Pharmacovigilance study of BCR-ABL1 tyrosine kinase inhibitors: a safety analysis of the FDA adverse event reporting system

Background With the increased use of BCR-ABL1 tyrosine kinase inhibitors (TKIs) in cancer patients, adverse events (AEs) have garnered considerable interest. We conducted this pharmacovigilance study to evaluate the AEs of BCR-ABL1 TKIs in cancer patients using the Food and Drug Administration Adverse Event Reporting System (FAERS) database. Methods To query AE reports from the FAERS database, we used OpenVigil 2.1. Descriptive analysis was then employed to describe the characteristics of TKIs-associated AE reports. We also utilized the disproportionality analysis to detect safety signals by calculating the proportional reporting ratio (PRR) and reporting odds ratios (ROR). Results From the FAERS database, a total of 85,989 AE reports were retrieved, with 3,080 significant AE signals identified. Specifically, imatinib, nilotinib, dasatinib, bosutinib, and ponatinib had significant AE signals of 1,058, 813, 232, 186, and 791, respectively. These significant signals were further categorized into 26 system organ classes (SOCs). The AE signals of imatinib and ponatinib were primarily associated with general disorders and administration site conditions. On the other hand, nilotinib, dasatinib, and bosutinib were mainly linked to investigations, respiratory, thoracic and mediastinal disorders, and gastrointestinal disorders, respectively. Notably, new signals of 245, 278, 47, 55, and 253 were observed in imatinib, nilotinib, dasatinib, bosutinib, and ponatinib, respectively. Conclusions The results of this study demonstrated that AE signals differ among the five BCR-ABL1 TKIs. Furthermore, each BCR-ABL1 TKI displayed several new signals. These findings provide valuable information for clinicians aiming to reduce the risk of AEs during BCR-ABL1 TKI treatment. Supplementary Information The online version contains supplementary material available at 10.1186/s40360-024-00741-x.


Introduction
Chronic myeloid leukaemia (CML) is a myeloproliferative neoplasm caused by the presence of the Philadelphia chromosome [1].The Philadelphia chromosome contains a BCR-ABL1 fusion gene that encodes a constitutively active cytoplasmic tyrosine kinase.This kinase activates various signals involved in promoting the proliferation and survival of myeloid progenitor cells [2].Therefore, the BCR-ABL1 kinase is the key target for CML therapy.Several BCR-ABL1 tyrosine kinase inhibitors (TKIs) have been developed and approved for the treatment of CML [3].In addition to being used to treat CML, these TKIs are also used to treat other malignancies, such as acute lymphoblastic leukemia (ALL), dermatofibrosarcoma protuberans (DFSP), and gastrointestinal stromal tumors (GIST) [2][3][4].The first-generation TKI is imatinib, while dasatinib, nilotinib, and bosutinib are second-generation TKIs, and ponatinib is the third-generation TKI [3,4].These TKIs inhibit the activity of BCR-ABL tyrosine kinase by binding to it in an inactive form, leading to the death of tumor cells.Although BCR-ABL1 TKIs have significantly improved the survival of patients with CML, they are not without adverse events (AEs) [5,6].AEs induced by BCR-ABL1 TKIs may reduce therapeutic adherence; therefore, pharmacovigilance studies of these drugs are essential for successful CML treatment.
Most of the efficacy and safety data of BCR-ABL1 TKIs are obtained from clinical trials.However, clinical trials have limitations in fully reflecting safety data from real clinical settings due to strict inclusion criteria, relatively small sample size, or limited follow-up durations.Therefore, there may be unknown adverse reactions occurring in real-world clinical settings.The Food and Drug Adverse Event Reporting System (FAERS) is one of the largest spontaneous reporting database in the world, providing sufficient data to verify and supplement the findings of clinical trials [7].As the FAERS database is publicly available and reflects complete AE reports in real-world clinical settings, it is widely used to detect potential drug-associated AEs [8].In this study, we conducted a pharmacovigilance analysis using the FAERS database to evaluate and compare the safety of BCR-ABL1 TKIs.

