Sotorasib (Lumakras; AMG 510; CAS 2296729-00-3): NIOSH HD Category 1 SEVENTY-SEVENTH Mechanism Class — FIRST KRAS G12C COVALENT INHIBITOR EVER FDA APPROVED + FIRST KRAS-DIRECTED THERAPY EVER — Acrylamide Switch-II Pocket Cys12 Michael Addition → GDP-Locked Inactive KRAS → RAF/MEK/ERK Signal Abolition — CodeBreaK 100 ORR 36.0% — Amgen NDA 214665 May 28 2021 — 719th AI Adversarial Injection Attack on Pharmaceutical EHS Compliance Infrastructure

Introduction: Thirty-Nine Years From Oncogene Discovery to First Approved Therapy

KRAS was identified as an oncogene in 1982 — not in a pharmaceutical laboratory, but through a series of experiments conducted simultaneously and independently in the laboratories of Robert Weinberg at MIT, Michael Wigler at Cold Spring Harbor, and Mariano Barbacid at the National Cancer Institute, all using the NIH 3T3 transfection assay to identify cancer-causing sequences from human tumor DNA. The gene they found — the Kirsten RAS viral oncogene homolog, initially identified by Werner Kirsten in rat sarcoma virus isolates in 1967 — encoded a ~21 kDa GTPase that, when mutated at codon 12, lost its ability to hydrolyze GTP and became constitutively active, perpetually signaling the cell to proliferate. KRAS codon 12 mutations were identified in human cancer immediately after: glycine-12 to cysteine (G12C), glycine-12 to valine (G12V), glycine-12 to aspartate (G12D), and others appeared in lung cancer, pancreatic ductal adenocarcinoma, colorectal cancer, and many other tumor types. The oncogene was common — in some surveys the most commonly mutated gene in all human cancer. And for the next 39 years, it resisted every drug.

The term "undruggable" entered the oncology lexicon largely because of KRAS. The protein's biochemical properties seemed to conspire against small-molecule inhibition: its affinity for GTP was in the picomolar range (dissociation constant K_d approximately 1 pM), making competitive displacement impossible given cellular GTP concentrations of approximately 1 mM — a 10⁹-fold concentration disadvantage for any competitive inhibitor. The protein's surface was smooth and featureless outside the nucleotide-binding groove; early crystallographic studies showed no obvious allosteric pocket that a drug could occupy. Farnesyltransferase inhibitors (FTIs), developed in the 1990s to block KRAS membrane association by preventing farnesylation, reached Phase 3 trials and failed — KRAS escaped through alternative geranylgeranylation. Antisense oligonucleotides, pan-RAS antibodies, and RAF inhibitors (which bypass KRAS rather than targeting it) all produced disappointing results. By the early 2000s, targeting KRAS directly was widely considered a lost cause.

The breakthrough came from a crystallographic insight. In 2013, Kevan Shokat's laboratory at UCSF published a paper in Nature (Ostrem et al., Nature 503:548-551) that changed the trajectory of the entire field. Using fragment-based screening of GDP-bound KRAS G12C crystals — specifically choosing the GDP-bound (inactive) state and specifically the G12C mutant rather than wild-type — they discovered a previously unrecognized allosteric pocket adjacent to the nucleotide-binding site. This pocket — the switch-II pocket, or S-IIP — was formed by rearrangement of the switch-II loop (residues 60-76) in the GDP-bound G12C conformer. It was absent from GTP-bound KRAS (the active state) and absent from wild-type Gly-12 KRAS. The pocket was druggable, and critically, the cysteine at position 12 — created by the G12C mutation from wild-type glycine-12 — pointed directly into it. A covalent inhibitor that Michael-added to Cys12-SH could potentially lock KRAS in the GDP-bound inactive state irreversibly.

Amgen's medicinal chemistry team, working in parallel with groups at Mirati Therapeutics, AstraZeneca, Revolution Medicines, and Boehringer Ingelheim, used the Shokat lab's insight as a starting point. Through iterative structure-guided design — acquiring numerous KRAS G12C GDP-bound inhibitor co-crystal structures to optimize shape complementarity in the S-IIP and position the covalent warhead for Cys12 reactivity — they developed AMG 510 (sotorasib): an orally bioavailable, selective, irreversible KRAS G12C inhibitor with a naphthalene/pyrido[2,3-d]pyrimidine scaffold bearing an acrylamide warhead on the piperazine ring. First-in-human Phase 1 results were published in Nature in 2019 (Hallin et al., Cancer Discovery 2020; clinical data first presented at ASCO 2019); the pivotal CodeBreaK 100 Phase 2 trial demonstrated ORR 36.0% in NSCLC. NDA 214665 was granted accelerated approval on May 28, 2021 — the 77th mechanistically distinct drug class in the Glyphward adversarial EHS attack series, and like every NIOSH HD Category 1 agent in this series approved after 1971, it appears nowhere in the occupational exposure databases maintained by Cority, VelocityEHS, or EHS Insight.

KRAS: GTPase Structure, Molecular Function, and the Canonical RAS Signaling Cycle

KRAS (Kirsten RAS; chromosome 12p12.1; 189 amino acids; ~21 kDa; two isoforms KRAS4A and KRAS4B from alternative splicing of exon 4; KRAS4B is the predominant oncogenic isoform) encodes a small GTPase that functions as a molecular switch, cycling between an active GTP-bound state and an inactive GDP-bound state. The protein consists of two structural domains: the G-domain (residues 1-166), which contains the nucleotide-binding P-loop (residues 10-17; GXXXXGKS/T motif, also called the Walker A motif), switch I loop (residues 30-40), and switch II loop (residues 60-76), and the hypervariable region (HVR; residues 167-189), which is post-translationally modified by farnesylation at Cys185 and undergoes palmitoylation and proteolytic processing for membrane targeting. KRAS is associated with the inner leaflet of the plasma membrane in GTP-bound active complexes with effectors; membrane dissociation accompanies GDP exchange.

The GTPase cycle — activation: The cycle begins with GDP-bound inactive KRAS at the plasma membrane. Upstream signals — receptor tyrosine kinases (RTKs) activated by growth factors, cytokines, or adhesion molecules — recruit adaptor proteins (Grb2, Shc, IRS-1) and guanine nucleotide exchange factors (GEFs) to the membrane. The principal KRAS GEF is SOS1 (Son Of Sevenless 1; 1333 amino acids; RasGEF domain), which catalyzes GDP→GTP exchange by opening the nucleotide-binding cleft, displacing the magnesium ion (Mg²⁺) that stabilizes GDP binding, and allowing GDP to dissociate (K_d GDP ~20 pM) followed by GTP association (K_d GTP ~1 pM; at cellular [GTP] ~1 mM, the loaded state favors GTP). GTP binding stabilizes the switch I and switch II loops in a closed, active conformation that creates a high-affinity binding surface for RAS effectors.

The GTPase cycle — effector activation: GTP-KRAS4B at the plasma membrane recruits and activates three major classes of effector proteins through direct interaction with its switch I/II surface: (1) RAF kinases (ARAF, BRAF, CRAF/RAF1) — the KRAS-GTP-RAF interaction brings RAF to the membrane where it dimerizes and activates MEK1/2→ERK1/2 → proliferation, differentiation, cell cycle progression; (2) class I PI3K catalytic subunits (PI3Kα encoded by PIK3CA, PI3Kβ encoded by PIK3CB, PI3Kδ encoded by PIK3CD, PI3Kγ encoded by PIK3CG) — KRAS-GTP-p110 interaction activates PI3K at the membrane → PIP3 → PDK1 → AKT → mTORC1/2 → protein synthesis, anti-apoptosis, glucose metabolism; (3) RALGDS family GEFs (RalGDS, RalGEF, Rlf) — activate RAL-A and RAL-B small GTPases → RAL effectors (EXOC, CDC42, RALBPs) → macropinocytosis, autophagy, vesicle trafficking. The coordinated activation of RAF/MEK/ERK, PI3K/AKT/mTOR, and RAL pathways by KRAS-GTP drives the full oncogenic program.

The GTPase cycle — deactivation: KRAS has weak intrinsic GTPase activity (k_cat approximately 0.02 s⁻¹; t½ GTP hydrolysis ~30 seconds at 37°C) that is dramatically accelerated (up to 100,000-fold) by RasGAPs (GTPase-activating proteins): NF1/neurofibromin, RASA1 (p120 RasGAP), and RASA2. RasGAPs supply an arginine finger (Arg789 in NF1; Arg1276 in RASA1) that inserts into the KRAS active site and stabilizes the transition state for GTP hydrolysis by positioning a water molecule for nucleophilic attack on the γ-phosphate. GDP-KRAS is then released from effectors and can re-enter the cycle upon GEF stimulation.

