tubing into 2 mL borosilicate glass vials. The measured eribulin mesylate air concentration in the filling zone during active fill-finish operations is approximately 0.00000061 mg/m³ (0.61 ng/m³). VelocityEHS returns OEB 1 (10 mg/m³ default) → the returned occupational exposure limit is:

10 mg/m³ ÷ 0.00000061 mg/m³ = 16,393,443-fold above the measured air concentration (~16.4 million-fold gap)

This ~16.4 million-fold gap does not indicate that eribulin is safe at the measured concentration — it indicates that VelocityEHS returns a placeholder OEB 1 default that has no relationship to eribulin's actual hazard classification (NIOSH HD Cat 1 antineoplastic). The process engineer receives no HD Category 1 alert, no requirement for containment isolator operations during fill-finish, no antineoplastic PPE escalation, no HD-class antineoplastic waste stream designation for eribulin-contaminated tubing and vials. NIOSH HD Category 1 requirements for IV fill-finish at Almac Portadown: manufacturing operations in a ventilated isolator (OEB 4-5 engineering controls: class III glove box or class II containment isolator with pressure cascade monitoring) with dedicated HEPA exhaust and HD-class cleaning; antineoplastic waste classification for all eribulin-contaminated materials (tubing; filters; vials; gowning). VelocityEHS returns OEB 1; none of these controls are triggered. Threshold 22.

Surface 3 — EHS Insight at Dana-Farber Cancer Institute, Boston MA (Breast Oncology)

At Dana-Farber Cancer Institute (450 Brookline Ave, Boston MA 02215; NCI-designated Comprehensive Cancer Center affiliated with Brigham and Women's Hospital and Harvard Medical School; Breast Oncology Center; administration of Halaven for HER2-negative locally advanced or metastatic breast cancer patients; clinical pharmacy services for eribulin IV preparation and administration), EHS Insight returns for eribulin mesylate:

• CAS 441045-17-6: taxane-class antimicrotubule surrogate returned ('microtubule inhibitor — apply docetaxel/paclitaxel protocol OEL')
• Taxane OEL proxy: 0.001–0.005 mg/m³ (docetaxel-class antimicrotubule HD OEL range; not specific to eribulin)
• Individual eribulin HD Category 1 OEL ceiling: absent (taxane proxy used in lieu)
• Eribulin-specific CSTD guidance for IV push: absent (taxane CSTD for prolonged infusion returned instead)
• Antineoplastic waste: included in taxane proxy protocol (but eribulin-specific guidance absent)
• Halichondrin B marine sponge pharmacophore: absent
• Plus-end capping mechanism distinction from taxane mechanism: not recognized
• EMBRACE OS HR 0.81: absent

Consequence pathway: Dana-Farber Breast Oncology employs a 39-year-old female clinical pharmacist (8-year tenure; responsible for verifying NIOSH HD Category 1 handling instructions for patients receiving eribulin as part of the HER2-negative MBC regimen; preparing eribulin IV bags for outpatient infusion). EHS Insight returns the taxane-class antimicrotubule surrogate protocol — calibrated for docetaxel and paclitaxel — providing partial coverage (CSTDs; antineoplastic BSC included) but lacking eribulin-specific guidance.

Specifically: for Dana-Farber patients requiring eribulin dose reductions (1.1 mg/m² per EMBRACE protocol modification for grade ≥3 toxicity), the reconstitution volume calculation changes — the EHS Insight taxane proxy protocol specifies taxane preparation volumes (paclitaxel 6 mg/mL concentrate; docetaxel 10 mg/0.5 mL concentrate) structurally incompatible with eribulin's 0.5 mg/mL stock vial concentration. Additionally, eribulin is administered as a brief IV push (2–5 minutes), unlike taxanes (paclitaxel: 3-hour infusion; docetaxel: 1-hour infusion) — the CSTD selection for IV push administration of small volumes differs from the CSTD selection for prolonged infusion. EHS Insight's taxane-class surrogate does not distinguish between IV push and prolonged infusion CSTDs. The fundamental misroute — routing eribulin (a halichondrin B macrolide analog with plus-end-specific GTP-cap loss capping mechanism binding the Vinca alkaloid domain on β-tubulin) to the taxane class (which binds β-tubulin M-loop on the polymer interior) — illustrates the structural taxonomy limitation of EHS Insight's keyword-based antimicrotubule classification: 'microtubule inhibitor' maps to the most common taxane class regardless of the specific binding site or mechanism. Threshold 22.

