Entrectinib Safety Signals in NSCLC Replicated Across FAERS and JADER, With Early Neurologic, Cardiac and Renal Events
核心洞察
A multi-source pharmacovigilance analysis of FAERS and JADER identified 23 preferred-term signals for entrectinib in NSCLC replicated across both databases, spanning neurologic, cardiac, renal and laboratory domains.
Nervous system disorders showed the strongest system-organ-class signal in both databases, with reporting odds ratios of 4.48 in FAERS and 4.76 in JADER.
Time-to-onset analysis in FAERS found a median onset of 13 days, with more than 70% of evaluable reports occurring within 30 days of treatment initiation.
A multi-source pharmacovigilance analysis of the FDA Adverse Event Reporting System (FAERS) and the Japanese Adverse Drug Event Report database (JADER) has characterized the real-world adverse event reporting profile of entrectinib in non-small cell lung cancer (搜索) (NSCLC), identifying a reproducible core of neurologic, cardiac, renal and laboratory-related signals across both spontaneous reporting systems.
The study analyzed FAERS data from 2020Q1 to 2025Q4 and JADER data from 2020Q1 to 2025Q3, retrieving reports using the generic name entrectinib and the brand name Rozlytrek in FAERS and the Japanese drug 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. After deduplication, 520 eligible patients and 1,573 preferred term (PT) events were identified in FAERS, while JADER yielded 254 unique patients and 393 PT events. All adverse events were coded using MedDRA Version 28.1.
Signal Detection Methods and Cross-Database Replication
Four established disproportionality methods were applied: reporting odds ratio (ROR), proportional reporting ratio (PRR), Bayesian confidence propagation neural network (BCPNN) and multi-item gamma Poisson shrinker (MGPS). ROR served as the primary criterion for defining positive signals, with the other three methods used to assess robustness. A PT was considered a positive signal when at least three events were reported and the lower bound of the 95% confidence interval for the ROR exceeded 1. Analyses were performed within the NSCLC reporting background rather than against the entire database.
At the system organ class (SOC) level, nervous system disorders exhibited the strongest positive signal in both databases, with an ROR of 4.48 (95% CI 3.95–5.08) in FAERS and 4.76 (95% CI 3.63–6.26) in JADER. In FAERS, additional positive SOC-level signals included cardiac disorders (ROR 3.20, 95% CI 2.70–3.80), renal and urinary disorders (ROR 2.04, 95% CI 1.66–2.52), psychiatric disorders (ROR 1.87, 95% CI 1.35–2.58), investigations (ROR 1.39, 95% CI 1.18–1.64) and eye disorders (ROR 1.55, 95% CI 1.03–2.35). JADER showed positive signals for cardiac disorders (ROR 5.43, 95% CI 4.12–7.15), renal and urinary disorders (ROR 3.92, 95% CI 2.88–5.33), investigations (ROR 2.44, 95% CI 1.84–3.24), psychiatric disorders (ROR 4.86, 95% CI 2.47–9.56) and surgical and medical procedures (ROR 8.05, 95% CI 2.47–26.22).
Twenty-three PT signals were identified in both databases, mainly involving neurologic, renal, cardiac, vascular, hepatic, metabolic and laboratory-related events. Representative shared signals included dizziness (搜索), taste disorder (搜索), renal impairment (搜索), blood creatinine increased, cardiac failure (搜索), cognitive disorder (搜索), ataxia (搜索), syncope, renal disorder, myocarditis (搜索), cardiomyopathy, electrocardiogram QT prolonged, arrhythmia, troponin I increased, left ventricular dysfunction and hyperuricaemia (搜索). Because these PTs were detected in both systems, the investigators considered them more reproducible than database-specific signals. In contrast, 58 positive PTs were unique to FAERS and 8 were unique to JADER.
Prominent PT-Level Findings
In FAERS, the most frequent and clinically notable PTs included dizziness (搜索) (n = 59; ROR 12.94, 95% CI 9.87–16.97), taste disorder (搜索) (n = 45; ROR 38.03, 95% CI 27.26–53.06), renal impairment (搜索) (n = 45; ROR 6.89, 95% CI 5.09–9.34), blood creatinine increased (n = 41; ROR 8.31, 95% CI 6.04–11.43), cardiac failure (搜索) (n = 30; ROR 7.57, 95% CI 5.22–10.97) and cognitive disorder (搜索) (n = 25; ROR 18.04, 95% CI 11.86–27.45). Additional notable signals included ataxia (搜索) (n = 19; ROR 51.73, 95% CI 30.47–87.81), cardiomyopathy (n = 12; ROR 13.27, 95% CI 7.32–24.04) and myocarditis (搜索) (n = 12; ROR 3.47, 95% CI 1.95–6.17).
In JADER, leading PT signals included renal impairment (搜索) (n = 29; ROR 8.18, 95% CI 5.53–12.10), cardiac failure (搜索) (n = 17; ROR 8.90, 95% CI 5.36–14.75), renal disorder (n = 16; ROR 21.60, 95% CI 12.48–37.39), cognitive disorder (搜索) (n = 16; ROR 78.29, 95% CI 40.54–151.22), syncope (n = 14; ROR 89.45, 95% CI 43.35–184.58) and blood creatinine increased (n = 13; ROR 24.99, 95% CI 13.51–46.21). Additional signals included electrocardiogram QT prolonged (n = 9; ROR 5.35, 95% CI 2.72–10.54), myocarditis (搜索) (n = 9; ROR 5.91, 95% CI 3.00–11.67) and neutropenia (n = 8; ROR 2.53, 95% CI 1.24–5.14).