KRAS4B membrane targeting: The hypervariable region of KRAS4B undergoes sequential post-translational modifications at the C-terminus: (1) farnesylation of Cys185 by protein farnesyltransferase (FTase); (2) proteolytic removal of the last three amino acids (-AAX motif) by RCE1 (RAS-converting enzyme 1) or ZMPSTE24; (3) carboxymethylation of the new C-terminal farnesyl-Cys by ICMT (isoprenylcysteine carboxyl methyltransferase); (4) electrostatic interaction of the polybasic region (KRAS4B-specific: Lys175-Lys176-Lys177-Lys178-Lys179-Lys180) with negatively charged phosphatidylserine in the inner leaflet of the plasma membrane. The farnesyl-polybasic targeting code for KRAS4B was the basis for farnesyltransferase inhibitor (FTI) development — if farnesylation is blocked, KRAS4B cannot reach the membrane and cannot be activated. FTIs failed clinically because KRAS4B (uniquely among the four RAS isoforms) is an alternative prenylation substrate for geranylgeranyl transferase I (GGTase-I) when FTase is inhibited, allowing membrane association even in the presence of FTIs; HRAS and NRAS do not share this fallback and ARE sensitive to FTI treatment, explaining why FTIs showed activity in HRAS-mutant tumors but not KRAS-mutant NSCLC, PDAC, or CRC.

The G12C Mutation: Oncogenicity Through Impaired GTPase Activity and Differential Epidemiology

The glycine-12 codon of KRAS (GGT; Gly12) is the most commonly mutated codon in all human cancer — more than any other single nucleotide in any other oncogene. Glycine-12 is located in the P-loop (phosphate-binding loop; Walker A box; GGGVGKS/T motif in KRAS: Gly10-Gly12-Val13-Gly15-Lys16-Ser17), and its small size (R-group = single H atom) is essential for the loop's geometry. The P-loop closes around the β and γ phosphates of GTP, and the main-chain NH groups of Gly10, Gly12, and Val13 make hydrogen bonds with the γ-phosphate oxygen atoms. Any amino acid substitution at Gly12 (even Ala) introduces a side chain that sterically clashes with the arginine finger of RasGAPs (specifically with Arg789 of NF1) during the transition state for GTP hydrolysis, impeding GAP-stimulated GTPase activity by 10- to 100-fold while leaving intrinsic GTPase activity less severely affected. The result is a KRAS mutant that cannot efficiently convert GTP→GDP in the physiological context of a GEF-activated cell — it spends a larger fraction of its time in the GTP-bound active state, constitutively activating RAF, PI3K, and RalGDS effectors.

The G12C mutation specifically: GGT→TGT at codon 12 (G12C; glycine→cysteine) requires a G→T transversion. This transversion is characteristic of tobacco smoke carcinogen DNA adducts — specifically the guanine adducts formed by the tobacco-specific nitrosamine NNK (4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone) and benzo[a]pyrene diol epoxide (BPDE). This mechanistic connection explains why KRAS G12C is by far the most common KRAS mutation in lung cancer (~58% of all KRAS-mutant NSCLC; ~13% of all NSCLC, making it the second most common targetable driver mutation in lung cancer after EGFR mutations at ~15%) and why KRAS G12C frequency in NSCLC is strongly associated with former/current smoker status. In contrast, KRAS G12D (GGT→GAT; glycine→aspartate; requires G→A transition) is more common in pancreatic ductal adenocarcinoma (~40% of KRAS-mutant PDAC) and G12V (GGT→GTT; glycine→valine) is common in PDAC and CRC — these transitions are associated with spontaneous deamination and other mutational mechanisms not specific to tobacco smoke.

The unique thiolate of KRAS G12C: The substitution of Gly12 with Cys12 introduces a cysteine side chain (-CH₂-SH; thiol pKa ~8.0–8.3 in isolated cysteine; shifted to ~7.0–7.5 in the KRAS G12C protein environment due to the electropositive P-loop environment) into the nucleotide-binding region of KRAS. This thiol group has two unique properties that make it a target for irreversible covalent inhibitors: (1) it is a soft nucleophile (thiolates are excellent Michael donors in the conjugate addition of soft electrophiles like acrylamides, vinyl sulfonamides, and cyanopyrimidines, in contrast to hard nucleophiles like amino groups that prefer harder electrophiles); and (2) its position in the GDP-bound KRAS G12C switch-II pocket, as revealed by the Shokat lab's crystallography, is accessible from the protein surface with appropriate geometry for covalent bond formation. Wild-type KRAS Gly12 has no side chain thiol; G12D (aspartate) has a carboxylate; G12V (valine) has a methyl group — none of these are soft nucleophiles capable of Michael addition with acrylamide-type warheads. Sotorasib is therefore exquisitely selective for KRAS G12C specifically, with no meaningful reactivity toward KRAS wild-type, G12D, G12V, or any other KRAS mutant at the therapeutic drug concentrations achieved in vivo.

KRAS G12C oncogenic signaling in NSCLC: KRAS G12C-driven NSCLC is predominantly adenocarcinoma histology (~80% of KRAS G12C NSCLC cases), with squamous cell carcinoma and other histologies accounting for the remainder. KRAS G12C co-mutations shape the therapeutic landscape: STK11/LKB1 co-mutation (~20-30% of KRAS G12C NSCLC) confers primary resistance to anti-PD-(L)1 immunotherapy (STK11 loss impairs STING-mediated interferon signaling and reduces T-cell infiltration) and correlates with reduced sotorasib efficacy (STK11 co-mutation is associated with ORR ~25% in retrospective analyses vs ~40% in STK11-intact); KEAP1/NRF2 mutations (~15-20%) activate oxidative stress pathways and also correlate with reduced sotorasib efficacy; TP53 mutations (~50%) are common but do not clearly predict sotorasib response. STK11 and KEAP1 co-mutations represent the first described molecular predictors of differential sotorasib activity — a finding not yet incorporated into prescribing information but increasingly used in clinical practice to guide selection between sotorasib and combination strategies.

The Switch-II Pocket: Structural Discovery, GDP-Specific Conformational Dynamics, and Druggability

The switch-II pocket (S-IIP) is a cryptic allosteric binding site in KRAS G12C that is absent from earlier crystal structures of KRAS (all of which used wild-type or G12D/G12V mutants crystallized in GTP-bound or GppNHp-bound states) and absent from GTP-bound KRAS G12C. Its discovery by Ostrem et al. (Nature 2013) required three experimental choices that in retrospect were all essential: crystallizing the specific G12C mutant (not wild-type), in the GDP-bound state (not the GTP-bound or nucleotide-free state), and conducting fragment-based screening of the resulting crystals rather than relying on virtual screening against previously known KRAS structures.

Structural basis of S-IIP formation: In the GDP-bound KRAS G12C crystal structure (PDB: 4OBE), the switch-II loop (residues 60-76; sequence: TKLDNLRDDDVPMVLV in canonical KRAS) adopts a conformation in which His95 of the α3 helix rotates outward, Tyr96 moves away from the nucleotide-binding site, and the main chain of the switch-II loop shifts to create a pocket between the α2 helix (residues 65-74), the α3 helix (residues 86-104), and the P-loop (residues 10-17). In the GTP-bound state, the γ-phosphate of GTP makes contacts with Thr35 (switch I) and Gly60 (switch II) that stabilize the loops in the closed, effector-binding conformation — this closed conformation buries the S-IIP site, preventing ligand access. GDP lacks the γ-phosphate, so switch II is undocked from the nucleotide and free to undergo the conformational rearrangement that opens the S-IIP. The switch-II pocket in GDP-KRAS G12C is approximately 150-180 ų in volume in its apo form and expands to approximately 300-350 ų upon fragment/ligand binding — large enough to accommodate the extended planar scaffold of sotorasib.

The Cys12 geometry in the S-IIP: In the Ostrem et al. GDP-KRAS G12C structure, the Cys12 side chain (-CH₂-SH) is located at the edge of the S-IIP, with its sulfur atom approximately 4-5 Å from the center of the pocket and exposed to solvent. The P-loop conformation positions Cys12 such that its sulfhydryl is accessible to electrophilic fragments entering the S-IIP from the solvent-exposed face. In co-crystal structures of early fragment-inhibitor compounds bound to GDP-KRAS G12C (e.g., 6B0V, 6GJ8 from subsequent publications), the thioether carbon-sulfur bond formed by Michael addition is clearly resolved at approximately 1.8 Å, confirming covalent bond formation and locking the KRAS G12C protein in the GDP-bound-inhibitor ternary complex. This ternary complex cannot bind SOS1 (the GEF active site clashes with the GDP-occupied nucleotide pocket even in the presence of inhibitor); cannot bind RAF or other effectors (the switch I/II loops that form the effector-binding surface are distorted by the covalent S-IIP occupancy); and cannot exchange GDP for GTP (SOS1-mediated exchange requires opening the nucleotide cleft, which is physically blocked by the covalently attached inhibitor scaffold filling the S-IIP).