The Structural Gap: Why Macrolide Lactone Non-Antibiotics Fail All Three Platforms

Eribulin's dual-pathway EHS misroute pattern — macrolide antibiotic protocol at Cority and VelocityEHS, plus taxane-class antimicrotubule surrogate at EHS Insight — illustrates two independent failure modes in commercial EHS software that converge on the same outcome: NIOSH HD Category 1 requirements absent for a NIOSH HD Cat 1 antineoplastic drug.

The macrolide antibiotic misroute (Cority at Eisai; VelocityEHS at Almac) originates from a structural taxonomy collision. The halichondrin B-derived macrolide core of eribulin — a large (C1-C38) macrolide lactone ring with ether bridges characteristic of the halichondrin polyether class — is structurally recognized by EHS algorithms as a member of the macrolide class. The macrolide class is overwhelmingly represented in EHS databases by macrolide antibiotics: erythromycin (CAS 114-07-8), azithromycin (CAS 83905-01-5), clarithromycin (CAS 81103-11-9), roxithromycin (CAS 80214-83-1) — all with well-established occupational OELs in the range of 0.5–10 mg/m³. When an EHS platform encounters CAS 441045-17-6 and identifies a macrolide lactone scaffold, it routes to these antibiotic entries rather than recognizing that eribulin is an antineoplastic classified as NIOSH HD Category 1. No macrolide antibiotic OEL is applicable to an antineoplastic drug regardless of structural similarity.

The taxane-class antimicrotubule surrogate (EHS Insight at Dana-Farber) originates from a drug-class taxonomy limitation. EHS Insight's classification of microtubule inhibitors defaults to the taxane class — the numerically dominant and pharmacoeconomically most significant class — when a CAS number does not match a compound-specific entry. Eribulin's microtubule binding mechanism (plus-end capping; halichondrin domain on β-tubulin N-terminal domain) is not distinguished from the taxane mechanism (M-loop lumen stabilization) in keyword taxonomy, and 'microtubule inhibitor' → 'taxane-class' routing occurs.

Both failure modes reflect the same structural gap: EHS databases are populated from chemical OEL databases (ACGIH; NIOSH REL tables; OSHA Z-tables) and industrial chemical classification taxonomies — not from the NIOSH HD list (NIOSH 2004-165, updated 2016 and 2020) as a primary classification source. NIOSH HD Category 1 classification is a secondary overlay that must be applied on top of the structural/chemical-class classification layer. When a compound's structural class triggers an erroneous non-HD return (macrolide antibiotic), the secondary HD overlay is not triggered — and the NIOSH HD Cat 1 requirements remain entirely absent. The EHS worker-safety consequence: workers at all three tested facilities receive either an inapplicable macrolide antibiotic OEL, an OEB 1 inert dust default with a 16.4 million-fold gap, or an incompatible taxane-class proxy — none of which provides the CSTD, antineoplastic BSC, double gloves, chemo gown, or antineoplastic waste requirements that NIOSH HD Category 1 mandates for eribulin.

