Sage Journals: Discover world-class research

Abstract

Objective

This study evaluated the hepatotoxicity risk of anaplastic lymphoma kinase (ALK)/ROS1- tyrosine kinase inhibitors (TKIs) in non-small cell lung cancer (NSCLC) using the Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS).

Methods

A retrospective analysis of 19,611 FAERS reports related to ALK/ROS1-TKIs (2011–2025) identified 1495 hepatotoxicity cases. Disproportionality analysis (reporting odds ratio, proportional reporting ratio, empirical Bayes geometric mean) assessed six ALK-TKIs and three ROS1-TKIs. Temporal trends, clinical outcomes, and differences across TKI generations were examined.

Results

Second-generation ALK-TKIs (e.g. alectinib) showed the most favorable hepatic safety profile, with the lowest observed mortality (3.5%) and hospitalization (13.0%). First-generation TKIs (crizotinib, ceritinib) and newer agents (lorlatinib, entrectinib) were associated with higher fatality rates (12.0%–13.2%). Most hepatotoxic events (51.9%) occurred within one month; however, 8.37% developed after one year. Mechanistically, the lower risk with second-generation TKIs may reflect less CYP-dependent metabolism and greater target selectivity.

Conclusion

Analysis of FAERS reports indicates substantial heterogeneity in hepatotoxicity risk among ALK inhibitors. Second-generation agents showed lower mortality and hospitalization than first- and third-generation counterparts. Crizotinib, ceritinib, and lorlatinib were associated with notably high fatality rates, warranting continuous hepatic monitoring. ROS1 inhibitors (e.g. entrectinib) may exhibit distinct hepatotoxicity profiles. These findings should be interpreted cautiously given inherent reporting biases. Future priorities include prospective studies, biomarker-integrated risk stratification, mechanistic investigation, and AI-enhanced monitoring to optimize the risk–benefit profile of ALK/ROS1-TKIs.

Keywords

adverse events hepatotoxicity FDA adverse event reporting system

Introduction

Lung cancer remains a major global health burden. According to the International Agency for Research on Cancer, lung cancer is the leading cause of cancer-related mortality worldwide.¹ With over two million new cases diagnosed annually, non-small cell lung cancer (NSCLC) accounts for approximately 80–85% of all lung cancer cases.¹ Major NSCLC subtypes include adenocarcinoma, squamous cell carcinoma, and large-cell neuroendocrine carcinoma. Over recent decades, NSCLC treatment has shifted from surgery, chemotherapy, and antiangiogenic therapy to immunotherapy and targeted therapies, which are now central to precision medicine. Notably, anaplastic lymphoma kinase (ALK) fusion–positive (ALK-rearranged) NSCLC represents a clinically important and therapeutically actionable molecular subtype with direct implications for treatment selection. It occurs in approximately 3–7% of NSCLC overall and 6.6–9.6% of lung adenocarcinomas.² ROS1 fusions are enriched in younger patients who are never or light smokers.³ Structurally, ROS1 shares substantial sequence similarity with ALK, with 49% identity in the kinase domain and 77% identity at the ATP-binding site. High conservation within the kinase domain and ATP-binding site may partly explain the activity of many ALK inhibitors against ROS1, including inhibition of ROS1-positive cell proliferation and clinical responses in ROS1 fusion–positive NSCLC.^4,5 To date, the U.S. Food and Drug Administration (FDA) has approved six ALK tyrosine-kinase inhibitors (ALK TKIs) for ALK-positive NSCLC and three ROS1-TKIs for ROS1-positive disease as standard therapies. Crizotinib, the first-in-class ALK kinase inhibitor, is also the inaugural first-generation TKI approved for dual targeting of both ALK and ROS1.^6–8 Its antitumor effect results from inhibition of ALK and ROS1 kinase activity, which blocks oncogenic signaling and suppresses tumor-cell proliferation. Compared to chemotherapy, targeted therapies significantly improve outcomes in advanced ALK-positive NSCLC patients, regardless of prior chemotherapy exposure.^9,10 However, the increased incidence of adverse effects—particularly severe hepatotoxicity associated with ALK-TKIs, which often leads to elevated mortality rates^11,12—and the emergence of drug resistance have driven the development of second- and third-generation TKIs. Although hepatotoxicity has been reported in clinical trials for these agents,^13–17 a comprehensive real-world comparison of liver safety profiles among NSCLC patients receiving ALK/ROS1-TKIs remains lacking. Real-world data serve as a critical complement to conventional clinical trial findings, offering insights into long-term safety and heterogeneous patient populations.

The U.S. FDA Adverse Event Reporting System (FAERS), one of the world's largest pharmacovigilance databases, enables the detection of potential risk signals through post-marketing spontaneous reports of adverse events (AEs).¹⁸ Previous studies leveraging FAERS have successfully identified TKI-associated cardiotoxicity¹⁹ and interstitial lung disease signals,^20,21 validating its utility in drug safety research. However, data mining focused on hepatotoxicity of ALK/ROS1-TKIs in NSCLC remains unexplored in FAERS. This study represents the first systematic investigation of hepatotoxicity signals for six ALK-TKIs (crizotinib, ceritinib, alectinib, brigatinib, lorlatinib, ensartinib) and three ROS1-TKIs (crizotinib, entrectinib, repotrectinib) using the FAERS database. Multiple signal detection methodologies—including the reporting odds ratio (ROR) and Bayesian confidence propagation neural network (BCPNN)—were employed to quantify risks. By analyzing temporal patterns, clinical outcomes, and inter-drug disparities of hepatotoxic events, this research aims to inform clinical decision-making, optimize liver function monitoring protocols, and stratify high-risk patients. Furthermore, it elucidates real-world characteristics and challenges associated with ALK/ROS1-TKI-induced hepatotoxicity.