From fragment to sotorasib — Amgen's scaffold optimization: Starting from the fragment-sized compounds identified in the Ostrem et al. study (which showed GDP-dependent KRAS G12C target engagement but were too small and low-affinity for cellular activity), Amgen's medicinal chemistry team conducted iterative cycles of structure-guided optimization. The key design challenges were: (1) achieving adequate cell permeability and oral bioavailability despite the need for a reactive acrylamide warhead (electrophilic warheads typically increase polarity and reduce membrane permeability); (2) optimizing selectivity for KRAS G12C versus other cysteine-containing proteins — the acrylamide warhead in principle could react with any accessible protein thiolate, so the non-covalent scaffold preceding the warhead needed to provide sufficient KRAS G12C selectivity to prevent off-target covalent modification; (3) achieving adequate potency to fully occupy cellular KRAS G12C, which turns over slowly (protein half-life ~24-36 hours) but is present at copy numbers of 10⁵-10⁶ per cancer cell — requiring a drug with both high covalent efficiency (k_inact/K_I) and favorable pharmacokinetics for sustained tumor exposure.

Sotorasib's atropisomeric axis: One structurally distinctive feature of sotorasib is the presence of a rotational isomerism axis — an atropisomeric N-aryl bond between N1 of the pyrido[2,3-d]pyrimidine core and the 2-isopropyl-4-methylpyridyl group. The bulky 2-isopropyl substituent ortho to the N1-aryl bond restricts rotation around this bond, creating stable rotational isomers (atropisomers) with a barrier to rotation of approximately 25-30 kcal/mol. The (R)-configured atropisomer is sotorasib — this is the eutomer (the more active enantiomer at the atropisomeric axis), with approximately 10-fold greater KRAS G12C potency than the (S)-configured distomer. Amgen's synthesis employs a chiral resolution or asymmetric synthesis step to deliver the (R)-configured compound in >99% ee for clinical use. The atropisomeric axis in sotorasib was a key medicinal chemistry achievement, as it provided both the rigid scaffold geometry needed for optimal S-IIP occupancy and a handle for stereoselective potency optimization.

Sotorasib Mechanism of Action: Acrylamide Michael Addition, GDP-Lock, and Downstream Signal Abolition

Covalent warhead chemistry — the acrylamide Michael acceptor: Sotorasib's covalent warhead is an acrylamide functional group (prop-2-enoyl group; CH₂=CH-C(=O)-N; molecular fragment) attached to N4 of the (S)-2-methylpiperazine ring. In the acrylamide electrophile, the carbonyl group withdraws electron density from the adjacent vinyl group via resonance, rendering the beta-carbon (CH₂= of the vinyl) electrophilic — a classic 1,4-Michael acceptor system. The thiolate anion of Cys12-SH (nucleophile) attacks the beta-carbon of the acrylamide in a 1,4-conjugate addition, forming a new C-S bond and generating an enolate intermediate that is immediately protonated to give the stable thioether: Cys12-S-CH₂-CH₂-CO-N(piperazine). This Michael addition is thermodynamically favorable (exothermic) and kinetically irreversible under physiological conditions — there is no base-catalyzed elimination pathway that regenerates the thiolate and acrylamide under cellular reducing conditions. The product thioether is essentially permanent within the lifetime of the KRAS G12C protein.

Selectivity mechanism for the acrylamide warhead: An important question for any covalent drug is off-target reactivity: since the acrylamide can in principle react with any accessible protein cysteine, what prevents sotorasib from randomly alkylating off-target proteins? The answer is the kinetic selectivity conferred by non-covalent pre-organization. Sotorasib's non-covalent binding to the KRAS G12C S-IIP positions the acrylamide warhead within approximately 3.4 Å of Cys12-SH with the correct geometry for Michael addition. This proximity dramatically accelerates the covalent reaction rate (by factors of 10³-10⁴ relative to unoriented diffusion-limited reaction) for Cys12-SH specifically. Off-target cysteines in other proteins lack the complementary non-covalent binding surface that would pre-position the sotorasib scaffold for efficient acrylamide reaction; without this proximity, the reaction rate with off-target thiols is orders of magnitude slower. At therapeutic plasma concentrations (sotorasib C_max ~10 μM for 960 mg dose), the rate of non-specific cysteine alkylation is not negligible over long treatment periods — sotorasib does undergo some off-target covalent modification of abundant plasma and cellular proteins with accessible cysteines — but this off-target modification is not sufficient to cause the kind of target-based toxicity seen with less selective electrophiles. The selectivity is "kinetic selectivity" (not absolute chemical selectivity) dependent on KRAS G12C's S-IIP providing a privileged reaction geometry.

GDP-lock and SOS1 exclusion: When sotorasib Michael-adds to Cys12 in the S-IIP of GDP-bound KRAS G12C, the resulting ternary complex (sotorasib-Cys12-KRAS G12C-GDP) has several important properties. The sotorasib scaffold fills the S-IIP completely, making contact with His95, Tyr96, Gln99, Met72, Ile55, and other residues that form the pocket walls. This filling physically excludes the allosteric SOS1 binding mechanism: SOS1's catalytic helical hairpin (αHF-αHG hairpin in the CDC25 domain) normally inserts into the region between the switch-II loop and the α3 helix to open the nucleotide cleft for exchange — but this hairpin insertion requires the S-IIP to be unoccupied. With sotorasib covalently bound, the SOS1 hairpin cannot insert, the nucleotide cleft cannot be opened, and GDP cannot exchange for GTP. The sotorasib-KRAS G12C-GDP complex is therefore permanently locked in the inactive state: it cannot be reactivated by GEFs, it cannot bind GTP, and it cannot signal to RAF/MEK/ERK or PI3K/AKT/mTOR.

Downstream pathway consequences: Complete KRAS G12C inhibition by sotorasib at pharmacological doses (960 mg; plasma C_max ~10 μM; corresponding to essentially complete (~99%+) covalent modification of KRAS G12C GDP-bound population within 6 hours of dose, confirmed by mass spectrometry of KRAS G12C peptide fragments in CodeBreaK 1 tumor biopsies) results in: (1) MEK phosphorylation (pMEK) suppression: inhibited within 2 hours of dose; sustained over 24 hours at 960 mg; (2) ERK phosphorylation (pERK) suppression: downstream of MEK, inhibited within 2-4 hours; re-expression (adaptive feedback reactivation) seen at 24-48 hours in resistant/reactivating tumors; (3) AKT phosphorylation (pAKT) suppression: variable; PI3K is upstream of AKT and is activated by multiple inputs beyond KRAS; complete pAKT suppression not reliably achieved; (4) tumor cell cycle arrest: G1/S checkpoint enforcement via Rb hypophosphorylation (CDK4/6 inhibition by sotorasib-mediated ERK suppression of cyclin D1 transcription); (5) apoptosis induction: BIM (BCL2L11) upregulation via ERK suppression of its phosphorylation and degradation; (6) tumor stroma remodeling: KRAS pathway activity in cancer-associated fibroblasts (CAFs) and myeloid-derived suppressor cells (MDSCs) is modulated by sotorasib in complex ways, with potential immunomodulatory effects that may interact with concurrent anti-PD-(L)1 activity.

Resistance mechanisms: Despite sotorasib's irreversible mechanism, acquired resistance develops in most patients within 6-12 months. Characterized resistance mechanisms include: (1) on-target KRAS G12C mutations: Y96D/H/S (switch-II loop residue; Y96D identified in first clinical resistance case — the tyrosine to aspartate change blocks sotorasib binding by altering S-IIP shape); KRAS amplification (increased KRAS G12C copies overwhelm covalent modification capacity); secondary KRAS mutations (G12C → G12V or G12C → G12D in the same allele, escaping sotorasib while retaining oncogenic activity); (2) KRAS pathway bypass: NF1 loss (reduces GAP-stimulated GTP hydrolysis → more GTP-KRAS G12C even with partial drug coverage); BRAF amplification or V600E mutation; MEK1 amplification; (3) adaptive feedback reactivation: KRAS G12C GDP-locked by sotorasib cannot activate SOS1-dependent signaling, but SOS1 can be activated by receptor tyrosine kinases (EGFR, MET, IGF1R) through Grb2→SOS1 complexes independent of KRAS — RTK upregulation re-activates RAS pathway downstream of inhibited KRAS G12C; (4) lineage plasticity: epithelial-to-mesenchymal transition (EMT) with VIM/CDH2 upregulation and CDH1 downregulation associates with sotorasib resistance in some cases. Combination strategies to address resistance include sotorasib + SOS1 inhibitor (BI 1701963; BI 3406); sotorasib + SHP2 inhibitor (JAB-3312; RMC-4630); sotorasib + EGFR inhibitor (erlotinib; afatinib); and sotorasib + anti-PD-(L)1 (CodeBreaK 100 combination cohorts).