Data sources and data collection
The data collection for this study utilized OpenVigil 2.11, which allowed us to retrieve the FAERS data from drug approval up to the third quarter of 2022 (Table 1).We collected specific clinical characteristics for each adverse event (AE) report, including individual safety reports (ISR), outcome, drug name, role code, dosage, indication, event, case ID, gender, reporter country, and age in the report.Given that the FAERS database is a compilation of submissions from various sources, duplicates can be found within the dataset.To address this, we utilized the case ID and ISR as key filters, choosing the higher ISR in cases where the case ID matched.Furthermore, in order to minimize confounding effects, preferred terms (PTs) associated with indication, off-label use, and product use issues were excluded from the analysis.

Adverse events and drug identification
We employed both the generic name and brand name, including "imatinib", "gleevec", "nilotinib", "tasigna", "bosutinib", "bosulif " "dasatinib", "sprycel", "ponatinib", and "iclusig", to identify AE records associated with the target drugs.Our search was specifically focused on AE reports in which the drug was considered the primary suspect, aiming to improve accuracy.The AEs were coded using the PTs according to the Medical Dictionary for Regulatory Activities (MedDRA) terminology.Additionally, we utilized MedDRA (version 22.1) to classify the AEs in each report into the corresponding system organ class (SOC) levels.

Statistical analysis
A descriptive analysis was conducted to summarize the clinical characteristics of the AE reports, including the event, outcome, gender, age, and reporting country.To study the correlation between the target drug and the target AEs, a disproportionality analysis was employed.The reporting odds ratio (ROR) and proportional reporting ratio (PRR) were calculated to generate signals of disproportionate reporting.The specific algorithm for the disproportionality analysis was outlined in Table 2 [9], while the equations and criteria were listed in Table 3 [9].A signal is considered significant when both algorithms yield positive results.Furthermore, a higher ROR or PRR value

PTs analysis
Two algorithms and their corresponding criteria were used to detect all AE signals of BCR-ABL1 TKIs.The numbers of significant AE signals for imatinib, nilotinib, dasatinib, bosutinib, and ponatinib were as follows: 1,058, 813, 232, 186, and 791.The top 20 most frequently reported AE signals that met the criteria were listed in supplemental Tables 1-5.Among these signals, platelet count decreased was detected in five drugs, while pleural effusion and rash were detected in four drugs.Abdominal pain, death, fatigue, fluid retention, malignant neoplasm progression, myalgia, and thrombocytopenia were detected in three drugs.Abdominal pain upper, anaemia, chest pain, constipation, diarrhoea, drug intolerance, decreased haemoglobin, headache, hospitalisation, neoplasm progression, oedema, pain in extremity, pancytopenia, pyrexia, second primary malignancy, and decreased white blood cell count were detected in two drugs.