Methods

Data sources

Quarterly AE reports were retrospectively retrieved from FAERS for first quarter (Q1) 2011 through Q1 2025. FAERS is a U.S. pharmacovigilance repository that documents drug-related AEs.^22,23 Reports are submitted by manufacturers, healthcare professionals, and consumers, and are supplemented by postmarketing clinical-trial data from domestic and international sources. The database encompasses multiple structured categories: demographic and administrative information (DEMO), drug-specific details (DRUG), reported adverse events (REAC), patient outcomes (OUTC), reporting sources (RPSR), drug therapy initiation/cessation dates (THER), indications (INDI), and deleted cases. The investigated agents included crizotinib, ceritinib, alectinib, brigatinib, lorlatinib, ensartinib, entrectinib, and repotrectinib. Target drugs were designated as the primary suspect (PS) in all analyses. AEs were codified using the Medical Dictionary for Regulatory Activities (MedDRA) at the Preferred Term (PT) level. This study was conducted entirely in silico (i.e. a “dry lab” computational study) using the FAERS database and standard pharmacovigilance signal-detection algorithms, consistent with best practices for computer-based analyses.²⁴

Data processing

Data packages in ASCII format downloaded from the FAERS database were imported into R (version 4.4.2) for integration, cleaning, and standardization. Duplicate reports in the DEMO table were deduplicated following FDA-recommended deletion criteria,²⁵ and the primaryid field was used to map and deduplicate records across six additional tables (e.g. DRUG). Reports with misspelled drug names, duplicates, or unrelated AEs were excluded. AEs were standardized using PTs from the MedDRA, version 27.1 and categorized by System Organ Class (SOC).²⁶ Drug names were compiled from FDA-approved listings, including generic and brand names, to query the drugname and prod_ai fields in the DRUG table. Only entries with role_cod (drug role code) designated as “PS” were retained. Clinical features of patients experiencing AEs—including age, sex, reporter type, geographic origin, and outcomes—were extracted to comprehensively characterize the safety profiles of each drug. To identify NSCLC-related reports, cases with “non-small cell lung cancer” in the FAERS “indication” field were included, while those with ambiguous or unrelated indications were excluded. For hepatotoxicity analysis, the REAC dataset was standardized prior to drug-specific queries. A predefined lexicon of liver injury PTs derived from MedDRA was applied (Table S2).²⁷ A flowchart summarizing the analytical workflow is provided in Figure S1. All data integration, deduplication, and signal computations were performed in silico using scripted pipelines, without wet-lab experimentation.

Statistical analysis

Quantitative and qualitative analyses were performed on clinical characteristics of ALK/ROS1-TKI-associated hepatotoxicity cases retrieved from the FAERS database. Signal detection was conducted using established disproportionality metrics: the proportional reporting ratio (PRR), ROR, BCPNN, and Multi-item Gamma Poisson Shrinker (MGPS).^28–32 MGPS employed an empirical Bayesian model to generate adjusted observed-to-expected ratios, reducing false-positive ROR inflation. RORs were calculated using 2 × 2 contingency tables incorporating drug-event co-occurrences, drug-only reports, event-only reports, and unrelated cases. In this study, we applied a case–noncase reference framework within FAERS to construct 2 × 2 contingency tables with cells defined as: a = target drug + hepatotoxicity terms; b = target drug + non-hepatotoxicity terms; c = other drugs + hepatotoxicity terms; and d = other drugs + non-hepatotoxicity terms. The distribution of the same event among all other drugs served as the background control. All analyses were performed after FDA-recommended de-duplication and data cleaning. In the primary analysis, we preferentially retained reports in which the drug was designated the PS and, when possible, restricted the indication to NSCLC to reduce confounding. Signal detection thresholds were defined as follows. Signal thresholds were defined as ROR [95% confidence interval (CI) lower bound >1 with ≥3 total reports] or information component (IC)-based metrics (IC > 0 and IC–2SD > 0).³³ A positive PRR required PRR ≥ 2, χ² ≥ 4, and n ≥ 3. For the BCPNN, IC_025 > 0 was considered positive; for the MGPS, a positive signal required an empirical Bayes geometric mean (EBGM) lower bound EB05 > 1. To mitigate small-sample inflation and sparse cells, we prioritized conservative lower-bound indices (IC_025 and EB05) and interpreted signals based on concordance across methods. Accuracy was evaluated in three ways: (i) agreement across algorithms; (ii) benchmarking against established hepatotoxicity evidence from prior literature and product labeling to assess face validity; and (iii) sensitivity analyses, with results compared with the primary analysis for consistency. PRR and ROR demonstrated robust performance in identifying drug-event risk associations. Analyses were performed in R (v4.4.2), with statistical significance set at P < 0.05.

