From class effects to specificity FAERS evidence and network mapping of adverse events in NSCLC targeted therapy.

backgroundWhether targeted therapies for non-small cell lung cancer (NSCLC) share mechanism-driven class toxicities or mainly exhibit drug-specific risks remains unclear in real-world practice.

methodsWe analyzed reports from the U.S. FDA Adverse Event Reporting System (FAERS, 2004-2025) for 14 Food and Drug Administration (FDA)-approved agents across five classes: EGFR, ALK, ROS1, RET tyrosine kinase inhibitors (TKIs), and a KRAS G12C inhibitor. Adverse events (AEs) were standardized to MedDRA v27.0 preferred terms (PTs). Primary PT-level safety signals were defined based on concordance across four disproportionality methods: proportional reporting ratio (PRR; PRR ≥ 2, x2 ≥ 4, and ≥3 reports), reporting odds ratio (ROR; lower 95% confidence interval bound ROR025 > 1), Bayesian confidence propagation neural network (BCPNN; IC025 > 0), and the multi-item gamma Poisson shrinker (MGPS; EB05 ≥ 2). Cross-drug structure was evaluated via a Jaccard-based similarity network.

resultsAmong 34 948 individuals, per-drug signal counts ranged from 3 to 113. EGFR-TKIs were enriched for mucocutaneous and gastrointestinal events; ALK-TKIs for metabolic and laboratory abnormalities and selected cardiac findings; and RET-TKIs for hepatotoxicity and hypertension. Osimertinib showed prominent electrocardiographic signals (e.g., QT prolongation); lorlatinib exhibited a distinctive dyslipidemia signature. Brigatinib and crizotinib aligned with creatine kinase elevation and visual effects, respectively. No single PT occurred across all 14 drugs. Recurrent cross-class PTs included increased blood pressure, QT prolongation, dry skin, and edema. The similarity network revealed tight within-class modules (EGFR, ALK), a binary RET pair, and peripheral placement of repotrectinib and adagrasib, indicating limited overlap of their AE profiles.

conclusionThis first NSCLC-focused FAERS comparison integrating four-method signal detection with network analysis delineates reproducible class effects superimposed by drug-specific toxicities. Findings support tailored monitoring (e.g., dermatologic care for EGFR-TKIs; ECG/electrolytes for osimertinib; lipid/CK surveillance for ALK-TKIs; blood pressure/liver testing for RET-TKIs) to inform risk-aware first-line decisions.