Safety recommendations for ALK tyrosine kinase inhibitors in non-small cell lung cancer: evidence from FAERS and CVARDD real-world databases.

1
Introduction
Non–small cell lung cancer (NSCLC) is one of the leading causes of cancer-related morbidity and mortality worldwide. Over the past two decades, substantial progress has been made in its treatment. The introduction of small-molecule tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors has provided unprecedented survival benefits for selected patients. Nevertheless, the overall cure and survival rates of NSCLC remain unsatisfactory, particularly in metastatic disease (1, 2). Among the genetic drivers, rearrangements of the anaplastic lymphoma kinase (ALK) gene represent an important oncogenic alteration, accounting for approximately 3%–7% of all lung adenocarcinoma cases (3). Since the advent of ALK tyrosine kinase inhibitors (ALK-TKIs), both progression-free survival and overall survival have significantly improved, and ALK-TKIs have become the standard first-line therapy for ALK-positive NSCLC (4, 5). However, with longer treatment durations and broader real-world use, drug-related adverse drug events (ADEs) have increasingly attracted attention, highlighting the need for a deeper understanding of safety profiles and differential risks across commonly used ALK-TKIs.
In this study, we focused on crizotinib, alectinib, and brigatinib because they represent widely used agents spanning different generations of ALK inhibition and distinct therapeutic positioning in clinical practice (6). Crizotinib, as an earlier ALK inhibitor, historically served as a key first-line option and remains clinically relevant in certain settings; alectinib is a commonly used next-generation agent with established first-line use and improved central nervous system (CNS) disease control; brigatinib is also a next-generation inhibitor used for advanced ALK-positive NSCLC with activity against CNS metastases and resistance profiles (4–6). Importantly, these drugs differ in target spectrum, exposure profiles, and tissue distribution, which may translate into heterogeneous toxicity patterns. Prior studies and post-marketing evidence suggest that crizotinib is frequently associated with visual disturbances and hepatotoxicity, alectinib with constipation, weight gain, and edema, and brigatinib with hypertension, myalgia, and pulmonary toxicities (7–9). Because these agents are commonly considered as alternatives or sequential options in the treatment pathway, clarifying their comparative safety characteristics is clinically meaningful for individualized drug selection and monitoring strategies.
Beyond pharmacologic and structural differences among ALK-TKIs, patient-specific factors may also shape ADE susceptibility. Sex differences are increasingly recognized as determinants of drug response and toxicity, potentially mediated by differences in body composition, hormonal milieu, and pharmacokinetic processes such as absorption, distribution, metabolism, and elimination (4, 10). For example, sex-related variability in hepatic enzyme activity, transporter expression, and cardiovascular or metabolic vulnerability may contribute to differential toxicity manifestations across males and females (11, 12). However, existing literature has often emphasized efficacy endpoints or single-agent safety descriptions, whereas comparative, sex-stratified safety evaluations across multiple ALK-TKIs remain limited. This knowledge gap restricts evidence-based guidance for sex-informed risk assessment and targeted monitoring in routine practice.
Against this backdrop, real-world data (RWD) complement clinical trials by capturing broader populations, longer exposure periods, and rarer events. The U.S. Food and Drug Administration’s Adverse Event Reporting System (FAERS) and the Canadian Vigilance Adverse Reaction Database (CVARDD) collect large volumes of spontaneous adverse event reports from distinct healthcare systems and populations, enabling cross-database assessment of signal robustness (13, 14). Therefore, using FAERS and CVARDD, we systematically evaluated and compared the safety signal characteristics of crizotinib, alectinib, and brigatinib, and further examined sex differences, signal strength distributions, and time-to-onset patterns. By placing these agents within a comparative context aligned with their clinical use, our findings aim to clarify drug- and sex-specific safety discrepancies and provide practical evidence to support individualized risk management in ALK-positive NSCLC.