Results

Descriptive analysis

Following deduplication in accordance with FDA recommendations, the FAERS database yielded 19,611 AE reports associated with ALK/ROS1 TKI therapy in NSCLC patients from Q1 2011 to Q1 2025. Of these, 1495 reports of hepatotoxicity post-ALK/ROS1-TKI treatment were identified for further analysis (Figure S1). Drug-specific hepatotoxicity signals were distributed as follows: 481 cases (7.67%) among 6270 reports listing crizotinib as the PS, 550 cases (8.66%) among 6349 alectinib reports, 228 cases (13.28%) among 1716 ceritinib reports, 132 cases (5.91%) among 2232 brigatinib reports, 77 cases (3.80%) among 2027 lorlatinib reports, and 27 cases (3.28%) among 822 entrectinib reports. No hepatotoxicity signals were detected for ensartinib or repotrectinib. Demographic characteristics of the cohort are detailed in Table 1. Among the 1495 hepatotoxicity cases, males accounted for 53.38%, females 40.27%, and patients with unknown/missing age data 6.35%. The 18–64 age group exhibited the highest incidence of ALK/ROS1-TKI-associated hepatotoxicity, representing 49.97% of reported cases. Clinicians constituted the majority of reporters, while pharmacists submitted only 6.9% of cases, highlighting potential profession-based discrepancies in severity assessment (e.g. clinicians prioritized laboratory abnormalities, while nurses focused on symptomatic manifestations). Geographically, reports originated predominantly from the United States (20.47%) and Japan (19.93%).

Table 1.

Clinical characteristics of patients with ALK/ROS1-TKI-related hepatotoxicity.

ALK/ROS1-TKIs	crizotinib	ceritinib	alectinib	brigatinib	lorlatinib	entrectinib
Total	481	228	550	132	77	27
Sex
Female	255 (53.0%)	121 (53.1%)	295 (53.6%)	81 (61.4%)	38 (49.4%)	8 (29.6%)
Male	199 (41.4%)	95 (41.7%)	207 (37.6%)	46 (34.8%)	38 (49.4%)	17 (63.0%)
Missing	27 (5.6%)	12 (5.3%)	48 (8.7%)	5 (3.8%)	1 (1.2%)	2 (7.4%)
Age, years
18∼64	264 (54.9%)	120 (52.6%)	283 (51.5%)	63 (47.7%)	52 (67.5%)	15 (55.6%)
65∼85	154 (32.0%)	64 (28.1%)	139 (25.3%)	50 (37.9%)	20 (26.0%)	7 (25.9%)
>85	7 (1.5%)	1 (0.4%)	3 (0.5%)	1 (0.8%)	1 (1.3%)	1 (3.7%)
Other ages or missing	56 (11.6%)	43 (18.9%)	125 (22.7%)	18 (13.6%)	4 (5.2%)	4 (14.8%)
Reporters
Medical Doctor	243 (50.5%)	110 (48.2%)	343 (62.4%)	74 (56.1%)	35 (45.4%)	19 (70.4%)
Clinical Nurse	111 (23.1%)	46 (20.2%)	121 (22.0%)	19 (14.4%)	22 (28.6%)	3 (11.1%)
Pharmacist	33 (6.9%)	4 (1.8%)	33 (6.0%)	12 (9.1%)	2 (2.6%)	1 (3.7%)
Other	92 (19.1%)	51 (22.4%)	52 (9.4%)	18 (13.6%)	18 (23.4%)	4 (14.8%)
Missing	2 (0.4%)	17 (7.4%)	1 (0.2%)	9 (6.8%)	—	—
Reporter Country
USA	141 (29.3%)	24 (10.5%)	97 (17.6%)	26 (19.7%)	5 (6.5%)	3 (11.1%)
Japan	85 (17.7%)	64 (28.1%)	67 (12.2%)	54 (40.9%)	12 (15.6%)	16 (59.3%)
China	57 (11.9%)	22 (9.6%)	143 (26.0%)	10 (7.6%)	15 (19.5%)	4 (14.8%)
France	42 (8.7%)	30 (13.2%)	17 (3.1%)	15 (11.4%)	2 (2.6%)	—
India	30 (6.2%)	14 (6.1%)	6 (1.1%)	—	18 (23.4%)	—
The United Kingdom of Great Britain and Northern Ireland	5 (1.0%)	3 (1.3%)	3 (0.5%)	—	1 (1.2%)	—
Canada	11 (2.3%)	2 (0.9%)	6 (1.1%)	—	2 (2.6%)	—
Other countries or missing	110 (22.9%)	69 (30.3%)	211 (38.4%)	27 (20.4%)	22 (28.6%)	4 (14.8%)

Disproportionality analysis of hepatotoxicity signals

Disproportionality analysis revealed significant hepatotoxicity signals across ALK/ROS1-TKIs. Alectinib had the highest number of hepatotoxicity reports (N = 550), while entrectinib had the fewest (N = 27). Crizotinib demonstrated a ROR of 1.55 (95% CI, 1.43–1.68) based on 481 reports, ceritinib showed an ROR of 2.19 (95% CI, 1.95–2.46) from 228 reports, and alectinib exhibited the strongest signal (ROR = 2.43, 95% CI, 2.25–2.63) with 550 reports. Disproportionality metrics for all agents—including ROR, PRR, IC, and EBGM—are comprehensively summarized in Table 2.

Table 2.

Association between ALK/ROS1-TKIs and hepatotoxicity.

