Entrectinib is a ROS1-targeted tyrosine kinase inhibitor used to treat ROS1-positive non-small cell lung cancer (NSCLC). Clinical trials have demonstrated activity, including intracranial disease control, but trial populations may not reflect the full spectrum of postmarketing adverse events. To characterize real-world reporting patterns and identify reproducible safety signals, this study integrated adverse event reports from two spontaneous reporting systems: the U.S. FDA Adverse Event Reporting System (FAERS) and the Japanese Adverse Drug Event Report database (JADER).
Reports were retrieved for entrectinib (generic and brand name Rozlytrek in FAERS; Japanese name エヌトレクチニブ in JADER) with entrectinib designated as the primary suspect drug and the indication restricted to NSCLC. FAERS data covered 2020Q1–2025Q4 and yielded 520 eligible patients and 1,573 preferred-term (PT) events after FDA-recommended deduplication. JADER data covered 2020Q1–2025Q3 and produced 254 unique patients and 393 PT events after duplicate removal.
Adverse events were coded using MedDRA version 28.1 and mapped to PTs and system organ classes (SOCs). Because spontaneous reporting systems can include disease-course-related terms, PTs such as “disease progression” were retained when meeting statistical criteria but were classified as disease-course reporting rather than core toxicity signals for interpretation.
Four disproportionality methods were applied for signal detection: reporting odds ratio (ROR, primary criterion), proportional reporting ratio (PRR), Bayesian confidence propagation neural network (BCPNN), and multi-item gamma Poisson shrinker (MGPS). Subgroup analyses by age and sex were performed to explore reporting heterogeneity. Time-to-onset was evaluated in FAERS reports with valid onset dates using descriptive statistics and Weibull distribution modeling.
Across both databases, reproducible positive disproportionality signals at the SOC level included nervous system disorders, cardiac disorders, and renal and urinary disorders. Reported ROR values for these SOCs were: nervous system disorders (FAERS 4.48 / JADER 4.76), cardiac disorders (FAERS 3.20 / JADER 5.43), and renal and urinary disorders (FAERS 2.04 / JADER 3.92). These SOC-level signals indicate that reports mentioning entrectinib and these organ systems occurred more frequently than expected in both reporting systems.
At the PT level, the two-database comparison identified 23 PTs with positive signals in both FAERS and JADER. In addition, 58 PTs were unique to FAERS and 8 were unique to JADER. Representative PTs that were concordant across both databases included:
These PT-level signals span neurologic, cardiac, renal, and laboratory-related events. The exact set of PTs and the ROR estimates are reported in the study’s derived signal-detection tables.
Exploratory subgroup analyses suggested age- and sex-related reporting heterogeneity. The authors observed relatively more renal and mobility-related reports in older patients, and in FAERS there was a female predominance among reports of ataxia. These findings were described as exploratory and hypothesis-generating rather than definitive evidence of differential risk by age or sex.
In FAERS, 203 of 520 reports (39.0%) contained valid time-to-onset data and were evaluable. The median time to onset among these reports was 13 days. Seventy point nine four percent (70.94%) of evaluable reports had onset dates within 30 days of exposure. Weibull distribution modeling suggested an early-failure pattern, consistent with many events occurring soon after exposure began.
The dual-database approach identified reproducible disproportional reporting signals for entrectinib in NSCLC involving neurologic, cardiac, renal, and laboratory abnormalities across FAERS and JADER. Concordant signals in two independent spontaneous reporting systems increase the plausibility that the observed disproportionality is not solely a database-specific artifact. However, spontaneous-reporting disproportionality is susceptible to reporting biases, underreporting, and confounding by indication or comorbidity. Importantly, molecular fusion status (ROS1 confirmation) was not available for verification in the reports, which limits attribution to the intended molecular subgroup.
The authors emphasize that these findings should be treated as hypothesis-generating safety signals that may help prioritize focused monitoring—particularly early monitoring for neurologic, cardiac, and renal AEs—but do not establish incidence rates, risk magnitudes, or causality.
Entrectinib-associated reports in NSCLC showed reproducible neurologic, cardiac, renal, and laboratory-related disproportional reporting signals across FAERS and JADER. Time-to-onset patterns in FAERS suggested many events occurred early after exposure. These signals warrant consideration for early safety monitoring and further study but should not be interpreted as definitive evidence of causal risk.
Key limitations reported by the authors include the intrinsic constraints of spontaneous reporting systems: inability to calculate incidence, potential duplicate or incomplete reporting despite deduplication efforts, lack of verified molecular fusion status in case records, and the possibility that some PTs reflect disease course rather than drug toxicity. The subgroup and time-to-onset analyses were limited by missing or incomplete data in the source reports.
The study authors provide de-duplicated event-level analytic datasets and derived signal-detection tables as supporting information. Raw FAERS and JADER data are publicly available via the FDA and PMDA websites, respectively, as detailed in the source article.