2
Materials and methods
2.1
Data source
The adverse drug event (ADE) data analyzed in this study were obtained from the U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) database, which has been publicly accessible since 2004. FAERS collects spontaneous adverse event reports from diverse sources, including healthcare professionals, pharmaceutical manufacturers, and patients. To investigate ADEs associated with the ALK tyrosine kinase inhibitors (ALK-TKIs) crizotinib, alectinib, and brigatinib, all relevant reports were extracted from FAERS from its inception through the second quarter (Q2) of 2025. The data were subsequently imported into MySQL 15.0 and processed using Navicat Premium 15 software to facilitate comprehensive data management and analysis.

2.2
Data extraction and processing
In this study, crizotinib, alectinib, and brigatinib were identified as suspected drugs, and their names were standardized using the Medex_UIMA_1.8.3 encoding system. The extracted FAERS data were preprocessed using the Statistical Analysis System (SAS) and MySQL to ensure data completeness and integrity. Duplicate case reports with identical case identifiers in the DEMO table were removed. To address the inherent issue of duplicate reporting in spontaneous pharmacovigilance databases, a rigorous deduplication process was applied. In addition to removing reports with identical case identifiers in the DEMO table, we also cross-referenced reporting dates, patient demographics (age, sex), and reporting countries to identify and exclude overlapping or updated reports for the same patient event, retaining only the most recent and complete version.
The most recent version of the Medical Dictionary for Regulatory Activities (MedDRA version 27.1) was used to map the preferred terms (PTs) and corresponding system organ classes (SOCs) for ADEs associated with the three ALK-TKIs. Clinical characteristics of patients reporting ADEs related to these drugs were collected, including age, sex, weight, reporting country, indication, and reporting year.
The overall workflow for data extraction, preprocessing, and analysis of ALK-TKI–related adverse events is illustrated in Figure 1.

2.3
Data mining algorithms
To investigate the potential associations between ALK-TKI drugs (Crizotinib, Alectinib, Brigatinib) and Adverse Drug Events (ADEs), an disproportionality analysis was conducted. This analysis is considered a key tool in pharmacovigilance and aims to assess the relationship between drugs and ADEs by comparing observed frequency ratios in the exposed and unexposed populations using 2 × 2 contingency tables. This method is a cornerstone of pharmacovigilance, assessing the strength of the statistical association between a drug and an adverse event by comparing the observed frequency ratios in the exposed versus the unexposed populations. The study design and reporting of signal detection followed the Adverse Event Proportion Analysis Reporting (READUS-PV) statement for pharmacovigilance signal detection (15). Additionally, when selecting detection algorithms, we recognized the considerations and inherent limitations when using spontaneous reporting systems as a source of clinical data (16).
Sex differences analysis
To formally compare sex differences beyond descriptive counts, we conducted a between-sex comparative disproportionality analysis using the relative reporting odds ratio (rROR). For each drug–event pair (PT), we constructed a sex-stratified 2×2 table within reports of the target drug, where and denote the numbers of male reports with the PT and with other PTs, and and denote the corresponding counts in female reports. The sex contrast was quantified as:
Confidence intervals and hypothesis testing were performed on the log scale using the standard large-sample approximation for ln(rROR m/f), yielding 95% CIs and two-sided p-values. Multiple comparisons across PTs were adjusted using the Benjamini–Hochberg false discovery rate (BH-FDR).
An >1 indicates stronger reporting disproportionality in males, whereas <1 indicates stronger disproportionality in females.
In this study, four established disproportionality algorithms were applied for signal detection: the Reporting Odds Ratio (ROR) (17), Proportional Reporting Ratio (PRR) (18), Bayesian Confidence Propagation Neural Network (BCPNN) (19), and Multi-Item Gamma Poisson Shrinker (MGPS) method (20). Each algorithm offers distinct advantages: ROR helps correct biases related to low event counts; PRR provides higher specificity compared with ROR; BCPNN excels at integrating multisource data and performing cross-validation; and MGPS is particularly effective in identifying signals from rare events.