Drug	crizotinib	ceritinib	alectinib	brigatinib	lorlatinib	entrectinib
N	635	319	715	170	105	31
ROR (95%Cl)	1.55 (1.43–1.68)	2.19 (1.95–2.46)	2.43 (2.25–2.63)	1.01 (0.87–1.18)	0.59 (0.49–0.72)	0.66 (0.46–0.94)
PRR (χ²)	1.53 (113.03)	2.14 (191.07)	2.36 (535.47)	1.01 (0.02)	0.6 (29.16)	0.67 (5.28)
IC (EIC-2SD)	0.59 (0.47)	1.07 (0.9)	1.18 (1.07)	0.01 (−0.21)	−0.74 (−1.02)	−0.58 (−1.08)
EBGM (EBGML-EBGMU)	1.5 (1.38–1.63)	2.1 (1.88–2.36)	2.27 (2.1–2.45)	1.01 (0.87–1.18)	0.6 (0.49–0.73)	0.67 (0.47–0.95)

95% CI: 95% confidence interval; χ²: chi-squared; IC: information component; ROR: reporting odds ratio; PRR: proportional reporting ratio; EBGM: empirical Bayesian geometric mean.

Sensitivity analysis

In sensitivity analyses excluding concomitant medications, drug-event pairs classified as positive in the primary analysis remained positive across ROR, PRR, IC, and EBGM in the PS monotherapy dataset, with concordant directionality and greater magnitudes than in analyses including co-medication (see Supplemental Table S3). Effect sizes increased for several PTs, and cross-method concordance was unchanged. These findings indicate that the detected signals are robust and not dependent on the composition of concomitant therapies, consistent with recent FAERS studies showing similarly robust results after excluding combination treatment.^34,35

Analysis of hepatotoxicity onset timing

As illustrated in Figure 1, we analyzed the temporal distribution of ALK/ROS1-TKI-associated hepatotoxicity cases. After deduplication, 765 reports with precise onset dates were included. The median time to hepatotoxicity onset was 29 days. Overall, 51.9% of cases (397/765) occurred within the first month after treatment initiation, whereas reporting frequency was lowest during months 2–3. Additionally, 8.37% of cases occurred more than one year after treatment initiation. These findings support vigilant monitoring for hepatotoxicity throughout the treatment course, extending beyond the initial months of ALK/ROS1-TKI therapy.

Figure 1.

Time to onset of ALK/ROS1-TKI-related hepatotoxicity.

Analysis of hepatotoxicity-related outcomes

As detailed in Table 3, we analyzed 1910 hepatotoxicity outcomes from non-unique cases (multiple outcomes per patient permitted). Significant heterogeneity in hepatotoxicity severity was observed across ALK/ROS1-TKIs. First-generation agents crizotinib (12.0%, 80/668) and ceritinib (13.1%, 43/329) exhibited the highest mortality rates, while second-generation inhibitors alectinib (3.5%) and brigatinib (4.0%) demonstrated the lowest lethality. Notably, lorlatinib, despite limited sample size (N = 106), showed a mortality rate of 13.2%, warranting heightened vigilance. Hospitalization rates were highest with entrectinib (35.3%), followed by brigatinib (31.8%) and lorlatinib (26.4%). Alectinib displayed optimal safety profiles, with the lowest hospitalization (13.0%) and life-threatening event rates (1.0%). Disabling outcomes were rare (<2.0% across all agents), with crizotinib (1.0%) and alectinib (0.7%) showing minimal risks. Data completeness analysis revealed substantial missing outcomes for alectinib (47.2%), potentially attributable to its recent approval and reporting delays, whereas other agents had <18% missing data (lorlatinib: 0.9%). While multiple imputation addressed missingness, alectinib's high missingness may introduce residual bias. These findings suggest that second-generation ALK inhibitors (e.g. alectinib, brigatinib) may confer advantages in reducing hepatotoxicity-related mortality and hospitalization, while first-generation agents (crizotinib, ceritinib) and select newer TKIs (lorlatinib, entrectinib) require cautious monitoring for severe hepatic outcomes. Prospective studies are warranted to validate these signals.

Table 3.

Outcomes due to ALK/ROS1-TKI-related hepatotoxicity.

Drug	crizotinib	ceritinib	alectinib	brigatinib	lorlatinib	entrectinib
N	668	329	600	173	106	34
Death	80 (12.0%)	43 (13.1%)	21 (3.5%)	7 (4.0%)	14 (13.2%)	2 (5.9%)
Disability	7 (1.0%)	6 (1.8%)	4 (0.7%)	—	—	—
Hospitalization	180 (26.9%)	90 (27.4%)	78 (13.0%)	55 (31.8%)	28 (26.4%)	12 (35.3%)
Life Threatening	32 (4.8%)	8 (2.4%)	6 (1.0%)	1 (0.6%)	1 (0.9%)	2 (5.9%)
Other	299 (44.9%)	164 (49.8%)	208 (34.7%)	102 (59.0%)	62 (58.4%)	12 (35.3%)
Missing	69 (10.3%)	18 (5.5%)	283 (47.2%)	8 (4.6%)	1 (0.9%)	6 (17.6%)

Discussion

To our knowledge, this is the first systematic evaluation of hepatotoxicity risk among ALK/ROS1-TKIs in patients with NSCLC using the FAERS database. We observed substantial heterogeneity in hepatotoxicity incidence and clinical outcomes across TKI generations. Second-generation agents (e.g. alectinib) showed a more favorable safety profile, with lower mortality (3.5%) and hospitalization (13.0%) than first-generation TKIs (crizotinib, ceritinib) and selected newer agents (lorlatinib, entrectinib). These findings are consistent with previous studies,^36,37 which attribute the improved hepatic safety of second-generation ALK-TKIs to optimized molecular design and less CYP-dependent metabolism that may reduce hepatic oxidative stress. In real-world settings, rates of severe hepatotoxic events were higher than those reported in clinical trials.¹⁵ This discrepancy likely stems from stringent trial enrollment criteria (e.g. exclusion of patients with hepatic impairment) and the broader comorbidity profiles and polypharmacy patterns observed in real-world populations. These results support the need for robust postmarketing surveillance to fully characterize drug safety, particularly for therapies with narrow therapeutic indices.