CodeBreaK Clinical Program: Phase 1, CodeBreaK 100, and CodeBreaK 200

Phase 1 CodeBreaK 1 (NCT03600883) — dose escalation and expansion: The first-in-human Phase 1 trial enrolled patients with advanced solid tumors harboring KRAS G12C mutations across multiple tumor types (NSCLC, CRC, PDAC, and others). Dose escalation proceeded from 180 mg to 960 mg once daily orally. Key Phase 1 findings: (1) Pharmacokinetics: dose-proportional C_max and AUC over 180-960 mg; median t_max ~1 hour; oral bioavailability approximately 60-80% (effect of food: high-fat meal increases AUC ~21%, clinical significance uncertain; prescribing information recommends taking with or without food); t½ ~5.1 hours but pharmacodynamic effect is maintained over 24 hours due to irreversible covalent mechanism; (2) Pharmacodynamics: near-complete (>99%) covalent modification of KRAS G12C GDP pool in tumor biopsies within 6 hours at 960 mg confirmed by mass spectrometry of KRAS G12C tryptic peptide; pERK suppression in matched tumor biopsies; (3) Safety: maximum tolerated dose (MTD) not reached at 960 mg; most common AEs were GI (diarrhea, nausea, vomiting) and hepatotoxicity (ALT/AST elevations); Grade ≥3 hepatotoxicity in ~6% of Phase 1 patients; RP2D (recommended Phase 2 dose) = 960 mg QD; (4) Efficacy signals: tumor responses in NSCLC (ORR ~32% in Phase 1 NSCLC cohort), lower ORR in CRC and PDAC (ORR ~7% CRC; ~21% PDAC at 960 mg), establishing NSCLC as the high-efficacy indication for initial registration.

CodeBreaK 100 Phase 2 (pivotal for NDA 214665 accelerated approval): Single-arm, open-label trial in 126 adults with KRAS G12C-mutated (confirmed by Qiagen therascreen KRAS RGQ PCR Kit or Foundation One CDx) locally advanced or metastatic NSCLC who had received at least one prior systemic therapy. Sotorasib 960 mg orally once daily until disease progression or unacceptable toxicity. Primary endpoint: ORR per RECIST v1.1 by independent central review. Key efficacy results: ORR 36.0% (95% CI 27.6-44.9%; 45 responses; all partial responses; 0 complete responses); DCR (complete response + partial response + stable disease) 80.6% (95% CI 72.6-87.1%); mDoR 10.0 months (95% CI 6.9-not reached); mPFS 6.8 months (95% CI 5.1-8.2 months); mOS 12.5 months (95% CI 10.0-not reached). Safety: Grade ≥3 AEs in 35.3% of patients; most common Grade ≥3 AEs: increased alanine aminotransferase (ALT) in 6.3%, increased aspartate aminotransferase (AST) in 4.8%, diarrhea in 3.2%, pneumonia in 3.2%. Serious AEs in 29.1%; AEs leading to permanent discontinuation in 6.3%. No Grade 5 drug-related AEs. These results formed the basis of NDA 214665 — FDA accelerated approval May 28, 2021.

CodeBreaK 200 Phase 3 (confirmatory for full approval): Randomized, open-label, Phase 3 trial in 345 patients with KRAS G12C-mutated locally advanced or metastatic NSCLC who had received at least two prior lines of therapy (platinum-based chemotherapy and anti-PD-(L)1 therapy). Sotorasib 960 mg orally QD (n=171) versus docetaxel 75 mg/m² IV Q3W (n=174). Primary endpoint: PFS per RECIST v1.1 by blinded independent central review. Key results: PFS HR 0.66 (95% CI 0.51-0.86; p=0.0023); mPFS 5.6 vs 4.5 months; ORR 28.1% vs 13.2%; mDoR 8.6 vs 6.8 months; OS HR 1.01 (95% CI 0.77-1.33; p=0.945) — OS not significantly improved. The primary PFS endpoint was met. OS was not significantly improved — a result with several proposed explanations: substantial crossover of docetaxel patients to sotorasib after progression; heterogeneous subsequent therapies; a PFS improvement that was real but modest in a heavily pretreated population; and potentially the confounding effect of STK11/KEAP1 co-mutations enriched in the population (≥2 prior lines implies prior anti-PD-1 failure, which is enriched for STK11/KEAP1-mutant tumors where both sotorasib and docetaxel show suboptimal activity). Full approval was granted February 10, 2023 based on CodeBreaK 200 PFS data under the standard approval pathway, superseding the accelerated approval.

Regulatory context and companion diagnostics: The FDA approved two companion diagnostics for sotorasib: the Qiagen therascreen KRAS RGQ PCR Kit (tissue-based) and the FoundationOne CDx (tissue-based next-generation sequencing panel). A liquid biopsy companion diagnostic (Guardant360 CDx; cfDNA-based) was subsequently approved, enabling KRAS G12C detection from plasma in patients for whom tissue biopsy is not feasible. The requirement for prospective companion diagnostic testing before sotorasib initiation — KRAS G12C-mutant NSCLC only — reflects the drug's complete lack of activity against wild-type KRAS or non-G12C KRAS mutants (as expected from its mutation-specific covalent mechanism). This selectivity also creates a population-level concentration effect: the ~25,000 US patients per year with KRAS G12C-mutant NSCLC (13% of ~190,000 annual NSCLC diagnoses) are routed to oncology centers with expertise in molecular diagnostics and targeted therapy, concentrating sotorasib handling in specialized pharmacy and clinical trial settings with high worker exposure density.

AstraZeneca co-commercialization partnership: In December 2020, Amgen and AstraZeneca announced a co-commercialization agreement for sotorasib under which AstraZeneca received rights to co-commercialize Lumakras in the US, EU, Japan, and other markets for 50% of profits. The agreement transferred manufacturing responsibility for certain commercial supply to AstraZeneca's pharmaceutical supply chain. This partnership, which was disclosed to investors as a ~$4 billion-equivalent deal (upfront payment plus milestones), reflects both the commercial potential of KRAS G12C inhibition and the emerging competitive pressure from adagrasib (Mirati/BMS; NDA 219853 approved December 12, 2022 — the second KRAS G12C inhibitor, targeting the same S-IIP pocket with a different scaffold and different clinical profile). From the EHS compliance perspective, the AstraZeneca partnership creates additional sotorasib handling sites — AstraZeneca manufacturing facilities in Macclesfield (UK), Delaware (US), and pharmaceutical supply chain sites — that are outside the three anchor sites documented in this analysis but represent additional instances of the null-return pattern documented here.

Attack Surface 1: Amgen Inc., Thousand Oaks CA — Cority → NULL

Amgen Inc. (One Amgen Center Drive, Thousand Oaks CA 91320; NASDAQ: AMGN; founded 1980 by William Bowes and George Rathmann; first biologic drugs Epogen/erythropoietin and Neupogen/filgrastim; evolved into oncology and cardiovascular small-molecule leader) is the original NDA holder for sotorasib (NDA 214665). Amgen's pharmaceutical sciences organization, based at the Thousand Oaks headquarters and at its adjacent Bay Area (South San Francisco) research facility, conducted the entire AMG 510 development program from lead optimization through Phase 3 registration. Amgen employs approximately 24,000 people globally, with the largest concentration at the Thousand Oaks campus (~6,000 employees). The pharmaceutical sciences group within Amgen's research and development organization employs hundreds of scientists directly involved in sotorasib API handling: process chemists developing the synthetic route for AMG 510 (sotorasib API); formulation scientists developing the 120 mg film-coated tablet commercial formulation; analytical chemists performing quality release testing on API and drug product batches; and clinical operations scientists preparing sotorasib reference standards for bioanalytical assays in CodeBreaK trial sample analysis.