The mathematical formulations and significance thresholds for these four algorithms are summarized in Supplementary Table 1. All statistical analyses were performed using R software. Higher signal values indicate stronger associations between the target drug and specific ADEs, suggesting a greater likelihood of a causal relationship.

3
Results
3.1
Demographic information of adverse events for ALK-TKI drugs
As of the second quarter of 2025, the FAERS database received a total of 9,030 adverse event reports for Crizotinib, 8,486 for Alectinib, and 1,167 for Brigatinib.
For Crizotinib, the baseline data shows that the number of female patients slightly exceeds the number of male patients (48.9% vs. 39.9%). Among reports with clear age data, the most common age group was 18–65 years (40.4%). The majority of patients had a weight range between 50–100 kg (20.1%), though weight information was missing for 74.2% of the population. The predominant indication was LUNG NEOPLASM MALIGNANT (29.8%). The majority of reports came from the United States (44.8%), followed by India (11.6%) and Japan (7.8%).
For Alectinib, female patients were more prominent than male patients (54.6% vs. 36.5%). Among reports with clear age data, the most common age group was also 18–65 years (34.3%). The weight range for most patients was 50–100 kg (19.8%), with 76.8% missing weight data. The predominant indication was LUNG NEOPLASM MALIGNANT (31.9%). The majority of reports came from the United States (51.0%), followed by Japan (8.2%) and China (7.1%).
For Brigatinib, female patients slightly outnumbered male patients (49.4% vs. 40.7%). The most common age group among reports with clear age data was 18–65 years (24.0%). The majority of patients had a weight range between 50–100 kg (12.7%), with weight information missing for 84.7% of the population. The predominant indication was LUNG NEOPLASM MALIGNANT (30.4%). The majority of reports came from Japan (47.9%), followed by the United States (31.1%) and France (3.8%). Detailed information is provided in Table 1.
Figure 2A shows that since 2011, Crizotinib, the first approved ALK-TKI, reached a preliminary peak of adverse event reports in 2015 (approximately 1,109 cases), followed by a gradual decline. Since 2016, reports for Alectinib have risen rapidly, surpassing Crizotinib in 2020 to become the leading contributor, peaking in 2024 with about 2,953 cases. Brigatinib began receiving reports in 2017, with relatively few reports overall, but a consistent increase year by year. Overall (ALL), the total number of adverse event reports peaked in 2024 (3,778 cases), indicating a significant increase in safety reports for ALK-TKI drugs as clinical applications expanded and their market presence grew.
Further polynomial curve fitting analysis of the total reports for the three drugs was conducted, as shown in Figure 2B. The adverse event reporting frequency for ALK-TKI drugs has shown a rapid growth trend. The coefficient of determination (R² = 0.8137) indicates that the model explains 81.37% of the data variability, which demonstrates a high level of reference value for this trend.

3.2
SOC-level signal analysis
Figure 3A presents the System Organ Class (SOC)–level risk signals associated with ALK-TKIs. Among all reports, the most frequent SOC for crizotinib (N = 5,427) and alectinib (N = 4,050) was General disorders and administration site conditions, while Neoplasms benign, malignant and unspecified (including cysts and polyps) represented the predominant SOC for brigatinib (N = 528).
When restricting the analysis to cases with lung cancer as the reported indication (Figure 3B), the same pattern was observed: General disorders and administration site conditions remained the most frequent SOC for crizotinib (N = 1,484) and alectinib (N = 1,199), whereas Neoplasms benign, malignant and unspecified (including cysts and polyps) remained dominant for brigatinib (N = 230).
A further assessment of SOC-level disproportionality signals is illustrated in the forest plots in Figure 4.