Several low-frequency PTs showed disproportionately strong signals with wide confidence intervals, including dyslalia (n = 4; ROR 91.71, 95% CI 25.85–325.29), ligament rupture (n = 3; ROR 206.21, 95% CI 34.43–1234.98) and orthostatic hypotension (n = 6; ROR 15.89, 95% CI 6.81–37.05) in FAERS, and ataxia (搜索) (n = 7; ROR 351.45, 95% CI 72.78–1697.24), cardiac disorder (n = 7; ROR 175.72, 95% CI 51.23–602.72) and left ventricular dysfunction (n = 3; ROR 99.38, 95% CI 20.00–493.94) in JADER. The authors retained these in supplementary tables for transparency but cautioned that they are statistically unstable and should be interpreted as exploratory rather than confirmed adverse drug reactions.
Time-to-Onset and Reporting Patterns
Time-to-onset analysis, restricted to FAERS because of insufficient date-paired JADER reports, included 203 of 520 eligible reports (39.0%) with complete and valid onset-time data. The analysis suggested early clustering of reported events, with a median onset of 13 days and more than 70% of evaluable reports having an onset date within the first 30 days after treatment initiation. A Weibull distribution model was consistent with an early-failure reporting pattern.
Annual reporting trends showed an early postmarketing increase followed by fluctuations. In FAERS, eligible reports rose from 19 in 2020 to 58 in 2021 and 129 in 2022, peaked at 161 in 2024 and declined to 52 in 2025. In JADER, reports numbered 29 in 2020, 72 in 2021, 84 in 2022, 23 in 2023, 34 in 2024 and 12 in 2025Q3. The investigators emphasized that these trends describe reporting activity rather than changes in the true incidence of adverse events.
Physicians were the predominant reporters in both databases, contributing 71.5% of FAERS reports and 79.9% of JADER reports. Female patients accounted for 48.7% of the FAERS cohort and 57.9% of the JADER cohort. Body weight data were frequently missing, recorded as absent in 68.3% of FAERS patients and 46.9% of JADER patients.
Subgroup Findings and Mechanistic Context
Subgroup analyses by age and sex were considered exploratory. In FAERS, signals relatively enriched in younger patients (<65 years) included disease progression (21 vs. 12 cases; ROR 2.27, 95% CI 1.10–4.70), hepatic function abnormal (11 vs. 3 cases; ROR 4.71, 95% CI 1.30–17.04) and metastases to central nervous system (8 vs. 2 cases; ROR 5.09, 95% CI 1.07–24.18). Renal impairment (搜索) (11 vs. 29 cases; ROR 0.45, 95% CI 0.22–0.92), gait disturbance and renal disorder were numerically more frequent in older patients. In sex-stratified analysis, ataxia (搜索) showed the most pronounced female predominance (16 vs. 2 cases; ROR 5.92, 95% CI 1.35–25.97), whereas blood creatinine increased (14 vs. 20 cases; ROR 0.48, 95% CI 0.24–0.97) and pleural effusion (5 vs. 11 cases; ROR 0.32, 95% CI 0.11–0.92) showed male-predominant patterns. Age-stratified subgroup analysis was not conducted for JADER because the publicly available dataset provides only broad age categories.
The authors noted that entrectinib inhibits TRK (搜索), ROS1 (搜索) and ALK (搜索) and has intracranial antitumor activity, providing a biologically plausible context for repeated reporting of neurologic events such as dizziness (搜索), cognitive disorder (搜索), ataxia (搜索), gait disturbance, syncope and taste disorder (搜索). However, the study cannot determine whether these reports were caused by direct central nervous system pharmacologic effects, brain metastases, prior cranial irradiation, concomitant medications or other patient-related factors. Similarly, replicated cardiac signals may reflect kinase-inhibition-related effects on cardiomyocyte signaling, electrophysiologic vulnerability, myocardial stress in advanced cancer, pre-existing cardiovascular disease, prior anticancer therapy or concomitant QT-prolonging medications. Renal and metabolic signals may arise through direct or indirect renal effects, dehydration, hemodynamic instability, concomitant nephrotoxic drugs, tumor-related deterioration or tumor lysis-related metabolic disturbance.
Disease progression met the statistical signal criterion in FAERS but was categorized as a disease-course-related reporting term rather than part of the core toxicity profile, as it may reflect treatment failure, tumor aggressiveness, acquired resistance or the underlying disease course.
Limitations and Clinical Interpretation
The investigators outlined several limitations. FAERS and JADER are spontaneous reporting systems subject to underreporting, stimulated reporting, incomplete documentation, heterogeneous report quality, delayed submissions and reporting bias. Because these databases lack reliable exposure denominators, they cannot be used to estimate incidence rates, absolute risks or comparative safety profiles relative to alternative therapies. Although analyses were conducted within an NSCLC reporting background, residual confounding could not be fully excluded; ROS1 (搜索) fusion status could not be verified, and the NSCLC reference population could not be considered molecularly comparable to the entrectinib group.
The analysis involved a large number of PT-level comparisons across multiple methods without formal multiplicity adjustment, and a single case report could contribute multiple PT events, meaning event-level observations were not fully independent. The time-to-onset analysis included only 203 of 520 eligible FAERS reports, and missing onset dates may have introduced selection bias.
The authors concluded that entrectinib-associated adverse event reporting in NSCLC is characterized by a reproducible core of neurologic, cardiac, renal and laboratory-related signals across FAERS and JADER, alongside additional database-specific and low-frequency exploratory signals. They stated that the clinical value of these findings lies in prioritizing areas for awareness and monitoring rather than establishing causality, and that future cohort-based studies, registry analyses and mechanistic investigations with detailed clinical covariates will be needed to determine which signals represent true drug-related toxicities and how they should be incorporated into risk-adapted monitoring strategies.