A predominance of AE reports among female patients was observed (Table S1), consistent with established sex-based differences in drug-related toxicities across therapeutic classes.^38,39 This pattern may reflect hormone-mediated modulation of drug transporters, receptors, and drug-metabolizing enzymes.

In FAERS, we observed a relatively strong hepatotoxicity signal for alectinib (ROR = 2.43), indicating the need for heightened liver-function monitoring in real-world practice. However, when clinical severity outcomes were compared (Table 3), mortality (3.5%) and hospitalization (13.0%) associated with alectinib were lower than those for crizotinib (12.0% and 26.9%, respectively) and lorlatinib (13.2% and 26.4%, respectively), suggesting a more favorable profile for severe outcomes. Furthermore, updated safety data from the ALEX trial⁴⁰ reported alanine aminotransferase (ALT) elevations in 17.1% and aspartate aminotransferase (AST) elevations in 15.8% of patients receiving alectinib, with grade 3–5 ALT and AST elevations of 4.6% and 5.3%, respectively. Although transaminase elevation was the most common AE leading to treatment discontinuation, discontinuation rates were lower with alectinib than with crizotinib (e.g. due to ALT elevation: 1.3% vs 6.0%; due to AST elevation: 1.3% vs 4.0%; due to pneumonitis/interstitial lung disease: 0.7% vs 3.3%). These observations underscore that the “strength” of an ROR signal does not correspond one-to-one with toxicity grade, clinical severity, or the risk of treatment discontinuation.

ROR reflects the degree of disproportionality in spontaneous reporting for a given drug–event pair and is influenced by factors such as the size of the exposed population, duration of therapy, intensity of laboratory surveillance, labeling or media warnings, and reporting behavior. It is not equivalent to a true incidence rate and does not discriminate toxicity grades or clinical severity. Accordingly, a high ROR identifies a signal warranting attention but cannot, by itself, adjudicate comparative clinical safety. An additional limitation is that FAERS does not include Criteria for Adverse Events (CTCAE) grading, which precludes direct comparisons of grade 3–4 severe hepatotoxicity; therefore, we used severe outcomes (death, hospitalization, life-threatening events) as proxy indicators of real-world severity.

Mechanistically, ceritinib (second-generation ALK-TKI) is primarily metabolized by CYP3A enzymes, whereas alectinib exhibits reduced CYP dependency, favoring non-enzymatic pathways such as glucuronidation. This metabolic divergence likely mitigates hepatic damage by limiting the generation of reactive metabolites. Second-generation TKIs demonstrate superior target specificity, with alectinib showing 10-fold higher selectivity for ALK over off-target kinases (e.g. ROS1, IGF-1R), thereby reducing hepatocyte stress from off-target effects. Enhanced central nervous system penetration in this class may lower systemic drug exposure, particularly hepatic concentrations.⁴¹ Emerging evidence suggests second-generation agents may attenuate oxidative stress via ferroptosis-related pathways, including NQO1-mediated reduction of lipid peroxides.^42,43 Notably, third-generation lorlatinib exhibited mortality rates (13.2%) comparable to first-generation TKIs, potentially attributable to increased systemic exposure due to enhanced blood-brain barrier penetration,^44,45 underscoring the need for dose optimization to prevent concentration-dependent hepatotoxicity. Temporal analysis aligned with TKI pharmacokinetics: 51.9% of hepatotoxicity cases occurred within the first treatment month, reflecting early metabolic adaptation phases. Notably, 8.37% of hepatotoxic events occurred after one year of treatment, challenging the conventional paradigm that TKI-induced hepatotoxicity predominantly manifests during early treatment phases.

Although ROS1 shares homology with ALK, approved ROS1 inhibitors display distinct target selectivity and pharmacokinetic properties. For instance, entrectinib not only inhibits ROS1 but also effectively suppresses TRKA/B/C,⁴⁶ whereas repotrectinib utilizes a unique macrocyclic structure and binding mode to circumvent resistance mutations.⁴⁷ These differences in target engagement and off-target effects likely contribute to divergent safety profiles (including hepatotoxicity risk) warranting prospective multicenter studies to assess their hepatic risks independently. Moreover, ROS1 inhibitors exhibit varying metabolic pathways: entrectinib is predominantly metabolized by CYP3A4,⁴⁶ unlike some ALK inhibitors (e.g. alectinib), which rely minimally on CYP-mediated metabolism. Differences in metabolic activation mechanisms and the propensity to generate reactive metabolites are key factors influencing idiosyncratic liver injury risk. Our stratified analysis reveals notable differences: compared to the limited data available for ROS1 inhibitors, ALK inhibitors—particularly first-generation agents—typically exhibit stronger hepatotoxicity signals and higher mortality rates. Although entrectinib demonstrates a lower mortality rate, its hospitalization rate is significantly elevated. While the relatively small number of reported ROS1 inhibitor cases precludes definitive conclusions regarding class-wide differences, these observed variations underscore that the safety profiles of ALK inhibitors cannot be assumed to mirror those of ROS1 inhibitors.