In the overall population (Figure 4A), the strongest risk signals were as follows: Crizotinib: Neoplasms benign, malignant and unspecified (ROR = 3.33, 95% CI = 3.18–3.48), Alectinib: Hepatobiliary disorders (ROR = 2.67, 95% CI = 2.43–2.93), Brigatinib: Neoplasms benign, malignant and unspecified (ROR = 8.22, 95% CI = 7.48–9.03), In the lung cancer subgroup (Figure 4B), the top risk signals were: Crizotinib: Eye disorders (ROR = 2.58, 95% CI = 2.24–2.97) Alectinib: Pregnancy, puerperium and perinatal conditions (ROR = 5.36, 95% CI = 1.91–15.06), Brigatinib: Neoplasms benign, malignant and unspecified (ROR = 3.94, 95% CI = 3.40–4.58).

3.3
PT-level signal analysis
Figure 5 presents the forest plots of preferred term (PT)–level risk signals for the three ALK-TKIs. For crizotinib, the highest signals were observed for peripheral oedema (N = 525, ROR = 5.38, 95% CI = 4.77–6.07), vomiting (N = 482, ROR = 2.61, 95% CI = 2.38–2.85), and visual impairment (N = 294, ROR = 5.82, 95% CI = 5.19–6.53). For alectinib, strong signals were detected for fatigue (N = 579, ROR = 2.50, 95% CI = 2.30–2.72), constipation (N = 499, ROR = 7.97, 95% CI = 7.29–8.71), and oedema (N = 415, ROR = 9.47, 95% CI = 8.09–11.09). For brigatinib, similar risk signals were observed, with fatigue, constipation, and oedema showing relatively high ROR values.
To further compare overlapping adverse events across the three ALK-TKIs, six common PT-level risk signals were visualized in a heatmap (Figure 6) according to their ROR values. Overall, the three drugs shared partially overlapping ADE profiles, but the strength of signal varied among them, particularly for hepatic and pulmonary events, which consistently showed higher ROR values. Specifically, brigatinib demonstrated the strongest signals for hepatic function abnormal (ROR = 13.95), pleural effusion (ROR = 11.03), and interstitial lung disease (ROR = 9.36), indicating a higher risk of hepatotoxicity and pulmonary toxicity. Alectinib showed pronounced signals for oedema (ROR = 9.47) and hepatic function abnormal (ROR = 9.21), suggesting a tendency toward peripheral edema and hepatobiliary adverse effects. In contrast, crizotinib displayed a more balanced distribution of signals, with the highest RORs observed for pleural effusion (ROR = 8.88) and hepatic function abnormal (ROR = 7.92).
A between-sex comparative disproportionality analysis was subsequently conducted using the relative reporting odds ratio (rROR) (Table 2). For crizotinib, several PTs retained statistically robust sex contrasts after BH-FDR adjustment: female-enriched reporting was observed for nausea (rROR=0.623, 95% CI 0.528–0.734, p_FDR=3.49×10−7), vomiting (0.575, 0.471–0.702, p_FDR=4.96×10−7), malaise (0.673, 0.509–0.890, p_FDR=0.0277), and visual impairment (0.717, 0.555–0.925, p_FDR=0.0425). In contrast, renal impairment showed a male-enriched signal (1.905, 1.338–2.714, p_FDR=0.00235). Other PTs did not remain significant after multiplicity control.
For alectinib, rROR analysis indicated male-enriched reporting for disease progression (1.568, 1.239–1.986, p_FDR=0.00371), drug ineffective (1.528, 1.160–2.013, p_FDR=0.0129), and blood creatinine increased (1.839, 1.243–2.721, p_FDR=0.0129), whereas female-enriched reporting was identified for weight increased (0.592, 0.436–0.805, p_FDR=0.00822), constipation (0.773, 0.635–0.940, p_FDR=0.0336), and anaemia (0.676, 0.503–0.907, p_FDR=0.0336). Signals such as pneumonia and asthenia showed borderline significance after correction (both p_FDR≈0.0508).