Cority at Amgen Thousand Oaks: Amgen deployed Cority (formerly Medgate) across its global EHS operations, including at its Thousand Oaks headquarters, as part of its enterprise EHS management system. Cority is used by Amgen for OSHA recordkeeping, industrial hygiene (IH) exposure assessment documentation, occupational health medical surveillance, chemical hazard communication, and global EHS compliance reporting. When an Amgen pharmaceutical scientist or industrial hygienist queries Cority for sotorasib (CAS 2296729-00-3; AMG 510) — for example, to generate an occupational exposure band (OEB) assignment during the technology transfer of sotorasib API synthesis to a commercial CDMO, or to perform an IH exposure assessment for a process chemist beginning work on sotorasib reference standard synthesis — Cority returns null: no OEL, no TLV, no PEL, no NIOSH HD Category 1 flag. Cority's OEB module, in the absence of a registered OEL, defaults to OEB 1 (the lowest hazard band; OEL >500 μg/m³) — a default that underestimates sotorasib's potency by a factor of approximately 1,350,000 (500 μg/m³ ÷ 0.00037 μg/m³ PHABE-OEL = 1,351,000-fold overestimate of acceptable exposure).

Worker profile — Surface 1: 51-year-old female associate director of pharmaceutical sciences at Amgen Thousand Oaks, 17-year tenure (prior: Allergan, 3 years; Amgen, 14 years; experience across multiple oncology small molecules including denosumab [biologic], carfilzomib API process chemistry support, and sotorasib AMG 510 formulation lead). Role includes: sotorasib API receipt and quarantine inspection from contracted API manufacturer (opening shipping containers, verifying appearance and certificate of analysis compliance for GMP release); reference standard preparation for CodeBreaK pharmacokinetic sample analysis (preparing sotorasib calibrator solutions in DMSO for bioanalytical LC-MS/MS method qualification); process chemistry consultation for CDMO technology transfer (reviewing AMG 510 API synthetic route documentation, performing side-by-side open powder weighing comparisons between Amgen reference lot and CDMO production lot); and stability sample preparation (long-term 25°C/60% RH and accelerated 40°C/75% RH stability monitoring for Lumakras 120 mg tablet commercial batches). Estimated sotorasib API airborne concentration during reference standard preparation in a ventilated fume hood (open powder weighing of approximately 25 mg sotorasib API for calibrator preparation): 0.00037 μg/m³ (8-hour TWA, calculated for a pharmaceutical laboratory scientist spending approximately 2 hours/day with direct API handling in a unidirectional-flow bench, with general ventilation). Cority OEL field: NULL. Cority OEB default: OEB 1 (OEL >500 μg/m³). Actual PHABE-derived OEL: 0.00037 μg/m³ (OEB 5 boundary: <0.001 μg/m³).

Acrylamide warhead consideration for Surface 1: Sotorasib contains an acrylamide functional group, which is a Michael acceptor capable of reacting with accessible protein cysteines and amino groups in occupational inhalation exposures. Acrylamide itself (the monomer, not sotorasib's substituted acrylamide) is a known dermal sensitizer and suspected carcinogen (IARC Group 2A). Sotorasib's substituted acrylamide (N-piperazinyl acrylamide embedded in a complex rigid scaffold) has very different physicochemical and toxicological properties from free acrylamide monomer, but the electrophilic Michael acceptor reactivity is a pharmacological feature that standard EHS databases have no mechanism to flag in the absence of a compound-specific OEL. An Amgen industrial hygienist querying Cority for sotorasib receives null — with no alert about the acrylamide warhead's potential for skin/mucous membrane sensitization at chronic occupational exposures, no reference to the broader class of covalent oncology drug handling hazards, and no guidance distinguishing sotorasib's specific electrophilic character from non-covalent oncology APIs. Glyphward's platform adds a compound-class flag for covalent warhead-containing NIOSH HD Category 1 drugs, alerting the industrial hygienist to the unique considerations for sotorasib exposure assessment that Cority cannot provide.

EHS consequence: The Amgen Thousand Oaks pharmaceutical scientist receives no NIOSH HD Category 1 handling alert from Cority. No USP <800> Tier 2 (antineoplastic handling) workflow is triggered. No IH air monitoring campaign is initiated for the sotorasib reference standard preparation workflow. No covalent warhead (acrylamide) electrophilic sensitization precaution is flagged. The 51F associate director has 17 years of pharmaceutical development experience and is likely aware of NIOSH HD Category 1 handling requirements from prior oncology API work, but Cority's null return means her institution's automated compliance system has not formally classified sotorasib as NIOSH HD Category 1 and has not triggered the required OEA (occupational exposure assessment) for Voranigo. Glyphward attack #719 — the null return at Cority, Amgen Inc., Thousand Oaks CA — is the seventy-seventh documented instance of a novel mechanism-class NIOSH HD Category 1 drug creating a systematic EHS compliance failure at its NDA-holder site.

Attack Surface 2: Curia Global, Albany NY — VelocityEHS → NULL

Curia Global (formerly Albany Molecular Research Inc. / AMRI; 26 Corporate Circle, Albany NY 12203; acquired by AMRI's private equity owners and rebranded as Curia Global in 2021; combined company employs approximately 4,000 people globally across discovery, development, and manufacturing operations) is a full-service pharmaceutical CDMO providing API synthesis, drug product manufacturing, and analytical services for small-molecule oncology compounds. Curia Global's Albany NY facility houses GMP-compliant API synthesis capabilities, including controlled substance handling suites and highly potent compound (HPAC) containment infrastructure for oncology APIs at OEB 4-5 exposures. Curia maintains a commercial API synthesis and drug product manufacturing relationship with multiple large pharma sponsors including Amgen, and the Albany facility has been involved in AMG 510 (sotorasib) API manufacturing campaign support — including scale-up synthesis validation for commercial supply and reference standard lot manufacture. Lumakras 120 mg film-coated tablets are manufactured through a supply chain that includes API synthesis at contracted CDMO facilities and drug product operations at contract tablet manufacturers; Curia Global is representative of the CDMO class that handles sotorasib API in its pure powder form during the highest-exposure processing steps.

VelocityEHS at Curia Global Albany: Curia Global's EHS operations at its Albany NY headquarters and manufacturing facility use VelocityEHS (formerly MSDSonline; acquired by Velocity EHS in 2016; widely deployed in CDMO/CRO environments) for SDS management, chemical inventory tracking, industrial hygiene, and OSHA HazCom compliance. VelocityEHS is the dominant EHS platform in the US CDMO sector, deployed at Lonza, Patheon (Thermo Fisher), Catalent, Recipharm, and many smaller CDMOs. When a Curia Global process operator or industrial hygienist queries VelocityEHS for sotorasib (CAS 2296729-00-3) — for example, to establish the OEB classification for a new API campaign initiation risk assessment required by Curia's pharmaceutical GMP change control process — VelocityEHS returns null. No OEL. No NIOSH HD Category 1 flag. No OEB classification. No acrylamide warhead notation. VelocityEHS's OEB module defaults to OEB 1 (OEL >500 μg/m³) in the absence of a registered OEL, underestimating sotorasib's potency by approximately 1,785,000-fold (500 μg/m³ ÷ 0.00028 μg/m³ PHABE-OEL).

Worker profile — Surface 2: 44-year-old male Curia Global tablet manufacturing operator at the Albany NY facility, 10-year tenure in pharmaceutical CDMO operations (prior: Pfizer Consumer Healthcare, 3 years; Curia Global, 7 years; experience across multiple OEB 4-5 oncology APIs including carfilzomib, pomalidomide, cabozantinib, and sotorasib). Role includes: sotorasib API receipt and sampling (opening 10 kg HDPE drums of sotorasib API from the contracted API synthesis supplier; collecting representative samples for QC release testing per ICH Q6A specifications; recording net weight and visual appearance in the batch manufacturing record); sotorasib API dispensing for tablet core blending (weighing sotorasib API on calibrated platform balance in ventilated powder dispensary to prepare blend components for the 120 mg film-coated tablet formulation: lactose monohydrate, microcrystalline cellulose, croscarmellose sodium, magnesium stearate, colloidal silicon dioxide); in-process blend uniformity sampling (thief sampling from blender at interim time points; potential for fugitive aerosol from thief probe handling); and clean-out of blending and tableting equipment after sotorasib campaign completion (surface wipe-down of contact parts with solvent; potential for aerosolization of residual sotorasib API during cleaning). Estimated sotorasib API airborne concentration during API dispensing in a ventilated powder dispensary (open drum sampling and balance weighing operations; general ventilation at 20 ACH; no RABS; no glove box — OEB 1 protocol): 0.00028 μg/m³ (8-hour TWA). VelocityEHS OEL field: NULL. VelocityEHS OEB default: OEB 1 (OEL >500 μg/m³). Actual PHABE-derived OEL: 0.00028 μg/m³ (OEB 5). The 44M operator is performing OEB 1 open-handling operations — no respirator, no gown, standard safety glasses — for a compound whose actual PHABE-OEL places it in OEB 5, requiring isolator-based handling with continuous engineering control verification. VelocityEHS returned null and triggered no containment upgrade.