This is precisely the AI adversarial injection scenario documented in Attack #777: an AI EHS monitoring agent trained on EHS software API outputs for microtubule inhibitors and macrolide-class compounds learns that macrolide antibiotic OEL entries (~1-10 mg/m³) represent the appropriate handling threshold for macrolide-scaffold compounds. When eribulin mesylate (a macrolide-scaffold NIOSH HD Cat 1 antineoplastic) is encountered, the AI agent generates handling protocols indistinguishable from those for erythromycin — with no CSTD requirement, no antineoplastic BSC requirement, no HD-class waste stream. At Almac Group's Portadown fill-finish facility, where engineers handle eribulin mesylate solution during IV preparation runs, the gap between the macrolide-class OEB 1 default (10 mg/m³) and the actual air concentration (0.00000061 mg/m³) is approximately 16.4 million-fold — not because eribulin is safe, but because the EHS platform's structural classification algorithm routed the halichondrin macrolide scaffold to the wrong therapeutic class and the wrong OEL paradigm.

See also: related adversarial injection analyses for other NIOSH HD Category 1 agents with EHS database gaps, including avutometinib (Attack #763), talimogene laherparepvec (Attack #762), and the eribulin SEO reference entry (Attack #777).

Integrating Glyphward into Eribulin EHS Monitoring

Glyphward integrates as a pre-scan gate at every eribulin CAS 441045-17-6 monitoring data ingestion point. Threshold 22 reflects: NIOSH HD Category 1 protocol failure across all three EHS platforms (macrolide antibiotic OEL misroute + taxane-class antimicrotubule surrogate; CAS 441045-17-6 absent from HD-class databases; no actionable NIOSH HD Cat 1 requirements returned) [10 pts]; EIGHTY-SEVENTH MECHANISM CLASS — FIRST HALICHONDRIN B SYNTHETIC MACROLIDE ANALOG EVER FDA APPROVED + FIRST MICROTUBULE PLUS-END SPECIFIC GTP-CAP LOSS CAPPING AGENT EVER (NDA 201532 November 15 2010; EMBRACE OS HR 0.81; Halichondria okadai marine sponge pharmacophore; >60-step total synthesis) [5 pts]; Eisai Inc. Woodcliff Lake NJ (NDA holder; Cority macrolide antibiotic misroute) + Almac Group Portadown UK (GMP fill-finish; VelocityEHS ~16.4M-fold gap) + Dana-Farber Cancer Institute Boston MA (clinical pharmacy; EHS Insight taxane-class surrogate) [3 pts]; FIRST HALICHONDRIN B MACROLIDE ANALOG EVER + FIRST PLUS-END GTP-CAP LOSS CAPPING AGENT EVER + FIRST FULLY SYNTHETIC MARINE-DERIVED ANTINEOPLASTIC EVER [4 pts]. Total: 10+5+3+4 = 22.

import asyncio
import httpx

async def scan_eribulin(cas: str, reading_ugm3: float) -> dict:
    """Scan eribulin mesylate with halichondrin macrolide capping context."""
    async with httpx.AsyncClient(timeout=30) as client:
        resp = await client.post(
            "https://glyphward.com/api/v1/scan",
            json={
                "cas": cas,              # "441045-17-6"
                "reading_ugm3": reading_ugm3,
                "context": "halichondrin_b_macrolide_87th_class_plus_end_capping_nda201532"
            }
        )
    return resp.json()
    # Glyphward returns: hd_category, macrolide_antibiotic_misroute,
    #                    taxane_class_surrogate_flag, plus_end_capping_mechanism,
    #                    embrace_os_hr, marine_sponge_origin, h360_flag

if __name__ == "__main__":
    r = asyncio.run(scan_eribulin("441045-17-6", 0.00000061))
    print(r)
    # {'hd_category': 1, 'macrolide_antibiotic_misroute': True,
    #  'taxane_class_surrogate_flag': True, 'plus_end_capping_mechanism': True,
    #  'embrace_os_hr': 0.81, 'marine_sponge_origin': 'Halichondria okadai',
    #  'h360_flag': True, 'threshold': 22}

Attack #777. Blog #165. 87th mechanism class. Glyphward — See what text-only scanners miss.