For brigatinib, no PT-level sex contrast remained statistically significant following BH-FDR adjustment (all p_FDR≥0.607). Although nausea tended to be female-enriched (rROR=0.594, 0.348–1.016, p=0.0572) and metastases to liver tended to be male-enriched (2.416, 0.961–6.075, p=0.0607), these differences did not withstand multiplicity correction, likely reflecting limited statistical power.
Collectively, the between-sex comparative results demonstrate drug-specific sex contrasts in reporting disproportionality, with the most consistent differences observed for crizotinib and alectinib, whereas brigatinib exhibited no robust sex-associated variation in the PTs analyzed.

3.4
External validation using the Canadian CVARDD database
To further ensure the accuracy and robustness of the findings, an external validation analysis was conducted using adverse event data from the Canadian Vigilance Adverse Reaction Database (CVARDD).
Figure 7A compares the preferred term (PT)–level ROR signals of crizotinib between the FAERS and CVARDD databases to assess the external consistency of signal detection. Overall, most adverse events exhibited similar signal trends across both datasets, indicating good reproducibility of the results. Notably, oesophageal varices (ROR = 429.45) and hypertransaminasaemia (ROR = 136.98) showed substantially stronger signals in CVARDD than in FAERS, suggesting that despite differences in population composition and reporting structure between the two databases, the direction of risk association remained consistent. In contrast, common adverse events such as diarrhoea and anaemia displayed moderate and stable signal intensities in both datasets, reflecting the reliability of these associations. On a logarithmic scale, the signal distributions and color gradients showed a strong positive correlation between FAERS and CVARDD, further confirming the cross-database consistency and external robustness of crizotinib-related risk signals.
Figure 7B presents the comparison of PT-level ROR signals for alectinib across the two databases. In CVARDD, oesophageal varices (ROR = 921.40) and haemolytic anaemia (ROR = 279.11) exhibited markedly higher signal intensities than those observed in FAERS, implying potential amplification effects in real-world clinical data while maintaining consistent directional associations. Meanwhile, oedema and disease progression demonstrated moderate and directionally concordant signals in both databases, reinforcing the stability and reliability of alectinib-related adverse events. The overall positive correlation between signal distributions across databases further supports the robustness and external validity of alectinib-associated safety signals.
As shown in Figures 7C, the PT-level signal analysis for brigatinib revealed that oesophageal varices exhibited an exceptionally strong signal in the CVARDD database (ROR = 3,042.52), significantly higher than its corresponding FAERS value (ROR = 7.01). This finding indicates a pronounced signal amplification effect in real-world clinical settings, although the association direction remained consistent. Additionally, drug resistance (ROR = 80.72) and malignant neoplasm progression (ROR = 12.74) both showed concordant positive signals across the two databases, further corroborating the reliability of these associations.
Taken together, all three ALK-TKIs—crizotinib, alectinib, and brigatinib—demonstrated consistent cross-database risk signal trends between FAERS and CVARDD, underscoring the high external robustness and reproducibility of this pharmacovigilance analysis.

3.5
Time-to-onset analysis of safety signals
In this study, the temporal characteristics of adverse events (AEs) were evaluated using median time-to-onset (TTO) and Weibull distribution modeling. The Weibull model provides two key parameters—shape (β) and scale (α)—to describe the temporal risk pattern of AEs. When the shape parameter β< 1 and its 95% CI< 1, the incidence of AEs is considered to decrease over time, representing an “early failure” pattern. When β ≈ 1 and the 95% CI includes 1, AEs occur randomly over time, indicating a “random failure” pattern.
As summarized in Table 3, a total of 2,348 valid reports for crizotinib, 2,211 for alectinib, and 424 for brigatinib included time-to-onset information. The median TTOs were 51 days, 120 days, and 35 days, respectively. All three drugs exhibited an early failure pattern (β< 1), suggesting that the risk of ADEs was highest during the initial treatment phase and declined thereafter. The corresponding Weibull distribution curves for each drug are shown in Figures 8B–D.