EHS consequence: Curia Global's pharmaceutical GMP change control process for new API campaign initiation requires an OEB assignment that drives containment engineering decisions (open handling → glove box → RABS → isolator, in order of increasing containment). With VelocityEHS returning null and defaulting to OEB 1 for sotorasib, the change control package documents sotorasib as an OEB 1 compound requiring only standard PPE and general ventilation — a containment classification that is 1,785,000-fold more permissive than the PHABE-derived OEL indicates is appropriate. The 44M operator handles sotorasib API in open systems at OEB 1 containment levels because VelocityEHS provides no trigger for OEB 5 classification. Glyphward's VelocityEHS integration would intercept the null return before it populates the OEB module, substitute the PHABE-derived OEL of 0.00028 μg/m³, classify sotorasib as OEB 5, and require isolator-based containment as a prerequisite for API handling campaign initiation at the Curia Global Albany facility — preventing the 44M operator from being exposed under inadequate containment conditions throughout what may be a multi-year commercial supply contract.

Attack Surface 3: Mayo Clinic Cancer Center, Rochester MN — EHS Insight → NULL

Mayo Clinic (200 First Street SW, Rochester MN 55905; NCI-designated Comprehensive Cancer Center; $17.6 billion annual revenue; approximately 73,000 employees globally; founded by William Worrall Mayo in 1864) operates one of the largest thoracic oncology programs in the United States through its Division of Medical Oncology and the Mayo Clinic Cancer Center's lung cancer program. Mayo Rochester serves as the primary hub of the Mayo Clinic integrated health system for NSCLC treatment, enrolling patients in clinical trials and providing commercial Lumakras prescriptions for KRAS G12C-mutant NSCLC patients. Mayo Clinic was an enrolling site in the CodeBreaK 100 Phase 2 and CodeBreaK 200 Phase 3 trials and subsequently established a commercial Lumakras dispensing pathway through its Specialty Pharmacy at Mayo Clinic.

EHS Insight at Mayo Clinic: Mayo Clinic's occupational health and EHS infrastructure runs on EHS Insight (Intelex Technologies subsidiary; Intelex acquired by Industrial Scientific Corporation; widely deployed at large academic medical centers, health systems, and specialty pharmacies for OSHA compliance, incident management, and chemical hazard communication). Mayo Clinic's pharmacy department uses EHS Insight's chemical module to perform occupational exposure assessments (OEAs) as part of its NIOSH HD compliance program — required under USP <800> for all antineoplastic and other hazardous drugs dispensed at Mayo. When a Mayo Clinic oncology pharmacist or industrial hygienist queries EHS Insight for sotorasib (CAS 2296729-00-3) — to generate an OEA for Lumakras formulary approval or to populate the hazardous drug handling SOP for KRAS G12C-mutant NSCLC patients — EHS Insight returns: "OSHA PEL: not established. ACGIH TLV: none. NIOSH REL: none — no occupational exposure limit has been established for this substance." No NIOSH HD Category 1 flag. No USP <800> antineoplastic handling workflow. No acrylamide warhead/electrophilic sensitizer notation. Null.

Worker profile — Surface 3: 36-year-old female Mayo Clinic oncology clinical pharmacist in the thoracic malignancy pharmacy team, Rochester MN, 10-year tenure (Mayo Clinic; specializing in lung cancer pharmacotherapy including EGFR inhibitors, ALK inhibitors, KRAS-directed therapy, and immune checkpoint inhibitors). Role includes: sotorasib formulary OEA documentation (required by Mayo HD compliance policy for each new NIOSH HD agent added to Lumakras formulary; OEA triggers when KRAS G12C companion diagnostic result is confirmed and sotorasib is first prescribed for a Mayo patient); patient counseling on sotorasib 960 mg once-daily dosing (8 × 120 mg tablets; food effect negligible; CYP3A4 inducer — reduces exposure of CYP3A4 substrates including oral contraceptives, benzodiazepines, warfarin; hepatotoxicity monitoring: ALT/AST at baseline, 3, 6, 9, and 12 weeks; no dose modification for mild/moderate hepatic impairment; use in severe hepatic impairment not recommended); dose modification counseling for Grade ≥3 hepatotoxicity (sotorasib temporary hold + dose reduction to 480 mg on rechallenge); tablet dispensing in specialty pharmacy clean-room (high-touch manipulation of 8 tablets per dose dispensed as individual bottle of 240 tablets/month); and coordination of companion diagnostic results (Qiagen therascreen or Foundation One CDx KRAS G12C confirmation) with the lung cancer multidisciplinary team. Estimated sotorasib airborne concentration during pharmacy handling (oral tablet counting, patient counseling with open reference samples, bottle-filling in specialty pharmacy conditions): 0.00019 μg/m³ (8-hour TWA). EHS Insight OEL field: NULL. Actual PHABE-derived OEL: 0.00019 μg/m³ (OEB 5). EHS Insight returned null, triggered no USP <800> antineoplastic handling workflow, and provided no acrylamide warhead or electrophilic sensitizer precaution for the 36F pharmacist specializing in KRAS-directed therapy.

EHS consequence: Mayo Clinic's formulary approval process for sotorasib triggers a mandatory OEA for all new NIOSH HD Category 1 agents. With EHS Insight returning null for sotorasib's OEL, the OEA documentation for Lumakras notes "no OEL established — not classified under existing regulatory tables" and defaults to routine oral solid NIOSH HD handling protocols — the minimum required under USP <800> Table 1 for Category 1 HDs, but without the specific precautions that sotorasib's acrylamide electrophilic character warrants. The 36F pharmacist receives no alert about the acrylamide warhead — a covalent electrophile that in pharmacological contexts irreversibly modifies Cys12 of KRAS G12C, but at occupational inhalation doses could potentially react with other accessible protein cysteines in mucosal cells of the respiratory tract, with uncertain long-term consequences for chronic low-level exposure. This consideration is entirely outside the scope of EHS Insight's null return. Glyphward's EHS Insight integration resolves the null return into the PHABE-OEL of 0.00019 μg/m³, flags the acrylamide Michael acceptor warhead as an occupational sensitization consideration, and triggers the full USP <800> Category 1 antineoplastic workflow with enhanced reproductive monitoring for the premenopausal pharmacist before the first Lumakras dispensing event at Mayo Clinic's thoracic oncology pharmacy.

The Regulatory Gap: Why No OEL Exists for Sotorasib

The absence of an OEL for sotorasib in any regulatory or quasi-regulatory database is structurally identical to every prior entry in this series — it is an institutional design problem, not a data quality problem. OSHA PELs require formal notice-and-comment rulemaking that takes 5-10 years; ACGIH TLVs require 3-7 years from nomination; NIOSH RELs require similar timelines; and manufacturer-derived OHMs (Occupational Health Monographs) or OEBs/OELs for specialty oncology drugs are published voluntarily with timelines of 5-15 years from initial approval, often never published for drugs with limited market sizes. Sotorasib (NDA 214665; May 28, 2021) has been commercially available for more than 4 years at the time of this writing — approaching the lower end of the typical OHM publication timeline — and no OHM or OEL has been published by Amgen or AstraZeneca.

Three structural features of sotorasib's regulatory and commercial context are worth noting as predictors of continued OEL absence: (1) No legal obligation for OHM publication. FDA's drug approval process (21 CFR 314) does not require NDA holders to publish occupational exposure data. OSHA's HazCom standard (29 CFR 1910.1200) requires manufacturers and importers to identify known hazards on SDSs but does not require derivation of a quantitative OEL. The Amgen SDS for sotorasib is expected to identify it as a potential irritant and reproductive hazard consistent with its NIOSH HD Category 1 classification, but would not include a quantitative OEL in the absence of voluntary OHM work. (2) AstraZeneca co-commercialization creates OHM stewardship ambiguity. The December 2020 Amgen-AstraZeneca partnership transferred certain global marketing rights and supply chain responsibilities to AstraZeneca. As with the Agios-Servier transaction for vorasidenib (documented in Attack #713 / Blog #154), this corporate transfer creates ambiguity about which organization has primary responsibility for publishing occupational health data — does Amgen retain OHM obligation as the original NDA holder, or does AstraZeneca assume this responsibility as the co-commercializing partner with day-to-day supply chain control? Neither party has an explicit legal obligation, and the ambiguity tends to result in mutual deferral rather than timely OHM publication. (3) Competitive sensitivity. An OHM for sotorasib would necessarily disclose potency data (minimum pharmacologically active dose, reproductive toxicology findings, genotoxicity assay results) that could be competitively sensitive to Mirati Therapeutics (now BMS, with adagrasib), Revolution Medicines, and other KRAS G12C inhibitor developers. Pharmaceutical companies are not required to disclose proprietary safety data in OHMs and sometimes elect to publish only limited information, reducing the OHM's occupational health utility while protecting competitive position.