Through systematic comparative analysis of label warnings and clinical monitoring parameters for ALK/ROS1-TKIs, this study revealed marked heterogeneity in hepatotoxicity risk profiles across agents. Leveraging pharmacovigilance data from the FAERS, our findings provide evidence-based guidance for tailoring individualized therapeutic regimens. The study further substantiates that implementing dynamic risk assessment algorithms, combined with therapeutic drug monitoring and hepatic function stratification strategies, enables precise modulation of benefit-risk ratios in targeted cancer therapy. This biomarker-informed risk stratification framework establishes a novel paradigm for optimizing longitudinal management in advanced NSCLC patients.³⁷

This study has several limitations: First, the voluntary reporting nature of the FAERS database may introduce underreporting bias, potentially underestimating the true incidence of hepatotoxicity. Second, FAERS does not include CTCAE grading, precluding direct comparison of grade 3–4 severe hepatotoxicity. Third, high rates of missing data for certain agents—particularly ROS1-TKIs—may compromise interpretation. The markedly lower hepatotoxicity reports for ROS1-TKIs compared to ALK-TKIs (Table S2) likely reflect the rare incidence of ROS1 fusions in NSCLC (1%–2%) and limited real-world data availability due to fewer approved agents (e.g. repotrectinib) and recent market entry (data cutoff: Q1 2025). The limited sample size for ROS1-TKIs reduces statistical power in signal detection, precluding robust conclusions akin to those for ALK-TKIs. Future efforts should extend data collection periods and integrate post-marketing surveillance data for newer ROS1-TKIs to refine safety profiles. Fourth, Pooling ALK/ROS1-TKIs was intended to improve the efficiency of signal detection; it should not be interpreted as treating patients with ALK rearrangements and those with ROS1 rearrangements as a single clinical population. In particular, given the temporal shift in crizotinib's labeled use over the study period (initially for ALK and now primarily for ROS1), we analyzed signals at both the individual-drug level and the target-class level to mitigate potential confounding. Nonetheless, residual heterogeneity may persist. Prospective multicenter studies are warranted to validate these pharmacovigilance signals, coupled with mechanistic investigations of novel serological biomarkers (e.g. extracellular vesicle-derived miRNAs or ferroptosis-related proteins) for early hepatotoxicity prediction.

To address the limitations of pharmacovigilance databases and advance clinical translation, future research should focus on three key domains. First, prospective multicenter studies incorporating baseline liver function, pharmacogenomic profiling (e.g. CYP3A4/UGT1A1 polymorphisms), and dynamic multi-omics monitoring are critical to validate FAERS-derived hepatotoxicity signals while controlling for confounders like comedications and preexisting liver disease. Second, mechanistic investigations using hepatocyte organoids and animal models should elucidate TKI-specific impacts on mitochondrial dysfunction (assessed via oxygen consumption rate), ferroptosis pathways (NQO1/GPX4 activity), and oxidative stress biomarkers (e.g. lipid peroxidation products), complemented by serum/exosome-based biomarker discovery (miR-122, cytokeratin-18) for early injury detection. Third, longitudinal cohorts must refine risk stratification algorithms integrating Child-Pugh scores, early ALT/AST trends, and pharmacogenetic data to enable AI-driven surveillance systems, while PK/PD-guided therapeutic drug monitoring—particularly for CYP3A4 ultrarapid metabolizers or hypoalbuminemia patients—could optimize dosing strategies and mitigate late-onset hepatotoxicity (8.37% incidence beyond 1 year; Table S2). These interdisciplinary approaches, synergizing real-world evidence with experimental validation, will enhance precision management of TKI-associated hepatic AEs.

Conclusion

Analysis of FAERS reports revealed substantial heterogeneity in hepatotoxicity risk among ALK and ROS1 inhibitors. Second-generation ALK-TKIs showed lower observed mortality and hospitalization than first- and third-generation agents. Crizotinib, ceritinib, and lorlatinib were associated with higher mortality, supporting vigilant liver function monitoring during initiation and throughout long-term follow-up. ROS1 inhibitors—particularly entrectinib—may exhibit distinct patterns characterized by a higher hospitalization rate and warrant separate evaluation. These findings should be interpreted cautiously given reporting bias, confounding from concomitant medications, and potential underreporting, especially for ROS1 TKIs. Prospective multicenter studies are needed to validate these signals; integration of biomarker panels, mechanistic studies of metabolic pathways, and AI-enabled monitoring may improve early detection and refine dynamic risk stratification to optimize the benefit–risk profile in patients with NSCLC.

Supplemental Material

sj-docx-1-sci-10.1177_00368504251387821 - Supplemental material for Hepatotoxicity of ALK/ROS1 tyrosine kinase inhibitors in non-small cell lung cancer patients: A pharmacovigilance study based on signal mining and analysis of the FDA adverse event reporting system database

Supplemental material, sj-docx-1-sci-10.1177_00368504251387821 for Hepatotoxicity of ALK/ROS1 tyrosine kinase inhibitors in non-small cell lung cancer patients: A pharmacovigilance study based on signal mining and analysis of the FDA adverse event reporting system database by Weixing Zhao, Xiaoni Jin, Yuan Li, Huiyuan Peng and Jun Jiang in Science Progress

Footnotes

Abbreviations

Acknowledgements

The authors used an AI-based language tool (for translation and polishing) to improve the English expression of this manuscript. The authors reviewed and edited the content as needed and take full responsibility for all aspects of the work.