A violin plot of TTO distributions across the three drugs (Figure 8A) further illustrates inter-drug variability. Crizotinib displayed the widest temporal dispersion and the broadest range of AE onset times, reflecting greater interindividual differences in its safety profile.
We further performed TTO analyses for two clinically important high-signal AEs—hepatic function abnormal and pleural effusion—as detailed in Table 3. For hepatic function abnormal, brigatinib showed the shortest median TTO (19 days), with all three drugs following early failure–type Weibull curves, indicating that liver injury tends to occur early in treatment. For pleural effusion, crizotinib demonstrated the shortest median TTO (34.5 days). Both crizotinib and alectinib exhibited early failure patterns, whereas brigatinib displayed a random failure–type curve, suggesting continuous risk of pleural effusion over time.

4
Discussion
This study conducted a systematic comparison of adverse drug event (ADE) signals associated with three representative anaplastic lymphoma kinase tyrosine kinase inhibitors (ALK-TKIs)—crizotinib, alectinib, and brigatinib—using two large global real-world pharmacovigilance databases: FAERS and CVARDD. The analysis focused on differences in adverse event profiles, sex-related variations, and time-to-onset characteristics (21). The results revealed significant heterogeneity in the safety profiles of these ALK-TKIs, which appear to be influenced by drug generation, structural characteristics, metabolic pathways, and individual patient factors. The findings provide real-world evidence to support clinical medication management and individualized risk control in patients with ALK-positive non–small cell lung cancer (NSCLC).
4.1
Adverse event characteristics and population differences
The results of the study indicate that all three ALK-TKIs primarily present with systemic disorders and administration site reactions at the System Organ Class (SOC) level. At the specific adverse event level, liver function abnormalities, edema, fatigue, pleural effusion, and visual disturbances are the most common types of adverse reactions. There are significant differences in the adverse event profiles of the different drugs: Brigatinib exhibits the strongest signals for liver injury and pulmonary toxicity (liver dysfunction ROR = 13.95, interstitial lung disease ROR = 9.36), suggesting that this drug should be used with caution in patients with impaired liver function or underlying pulmonary diseases (12, 22, 23); Alectinib’s major adverse events are concentrated in metabolic and fluid retention reactions (e.g., edema ROR = 9.47), which could exacerbate circulatory burden in patients with heart failure, renal insufficiency, or prolonged sedentary behavior (24, 25); Crizotinib’s adverse event distribution is relatively balanced, but the risks of visual disturbances and pleural effusion are higher (visual disturbances ROR = 5.82, pleural effusion ROR = 8.88), so monitoring should be strengthened for patients who drive or perform fine tasks, as well as for those with pleural diseases.
In recent years, according to a national analysis from Greece, the usage rate of Crizotinib has significantly decreased, while Alectinib’s usage has notably increased. From 2020 to 2022, the usage of Alectinib rose from 16% to 59%, reflecting the widespread adoption of next-generation ALK inhibitors in clinical practice (26). This trend indicates that Alectinib, along with other new-generation ALK-TKIs, is gradually establishing its dominant position in treatment, especially in patients who require better control of brain metastases and superior resistance, gradually replacing Crizotinib as the standard treatment option.
Age distribution results show that patients aged 18–65 years are the main affected group, often in long-term treatment stages, where prolonged drug exposure increases the likelihood of cumulative toxicity (27, 28). Elderly patients, due to reduced liver and kidney metabolism capacity, should have their dosage appropriately reduced at the beginning of treatment, with shorter monitoring intervals.
Geographic distribution analysis shows that the United States and Japan have the highest number of reports, suggesting potential specific metabolic differences in East Asian populations. Previous studies have shown that the CYP3A5 genotype distribution in Asian patients differs from that in Western populations, potentially leading to higher plasma concentrations of Alectinib and Brigatinib, which increases the risk of hepatic and metabolic adverse reactions. Therefore, it is recommended to adopt a conservative dose escalation strategy in Asian patients, with adjustments to the dosing regimen based on plasma drug concentration monitoring (29).