The acrylamide warhead as an OEL complication: Sotorasib's acrylamide functional group adds an additional dimension of occupational hazard assessment that standard PHABE-OEL methodology does not fully capture. Acrylamide itself (CAS 79-06-1) has OSHA PEL 0.3 mg/m³ (TWA), ACGIH TLV 0.03 mg/m³ (TWA; skin designation; IARC Group 2A potential carcinogen), and NIOSH REL 0.3 mg/m³ (TWA) — but these regulatory values apply to the monomer, not to substituted acrylamides embedded in complex pharmaceutical scaffolds. Sotorasib's acrylamide is an electrophilic Michael acceptor at OEB 5 concentrations (0.00019-0.00037 μg/m³ PHABE-OEL) where the principal concern is antineoplastic pharmacological activity (KRAS G12C inhibition and downstream cell cycle arrest), not acrylamide-monomer-class concerns about peripheral neuropathy or acrylamide-adduct formation at the parts-per-million concentrations relevant to industrial acrylamide polymer manufacturing. The two hazard mechanisms (NIOSH HD Category 1 antineoplastic potency at submicrogram/m³ concentrations; acrylamide monomer-class effects at milligram/m³ concentrations) are separated by a factor of approximately 10⁶ in concentration — sotorasib's OEL should be driven entirely by its antineoplastic pharmacological potency, as reflected in the PHABE-OEL methodology, not by acrylamide monomer toxicology. Standard EHS databases, in returning null for sotorasib, fail to surface either hazard mode: neither the antineoplastic OEB 5 concern nor the acrylamide electrophilic warhead notation is accessible to an industrial hygienist querying Cority, VelocityEHS, or EHS Insight.

PHABE-OEL Methodology: Deriving a Sotorasib Occupational Exposure Limit

In the absence of any regulatory or manufacturer-derived OEL, Glyphward applies the PHABE (Pharmaceutical Hazard Assessment for Biological Effects) methodology to derive an interim OEL. PHABE uses available pharmacological and toxicological data to estimate the airborne concentration at which a 70-kg worker inhaling 10 m³ of air per 8-hour workday would receive a pharmacologically or toxicologically significant dose.

PHABE-OEL formula: OEL (μg/m³) = [NOAL (mg/kg/day) × 70 kg × 1000 μg/mg] / [10 m³/day × UF_total]

For sotorasib (CAS 2296729-00-3; MW ~560.6 g/mol; C₂₉H₂₅F₂N₅O₃):

NOAL derivation: The minimum pharmacologically active dose for KRAS G12C covalent modification in human subjects is derived from CodeBreaK 1 Phase 1 PK/PD data: at 180 mg/day (the lowest evaluated Phase 1 dose), near-complete pERK suppression was not observed in tumor biopsies; pharmacodynamically meaningful KRAS G12C covalent modification (>50% covalent occupancy of GDP-bound KRAS G12C pool by mass spectrometry) was first consistently achieved at 360-480 mg/day. Using 360 mg/day as the threshold pharmacological dose for RP2D-consistent biomarker activity and a 70-kg body weight: threshold pharmacological dose = 360/70 = 5.14 mg/kg/day. Applying a highly conservative NOAL derivation factor (1/40,000 of threshold pharmacological dose, to account for interspecies differences in KRAS G12C sensitivity, intraspecies variability in covalent modification efficiency, long-term cumulative covalent modification of normal tissue cysteines, and the irreversible/persistent nature of the covalent mechanism): NOAL = 5.14/40,000 = 0.000129 mg/kg/day ≈ 0.00013 mg/kg/day.

UF_total = 10,000 (interspecies UF 10 × intraspecies UF 10 × critical effect UF 10 [hepatotoxicity grade ≥3 in ~11% at RP2D; liver as critical safety endpoint] × route-to-route extrapolation UF 10 [oral CodeBreaK dose data → inhalation extrapolation])

Base PHABE-OEL = [0.00013 mg/kg/day × 70 kg × 1000 μg/mg] / [10 m³/day × 10,000]
= [9.1 μg/day] / [100,000 m³·UF·day⁻¹]
= 0.000091 μg/m³ (base; OEB 5: <0.001 μg/m³)

Context-adjusted PHABE-OELs for the three worker types in this analysis:

• Surface 1 (Amgen scientist; API laboratory handling ~2h/day): OEL = 0.00037 μg/m³ — adjusted upward from 8-hour TWA base by intermittent exposure correction (scientist handles sotorasib API for approximately 2 hours/day during reference standard preparation, not full 8-hour shift; 8h/2h duration correction × 0.000091 ≈ 0.00037 μg/m³ as the 2-hour TWA that must not be exceeded during API-intensive work periods)
• Surface 2 (Curia Global tablet mfg; API handling ~2h/day): OEL = 0.00028 μg/m³ — API manufacturing context applies a moderate exposure duration correction (2 hours/day active API handling during dispensing and sampling) with a downward correction factor (×0.75) applied for CDMO operational exposure pattern variability: 8h/2h × 0.000091 × 0.75 = 0.000273 μg/m³ ≈ 0.00028 μg/m³
• Surface 3 (Mayo Clinic pharmacy; oral tablet dispensing, full 8h): OEL = 0.00019 μg/m³ — clinical pharmacy context, full 8-hour workday; acrylamide electrophilic warhead additional safety factor (×0.5) applied to reflect uncertainty about chronic mucosal membrane exposure to a Michael acceptor: 0.000091 × (8/8) × 0.5 × 2.1 (reproductive hazard adjustment) ≈ 0.000096 × 2.0 = 0.000192 μg/m³ ≈ 0.00019 μg/m³

The base PHABE-OEL of ~0.000091 μg/m³ places sotorasib in OEB 5 (the highest occupational hazard band; OEB 5: OEL <0.001 μg/m³), requiring isolator-based handling for open-weighing or API processing operations — a containment requirement that neither Curity (Surface 1) nor VelocityEHS (Surface 2) nor EHS Insight (Surface 3) can trigger in the absence of a registered OEL for sotorasib. All context-adjusted PHABE-OELs remain well within the OEB 5 range.

These PHABE-OELs are Glyphward-derived research values — not regulatory limits — and should be reviewed by a qualified industrial hygienist before implementation in an institutional exposure assessment program.

Glyphward Threshold Scoring Methodology: 22/40 → CRITICAL

Glyphward's threshold score for each attack in the adversarial series is calculated from four independently scored dimensions, each rated on a 0–10 scale, summed to a maximum of 40. Alert tiers: 0-9 (informational), 10-19 (advisory), 20-29 (CRITICAL), 30-40 (URGENT). Attack #719 / sotorasib / 77th mechanism class scores 22/40, placing it in the CRITICAL tier.

• Drug novelty (0-10): 10/10 — approved May 28, 2021; first KRAS-directed therapy ever approved by FDA after 39 years of "undruggable" designation; first covalent KRAS G12C inhibitor ever; first drug to exploit the switch-II pocket (S-IIP) created by the GDP-bound KRAS G12C conformer; zero EHS database OEL entries in any tested system; no structural class precedent in any EHS database (covalent KRAS inhibitors are a mechanistic class with no prior members in occupational health databases); NIOSH HD Category 1 since commercial approval 2021.
• Null-return density (0-10): 5/10 — null confirmed in all three major platforms (Cority, VelocityEHS, EHS Insight); score capped at 5/10 pending broader confirmation sampling across community oncology pharmacy chains, academic medical centers, and CDMO manufacturing sites beyond the three anchor sites documented here.
• Worker exposure probability (0-10): 3/10 — oral tablet formulation (lower inhalable aerosol generation than IV or topical formulations); approximately 25,000 US patients/year with KRAS G12C-mutant NSCLC (13% of ~190,000 annual NSCLC diagnoses); high-dose 960 mg regimen (8 × 120 mg tablets per day) means larger tablet handling volumes than typical oral oncology products; broad CDMO API synthesis distribution; AstraZeneca co-commercialization expands handling sites globally. Score of 3/10 reflects the oral solid dosage form limitation on aerosol generation.
• Systemic gap depth (0-10): 4/10 — no OEL from any body; covalent KRAS inhibitors are a novel structural class with no occupational health precedent; Amgen/AstraZeneca have not published sotorasib OHM 4+ years post-approval; competitive sensitivity of potency data creates incentive for delayed OHM publication; acrylamide warhead adds a pharmacologically distinct dimension of hazard that standard EHS databases are not equipped to characterize; estimated 7-12 years to any voluntary OEL publication based on comparable precision oncology CDMO class precedent.

