Age-stratified adverse-event reporting signals of seven ALK/ROS1 tyrosine kinase inhibitors in FAERS: a disproportionality and time to onset study
Abstract
Background ALK/ROS1 tyrosine kinase inhibitors (TKIs) have improved outcomes in molecularly selected malignancies, but trial populations do not fully represent the age and clinical heterogeneity encountered in routine care. Comparative post-marketing evidence on age-related and drug-specific reporting patterns remains limited. Methods We analyzed US Food and Drug Administration Adverse Event Reporting System data from 2004 Q1 through 2026 Q1. The study included 38,479 primary-suspect reports with known age for crizotinib, entrectinib, brigatinib, lorlatinib, repotrectinib, cabozantinib, or ceritinib. Reports were stratified as 0–17, 18–44, 45–64, or ≥65 years. A preferred term (PT) was considered positive only when prespecified criteria for ROR, PRR, BCPNN, and MGPS were all met. We evaluated formal drug-exposure-by-age interactions, drug-level PT-spectrum similarity, clinical priority, and reported time to onset (TTO) with Weibull modeling. Results Positive PT counts were 64, 159, 230, and 178 across the four age groups, respectively, and 299 in the all-age analysis. Of 255 PTs eligible for formal interaction testing, 127 had an interaction false-discovery rate <0.05; 95 remained after exclusion of concepts not interpretable as adverse events. Descriptive displays highlighted pediatric skeletal concerns and neurologic or functional vulnerability in older reports, while formal contrasts distinguished these patterns from within-stratum signal detection. Drug-level positive-PT spectra showed generally low pairwise similarity, and target-set overlap correlated weakly with signal-spectrum similarity (Spearman ρ = 0.212, P = 0.356). TTO was evaluable for 15,810 reports (41.1%). Weibull estimates were below one across fitted pooled age groups, consistent with an early-failure reporting pattern. Conclusion These findings support surveillance based on both age and the prescribed agent rather than a uniform class profile. The early reporting pattern favors close assessment soon after treatment initiation. FAERS signals remain hypothesis-generating and require confirmation in longitudinal patient-level data.