In summary, for patients with liver dysfunction or chronic liver disease, Alectinib can be considered the first-choice drug, while Brigatinib should be used with caution. For patients with heart failure or renal impairment (30), Crizotinib offers relatively better safety. For younger patients requiring long-term maintenance treatment, Alectinib shows more significant overall tolerability and quality of life improvement advantages (31).

4.2
Sex differences and drug-specific risk profiles
Using the between-sex comparative disproportionality approach (rROR with BH-FDR adjustment), we observed drug-specific and PT-specific sex contrasts rather than uniform sex effects across all ALK-TKIs. In the crizotinib group, female-enriched reporting was identified for nausea, vomiting, malaise, and visual impairment (all rROR<1; BH-FDR<0.05), whereas renal impairment showed a male-enriched signal (rROR>1; BH-FDR<0.01). These findings suggest that sex may modulate the clinical expression of crizotinib-related toxicities, with males showing a stronger propensity for renal-related events and females showing relatively higher reporting of gastrointestinal and visual symptoms. Potential explanations include sex-related differences in body composition and pharmacokinetic handling (e.g., distribution volume and clearance), as well as hormonal influences on renal hemodynamics and symptom perception thresholds (32).
In the alectinib group, rROR analysis indicated male-enriched reporting for disease progression, drug ineffective, and blood creatinine increased (rROR>1; BH-FDR<0.05), whereas female-enriched reporting was observed for weight increased, constipation, and anaemia (rROR<1; BH-FDR<0.05). The sex contrasts for laboratory- and symptom-based events (e.g., creatinine increase, anaemia, constipation) may reflect sex-related variability in exposure and organ vulnerability (33, 34). Notably, the presence of sex differences in “disease progression” and “drug ineffective” should be interpreted cautiously, as these terms may capture complex real-world clinical contexts (e.g., treatment line, disease burden, or reporting practices) rather than direct biological susceptibility alone.
For brigatinib, no PT-level sex contrast remained statistically significant after BH-FDR adjustment, suggesting comparatively weaker or less detectable sex-associated variation in the PTs analyzed. Although nausea tended to be female-enriched and metastases to liver tended to be male-enriched at nominal significance, these differences did not withstand multiplicity control, likely reflecting limited statistical power and smaller report counts (35).
Overall, the results indicate that sex-related differences in ALK-TKI safety are agent-specific and endpoint-specific, and that rROR-based between-sex comparisons with multiplicity control help distinguish robust contrasts from descriptive fluctuations. Clinically, these findings support sex-informed, toxicity-targeted monitoring rather than broad sex-based assumptions. For example, early attention to renal-related events may be particularly relevant in males treated with crizotinib, whereas proactive assessment of gastrointestinal symptoms and visual complaints may be more pertinent in females. For alectinib, monitoring priorities may differ by sex, with consideration of creatinine changes and effectiveness-related reports in males and vigilance for constipation, weight increase, and anaemia in females (35). Importantly, these implications should be interpreted as risk-stratification cues derived from pharmacovigilance signals; treatment modifications should follow established clinical guidelines and individual patient context.

4.3
Time-to-onset characteristics and risk management
The time-to-onset (TTO) analysis based on Weibull modeling indicated an early-failure pattern (β< 1) for the three ALK-TKIs, suggesting that the hazard of AE occurrence is relatively higher shortly after treatment initiation and decreases over time. Consistently, the median TTOs for crizotinib, alectinib, and brigatinib were 51, 120, and 35 days, respectively, indicating that AE onset is generally concentrated in the early treatment phase, although the distribution differs across agents (with a relatively longer onset profile observed for alectinib).