ORCID iDs

Weixing Zhao

Xiaoni Jin

Yuan Li

Huiyuan Peng

Jun Jiang

Ethics approval and consent to participate

Because this study was an observational study using global open database (FAERS) with anonymized information, not involving treatment intervention or collection of human samples, informed consent was exempted.

Author contributions

Weixing Zhao: writing—original draft, writing—review & editing, methodology, conceptualization, formal analysis; Xiaoni Jin: writing—original draft, methodology, data curation, visualization; Yuan Li: writing—review& editing, visualization; Huiyuan Peng: writing—review & editing, supervision; Jun Jiang: writing—review & editing, supervision, conceptualization.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Science and Technology Agency of Qinghai Province under Grant 2022-ZJ-719.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability

The datasets analyzed in the current study are derived from publicly available databases and can be accessed via their respective official websites. Further inquiries should be directed to the corresponding author.

Supplemental material

Supplemental material for this article is available online.

References

Bray

Laversanne

Sung

, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024; 74: 229–263.

Shaw

Yeap

Mino-Kenudson

, et al. Clinical features and outcome of patients with non-small-cell lung cancer who harbor EML4-ALK. J Clin Oncol 2009; 27: 4247–4253.

Drilon

Jenkins

Iyer

, et al. ROS1-dependent Cancers – biology, diagnostics and therapeutics. Nat Rev Clin Oncol 2021; 18: 35–55.

Lin

Shaw

. Recent advances in targeting ROS1 in lung cancer. J Thorac Oncol 2017; 12: 1611–1625.

Shaw

Solomon

Chiari

, et al. Lorlatinib in advanced ROS1-positive non-small-cell lung cancer: a multicentre, open-label, single-arm, phase 1-2 trial. Lancet Oncol 2019; 20: 1691–1701.

Yang

Kim

, et al. Phase II study of crizotinib in east Asian patients with ROS1-positive advanced non-small-cell lung cancer. J Clin Oncol 2018; 36: 1405–1411.

Topal

Akdag

Yildirim

, et al. Efficacy and clinical outcomes of Crizotinib in patients with ROS1-rearranged NSCLC: A multicenter study. Medicina (Kaunas) 2025; 61: 490. DOI:10.3390/medicina61030490

First-line Crizotinib versus chemotherapy in ALK-positive lung cancer. N Engl J Med 2015; 373: 1582.

Solomon

Kim

, et al. Final overall survival analysis from a study comparing first-line crizotinib versus chemotherapy in ALK-mutation-positive non-small-cell lung cancer. J Clin Oncol 2018; 36: 2251–2258.

10.

Parvaresh

Roozitalab

Golandam

, et al. Unraveling the potential of ALK-targeted therapies in non-small cell lung cancer: comprehensive insights and future directions. Biomedicines 2024; 12: 297. DOI:10.3390/biomedicines12020297

11.

Kreitman

Nair

Kothadia

. Successful treatment of crizotinib-induced fulminant liver failure: a case report and review of literature. Case Reports Hepatol 2020; 2020: 8247960.

12.

Zhou

Wang

Ying

, et al. Anaplastic lymphoma kinase tyrosine kinase inhibitor-induced hepatic failure in lung cancer patients: a study of signal mining and analysis of the FDA adverse event reporting system database. J Clin Pharm Ther 2021; 46: 1148–1154.

13.

Huber

Hansen

Paz-Ares Rodríguez

, et al. Brigatinib in crizotinib-refractory ALK+ NSCLC: 2-year follow-up on systemic and intracranial outcomes in the phase 2 ALTA trial. J Thorac Oncol 2020; 15: 404–415.

14.

Michels

Massutí

Schildhaus

, et al. Safety and efficacy of crizotinib in patients with advanced or metastatic ROS1-rearranged lung cancer (EUCROSS): a European phase II clinical trial. J Thorac Oncol 2019; 14: 1266–1276.

15.

Shaw

Gandhi

Gadgeel

, et al. Alectinib in ALK-positive, crizotinib-resistant, non-small-cell lung cancer: a single-group, multicentre, phase 2 trial. Lancet Oncol 2016; 17: 234–242.

16.

Stinchcombe

. Lorlatinib in the treatment of anaplastic lymphoma kinase-positive non-small-cell lung cancer. Ann Oncol 2021; 32: 587–589.

17.

Yang

Gong

. Lorlatinib for the treatment of anaplastic lymphoma kinase-positive non-small cell lung cancer. Expert Rev Clin Pharmacol 2019; 12: 173–178.

18.

van Puijenbroek

Bate

Leufkens

, et al. A comparison of measures of disproportionality for signal detection in spontaneous reporting systems for adverse drug reactions. Pharmacoepidemiol Drug Saf 2002; 11: 3–10.

19.

Wang

. Cardiac arrhythmias associated with anaplastic lymphoma kinase (ALK) inhibitors: an analysis of the FDA adverse event reporting system (FAERS). Expert Opin Drug Saf 2023; 22: 1127–1132.

20.

Zhao

Liu

Xie

, et al. Interstitial lung disease risk of anaplastic lymphoma kinase tyrosine kinase inhibitor treatment of non-small cell lung cancer: a real-world pharmacovigilance study. Expert Opin Drug Saf 2023; 22: 1309–1316.

21.

Dong

Deng

, et al. Interstitial lung disease associated with ALK inhibitors and risk factors: an updated comparative pharmacovigilance analysis. Front Pharmacol 2024; 15: 1361443.

22.

https://www.fda.gov/drugs/drug-approvals-and-databases/fda-adverse-event-reporting-system-faers-database (FDA adverse event reporting system (FAERS) database).