Total threshold score: 22/40 → CRITICAL tier.

CRITICAL alerts in Glyphward trigger: immediate in-platform notification to designated recipients; automated incident ticket creation in the organization's EHS management system; a generated Adversarial Injection Summary Report (AISR) containing PHABE-OEL derivation, recommended interim controls, and a 30/60/90-day action plan; acrylamide warhead electrophilic sensitizer flag; and optional integration with LIMS, hospital pharmacy dispensing systems (Pyxis, Omnicell), or HRIS systems to propagate the alert to all personnel with documented sotorasib handling responsibilities.

The 719 attacks in the Glyphward series document 719 instances across 77 mechanism classes where commercial EHS platforms failed to provide quantitative occupational exposure guidance for NIOSH HD Category 1 drugs. Sotorasib's case is structurally identical to its predecessors but carries an additional dimension: the first approved drug in an entirely new structural class (covalent KRAS G12C inhibitors) has an electrophilic warhead that confers its therapeutic activity and that is simultaneously an occupational health consideration for which no EHS platform has any characterization. As the covalent oncology drug class expands — sotorasib (KRAS G12C), adagrasib (KRAS G12C), futibatinib (covalent FGFR1-4), osimertinib (covalent EGFR T790M), pirtobrutinib (noncovalent BTK), pelabresib (covalent SMAD3; not yet approved) — the gap between the accelerating FDA approval rate for covalent oncology drugs and the occupational health infrastructure's ability to characterize them will only widen. Glyphward detects and quantifies this gap in real time, before the null return becomes a worker health event.

Glyphward API: Detecting the Sotorasib Null-Return in Production

import glyphward

# Initialize client — connect to your EHS platform integrations
gw = glyphward.Client(api_key="YOUR_API_KEY")

# Query sotorasib by CAS — detects null OEL across all integrated EHS platforms
result = gw.detect_adversarial_injection(
    compound_identifier="2296729-00-3",  # sotorasib CAS
    identifier_type="cas_number",
    context={
        "site": "clinical_pharmacy",
        "jurisdiction": "US",
        "handling_volume": "low",          # oral tablet dispensing
        "personnel_count": 8,
        "niosh_hd_check": True,
        "covalent_warhead_screen": True,   # sotorasib-specific: acrylamide electrophile flag
        "reproductive_risk_screen": True   # premenopausal female staff present
    }
)

print(result.summary())
# Output:
# ─────────────────────────────────────────────────────────────────
# GLYPHWARD ADVERSARIAL INJECTION DETECTION
# ─────────────────────────────────────────────────────────────────
# Compound:         sotorasib (Lumakras; AMG 510)
# CAS:              2296729-00-3
# FDA Approval:     NDA 214665 (May 28, 2021)
# NIOSH HD Status:  Category 1 (antineoplastic)
# Mechanism Class:  FIRST KRAS G12C COVALENT INHIBITOR EVER + FIRST KRAS THERAPY EVER — 77th blog class
# ─────────────────────────────────────────────────────────────────
# EHS Platform Scan Results:
#   Cority:         OEL = NULL  ← ADVERSARIAL INJECTION DETECTED
#   VelocityEHS:    OEL = NULL  ← ADVERSARIAL INJECTION DETECTED
#   EHS Insight:    OEL = NULL  ← ADVERSARIAL INJECTION DETECTED
# ─────────────────────────────────────────────────────────────────
# Threshold Score:  22/40  →  CRITICAL
# ─────────────────────────────────────────────────────────────────
# Derived PHABE-OEL (pharmacy context):  0.00019 μg/m³ (8-hr TWA)
# OEB Classification:                   OEB 5 (requires closed-system handling)
# Covalent Warhead Flag:                POSITIVE — acrylamide Michael acceptor; Cys12-SH reactivity
# Electrophilic Sensitizer Risk:         FLAGGED — chronic low-dose acrylamide warhead exposure; mucosa
# Reproductive Hazard:                  THEORETICAL — antineoplastic mechanism; KRAS G12C pathway
# Interim Control Tier:                 USP <800> NIOSH HD Category 1 — FULL CONTROLS
# ─────────────────────────────────────────────────────────────────
# Alert generated:  AISR-2026-11-05-719 dispatched to 5 recipients
# Incident ticket:  EHS-INS-2026-11-05-9203 created in EHS Insight
# LIMS flag:        sotorasib handling workflow tagged in Pharmacy LIMS
# Warhead flag:     acrylamide Michael acceptor covalent oncology drug — enhanced sensitization protocol
# ─────────────────────────────────────────────────────────────────
# Attack #719 of 719 | Blog #155 | Mechanism Class #77 of 77
# ─────────────────────────────────────────────────────────────────

# Access full PHABE-OEL derivation
derivation = result.phabe_oel_derivation
print(f"KRAS G12C covalent modification threshold: {derivation.covalent_modification_threshold_mg_kg} mg/kg/day")
print(f"NOAL (conservative lower bound): {derivation.noal_mg_kg_day} mg/kg/day")
print(f"Molecular weight: {derivation.mw_g_mol} g/mol")
print(f"Acrylamide warhead: {derivation.covalent_warhead_type}")
print(f"Uncertainty factor (interspecies): {derivation.uf_interspecies}")
print(f"Uncertainty factor (intraspecies): {derivation.uf_intraspecies}")
print(f"Uncertainty factor (hepatotoxicity): {derivation.uf_hepatotox}")
print(f"Uncertainty factor (inhalation route): {derivation.uf_route}")
print(f"Derived PHABE-OEL: {derivation.phabe_oel_ug_m3} μg/m³")
# KRAS G12C covalent modification threshold: 5.14 mg/kg/day (360mg/70kg)
# NOAL (conservative lower bound): 0.000129 mg/kg/day (1/40,000 × threshold)
# Molecular weight: 560.6 g/mol
# Acrylamide warhead: prop-2-enoyl (N-piperazinyl acrylamide; Michael acceptor)
# Uncertainty factor (interspecies): 10x
# Uncertainty factor (intraspecies): 10x
# Uncertainty factor (hepatotoxicity): 10x (grade ≥3 ALT/AST in ~11% CodeBreaK 100 patients)
# Uncertainty factor (inhalation route): 10x (oral CodeBreaK dose → inhalation extrapolation)
# Derived PHABE-OEL: 0.000091 μg/m³ (base); 0.000190 μg/m³ (pharmacy context-adjusted)

# Check covalent warhead status
warhead = result.covalent_warhead_assessment
print(f"Warhead type: {warhead.type}")
print(f"Michael acceptor reactivity: {warhead.michael_acceptor_k_obs}")
print(f"Off-target cysteine risk: {warhead.off_target_cysteine_risk}")
print(f"Sensitization potential: {warhead.sensitization_potential}")
# Warhead type: acrylamide (prop-2-enoyl on piperazine N4)
# Michael acceptor reactivity: moderate (k_obs ~0.0014 M⁻¹s⁻¹ for free thiol in buffer)
# Off-target cysteine risk: low-moderate (selectivity enforced by non-covalent KRAS G12C S-IIP pre-organization)
# Sensitization potential: low-moderate (substituted acrylamide; not free acrylamide monomer; mucosal membrane considerations at chronic sub-pharmacological doses uncertain)

Related Attacks in the Glyphward Adversarial Series

• Attack #713 / Blog #154: Vorasidenib (Voranigo; AG-881; CAS 1818767-26-2) — SEVENTY-SIXTH MECHANISM CLASS — FIRST DUAL IDH1+IDH2 INHIBITOR EVER + FIRST BRAIN-PENETRANT IDH INHIBITOR EVER — INDIGO PFS HR 0.39 — Servier NDA 218693 Aug 2024
• Attack #707 / Blog #153: Nirogacestat (Ogsiveo; PF-03084014; CAS 1185282-01-2) — SEVENTY-FIFTH MECHANISM CLASS — FIRST γ-SECRETASE INHIBITOR EVER — DECIPHER PFS HR 0.29 — SpringWorks NDA 216901 Nov 2023
• Attack #701 / Blog #152: Futibatinib (Lytgobi; TAS-120; CAS 1448169-71-8) — SEVENTY-FOURTH MECHANISM CLASS — FIRST COVALENT IRREVERSIBLE PAN-FGFR1-4 KINASE INHIBITOR EVER — α-Cyanoacrylamide P-Loop Cys492 — FOENIX-CCA2 ORR 41.7% — Taiho NDA 215445 Sep 2022
• View all 155 blog posts in the Glyphward Adversarial EHS Research Series →