At the PT level, hepatic function abnormal and pleural effusion were among the earliest events (median 14.5–55 days). Notably, brigatinib showed a short median onset for hepatotoxicity (19 days), consistent with an early and more predictable toxicity profile that may be amenable to proactive surveillance. These observations are supported by published case reports. For example, reversible hepatic dysfunction has been reported during crizotinib therapy and resolved after appropriate management (36). In addition, contralateral pleural effusion has been described within days after crizotinib initiation and improved after drug discontinuation (37). Together, these reports provide clinical context for the early-onset tendency of selected hepatic and pulmonary AEs associated with ALK-TKIs.
From a risk-management perspective, the early-failure onset pattern implies that enhanced monitoring should be prioritized soon after treatment initiation, particularly for organ systems implicated by early-onset signals. Clinicians may consider closer follow-up during the early treatment period (e.g., the first 8–12 weeks), including targeted assessment of liver function and evaluation of respiratory symptoms when indicated. Importantly, TTO patterns should be interpreted as supporting the timing of vigilance rather than prescribing uniform schedules, and management decisions should follow established clinical guidelines and individual patient risk profiles.

4.4
Limitations
External validation using the CVARDD database confirmed the robustness of FAERS-derived signal detection results. The three ALK-TKIs demonstrated consistent directional trends across both databases, particularly for hepatic function abnormal, anaemia, and oesophageal varices, supporting the reliability of cross-database pharmacovigilance analysis. Nonetheless, several limitations must be acknowledged.
First, both FAERS and CVARDD are spontaneous reporting systems and are therefore subject to reporting bias: severe AEs are more likely to be reported, whereas mild, self-limiting, or expected events may be underreported. Second, incomplete clinical information (e.g., body weight, dosing details, treatment line, and concomitant medications) limited more granular analyses, including adjustment for confounding and exploration of dose–response relationships. Third, in accordance with the READER-PV guidelines for pharmacovigilance studies, several inherent constraints of spontaneous reporting data must be explicitly recognized. These databases lack a valid denominator of exposed patients; thus, true AE incidence rates cannot be estimated. In addition, key background variables—baseline comorbidities, prior anticancer therapies, exact dosing regimens, and molecular profiles—are frequently missing, restricting subgroup stratification and limiting causal interpretation.
Importantly, disproportionality-based methods identify statistical associations rather than causality. Therefore, the observed sex contrasts and time-to-onset patterns should be interpreted as hypothesis-generating signals that inform risk stratification and monitoring priorities. Future research should aim to (i) validate key signals using clinically annotated real-world datasets (e.g., electronic health records or claims-linked registries) that enable adjustment for treatment line, disease burden, comorbidities, and concomitant drugs; (ii) conduct targeted subgroup analyses (e.g., age strata, CNS metastasis status, renal/hepatic dysfunction at baseline, and concomitant cardiovascular medications) to clarify effect modification; (iii) apply longitudinal designs with longer follow-up to evaluate persistence, recurrence, and outcomes of major toxicities; and (iv) integrate pharmacokinetic/pharmacogenomic or pharmacometabolomic evidence to elucidate mechanisms underlying sex-related and early-onset toxicities.

5
Conclusion
In summary, this study systematically characterized differences among three ALK-TKIs in terms of adverse event profiles, sex-specific variations, and time-to-onset patterns. All three drugs predominantly induced hepatobiliary and metabolic toxicities, with brigatinib presenting the highest risk of early hepatotoxicity, alectinib showing a greater metabolic burden, and crizotinib demonstrating an overall balanced safety profile.
Sex-related differences were most pronounced for cardiac–renal and hepatic–hematologic AEs, underscoring the necessity of implementing sex-specific pharmacovigilance strategies. Furthermore, the “early failure” distribution pattern observed across all agents highlights the need for intensive early-phase monitoring and timely intervention.
These findings provide real-world evidence to support personalized safety management for ALK-TKI therapy and lay the groundwork for future sex-specific drug safety research and precision risk stratification model development in the context of targeted cancer therapies.