23.

FDA adverse event reporting system (FAERS): Latest quarterly data files, https://www.fda.gov/drugs/fdas-adverse-event-reporting-system-faers/fda-adverse-event-reporting-system-faers-latest-quarterly-data-files.

24.

Ranjbar

Behzadi

Mammina

. Respiratory tularemia: francisella Tularensis and microarray probe designing. Open Microbiol J 2016; 10: 176–182.

25.

Banda

Evans

Vanguri

, et al. A curated and standardized adverse drug event resource to accelerate drug safety research. Sci Data 2016; 3: 160026.

26.

Brown

. Using MedDRA: implications for risk management. Drug Saf 2004; 27: 591–602.

27.

Yang

Tan

, et al. Safety profile and hepatotoxicity of anaplastic lymphoma kinase tyrosine kinase inhibitors: a disproportionality analysis based on FDA adverse event reporting system database. Toxics 2025; 13: 210. DOI:10.3390/toxics13030210

28.

Caster

Aoki

Gattepaille

, et al. Disproportionality analysis for pharmacovigilance signal detection in small databases or subsets: recommendations for limiting false-positive associations. Drug Saf 2020; 43: 479–487.

29.

Rothman

Lanes

Sacks

. The reporting odds ratio and its advantages over the proportional reporting ratio. Pharmacoepidemiol Drug Saf 2004; 13: 519–523.

30.

Evans

Waller

Davis

. Use of proportional reporting ratios (PRRs) for signal generation from spontaneous adverse drug reaction reports. Pharmacoepidemiol Drug Saf 2001; 10: 483–486.

31.

Heo

Jung

. Extended multi-item gamma Poisson shrinker methods based on the zero-inflated Poisson model for postmarket drug safety surveillance. Stat Med 2020; 39: 4636–4650.

32.

Zhang

Song

Xia

, et al. Adverse event profile of crizotinib in real-world from the FAERS database: a 12-year pharmacovigilance study. BMC Pharmacol Toxicol 2025; 26: 61.

33.

Bate

Evans

. Quantitative signal detection using spontaneous ADR reporting. Pharmacoepidemiol Drug Saf 2009; 18: 427–436.

34.

Lou

Chen

Cui

, et al. Safety evaluation of irinotecan: a real-world disproportionality analysis using FAERS and JADER databases during the time period 2004-2024. Front Pharmacol 2025; 16: 1516449.

35.

Huang

Zhong

, et al. The real-world safety profile of pemetrexed and platinum with or without pembrolizumab: insights from a comparative analysis of FAERS database. BMC Cancer 2025; 25: 67.

36.

Peters

Mok

Gadgeel

, et al. Updated overall survival (OS) and safety data from the randomized, phase III ALEX study of alectinib (ALC) versus crizotinib (CRZ) in untreated advanced ALK+ NSCLC. Am Soc Clin Oncol, 2020; 38: 9518.

37.

Luo

Zhang

Guo

, et al. Comparative safety of anaplastic lymphoma kinase tyrosine kinase inhibitors in advanced anaplastic lymphoma kinase-mutated non-small cell lung cancer: systematic review and network meta-analysis. Lung Cancer 2023; 184: 107319.

38.

Moyer

Matey

Miller

. Individualized medicine: sex, hormones, genetics, and adverse drug reactions. Pharmacol Res Perspect 2019; 7: e00541.

39.

Mielke

Miller

. Improving clinical outcomes through attention to sex and hormones in research. Nat Rev Endocrinol 2021; 17: 625–635.

40.

Camidge

Dziadziuszko

Peters

, et al. Updated efficacy and safety data and impact of the EML4-ALK fusion variant on the efficacy of alectinib in untreated ALK-positive advanced non-small cell lung cancer in the global phase III ALEX study. J Thorac Oncol 2019; 14: 1233–1243.

41.

Mok

TSK

Crino

Felip

, et al. The accelerated path of ceritinib: translating pre-clinical development into clinical efficacy. Cancer Treat Rev 2017; 55: 181–189.

42.

Bao

Lin

, et al. ALK Fusion promotes metabolic reprogramming of cancer cells by transcriptionally upregulating PFKFB3. Oncogene 2022; 41: 4547–4559.

43.

Wang

Duan

Huang

, et al. Tanshinone functions as a coenzyme that confers gain of function of NQO1 to suppress ferroptosis. Life Sci Alliance 2023; 6: e202201667. DOI:10.26508/lsa.202201667

44.

Kim

Soo

, et al. First-line lorlatinib versus crizotinib in Asian patients with advanced ALK-positive NSCLC: Five-year outcomes from the CROWN study. J Thorac Oncol 2025; 20: 955–968. DOI:10.1016/j.jtho.2025.02.021

45.

Shaw

Bauer

de Marinis

, et al. First-Line lorlatinib or crizotinib in advanced ALK-positive lung cancer. N Engl J Med 2020; 383: 2018–2029.

46.

Doebele

Drilon

Paz-Ares

, et al. Entrectinib in patients with advanced or metastatic NTRK fusion-positive solid tumours: integrated analysis of three phase 1-2 trials. Lancet Oncol 2020; 21: 271–282.

47.

Drilon

Cho

, et al. Repotrectinib (TPX-0005) is a next-generation ROS1/TRK/ALK inhibitor that potently inhibits ROS1/TRK/ALK solvent- front mutations. Cancer Discov 2018; 8: 1227–1236.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.12 MB