Discussion
There are several methods available for mining AE data from the FAERS database.These methods include using the programming language Ruby, the statistical environment R, and web-based services [13].However, the use of Ruby and R may prove difficult for users when it comes to mining and analyzing the FAERS data.In contrast, OpenVigil is a web-based pharmacovigilance tool that facilitates the extraction, filtration, data mining, and disproportionality analysis of AE reports from the FAERS database [9,14].It is worth noting that while there are other applications for mining pharmacovigilance data, they differ from OpenVigil in terms of their data cleaning techniques [15,16].OpenVigil, on the other hand, has undergone successful verification by the FDA and is widely employed in pharmacovigilance studies [14,17].Therefore, for the purposes of the present study, Open-Vigil 2.1 was utilized for the retrieval and analysis of the FAERS data.
Two disproportionality analysis methods are applied to detect any potential positive signal: frequentist analysis and Bayes analysis [18].The frequentist analysis includes ROR and PRR, while the Bayes analysis includes Bayesian confidence propagation neural network and multiitem gamma poisson shrinker.The frequentist analysis is simple, easy to calculate and understand.However, it is very sensitive to small samples and therefore prone to producing false positive signals when the number of reports is small [19].On the other hand, the Bayes analysis is more complex since it involves distributional assumptions and optimization of the likelihood function.In this study, to reduce bias caused by using a single algorithm, two different data mining algorithms (ROR and PRR) were employed for signal detection.A signal is considered significant when both algorithms yield positive results.
Among the five BCR-ABL1 TKIs, imatinib is associated with the largest number of AE reports and the broadest signal spectrum, which is consistent with clinical practice.This is because imatinib is the first BCR-ABL1 TKI approved by the FDA and EMA.The higher incidence rate of CML in patients aged 45-64 years old accounted for a slightly greater proportion compared to other age groups [20].It is worth noting that AEs were more likely to occur in males, which may be related to the higher prevalence of CML in males compared to females [20].Although the FAERS database theoretically includes global adverse event data, a majority of the data comes from the United States.Consequently, research based on the FAERS data mainly focuses on the reporting regions of AE reports in North America [21,22].Analysis of the SOC level reveals that certain disorders and conditions were not detected for specific TKIs.For example, bosutinib did not exhibit endocrine disorders, immune system disorders, reproductive system and breast disorders, and surgical and medical procedures.Additionally, bosutinib and ponatinib did not show any cases of pregnancy, puerperium and perinatal conditions, while dasatinib did not exhibit any cases of psychiatric disorders and social circumstances.Furthermore, there were variabilities in the case numbers, PT numbers, and signal intensity of SOCs across individual BCR-ABL1 TKIs.Therefore, when prescribing BCR-ABL1 TKIs for cancer patients, it is important for clinicians to consider both the adverse reaction characteristics of each drug and the specific details of the patient.
The adverse reactions most frequently reported in the labels of BCR-ABL1 TKIs were fluid retention events, gastrointestinal toxicities, hematologic toxicities, rash, pain, hepatic toxicities, and hemorrhage [23][24][25][26][27][28][29][30][31][32][33].These adverse reactions are consistent with the results of the AE signals analysis, providing credibility to this study.Among these adverse reactions, fluid retention events, such as edema, pleural effusion, and pericardial effusion, were the most common for BCR-ABL1 TKIs.Specifically, dasatinib showed a stronger association with edema (PRR = 7.168, ROR = 7.389), pleural effusion (PRR = 33.536,ROR = 36.388),and pericardial effusion (PRR = 13.327,ROR = 13.484)compared to the other four drugs.In terms of gastrointestinal toxicities, bosutinib had a stronger association with nausea and This study comprehensively revealed the AE signals of BCR-ABL1 TKIs based on real-world data, providing strong support for monitoring adverse drug reactions (ADRs) and rational clinical use of drugs.However, there are several limitations to this study.Firstly, the FAERS database, which was used in this analysis, is a self-reporting system with reporting randomness and massive missing data.As a result, the analysis results may deviate from the actual situation [34].Secondly, since the FAERS database only includes cases with AEs, it cannot provide the total number of patients receiving BCR-ABL1 TKIs treatment, making it impossible to estimate the incidence rate of AEs associated with each drug [35].Thirdly, due to the self-reporting nature of the FAERS database, it is difficult to determine the causal relationship between AEs and drugs [36].Therefore, further studies are required to validate our findings.Finally, there is a distinction between AEs and ADRs.While ADRs are caused by drugs, AEs can be caused by drugs as well as by the disease itself or other factors.Therefore, clinicians must assess if an AE is drug-related using clinical realities.Despite these limitations, the FAERS database remains a rich resource and an important tool for post-marketing surveillance.

Conclusion
Based on the FAERS database, we mined and analyzed the AE signals of different BCR-ABL1 TKIs in this study.The study revealed that AE signals associated with BCR-ABL1 TKIs varied.Imatinib and ponatinib mainly exhibited AE signals related to general disorders and administration site conditions.On the other hand, nilotinib, dasatinib, and bosutinib showed AE signals primarily associated with investigations, respiratory, thoracic and mediastinal disorders, and gastrointestinal disorders, respectively.Furthermore, new signals were observed for each BCR-ABL1 TKI, which have implications for clinical practice and drug monitoring.Clinical practices usually use official drug labels, and as we mentioned previously, similar studies are useful as idea generating that need to be further investigated taking into account all available evidence (review of actual reports from the database, clinical data, studies etc.).

Fig. 2 Fig. 1
Fig. 2 Proportion of reported cases for different BCR-ABL1 TKIs in SOC level

Table 2
Disproportionality analysis algorithm

Table 3
The equations and criteria for the algorithm

Table 4
Characteristics of AE reports for different BCR-ABL1 TKIs