Introduction

The use of taxanes in cancer treatment has been gradually increasing, particularly in non-small cell lung cancer and breast cancer1. Taxanes, which include paclitaxel, docetaxel, and cabazitaxel, as well as various formulations such as nab-paclitaxel and paclitaxel liposomes, have become integral parts of contemporary oncological treatment regimens2,3. Recent studies have reported that the administration of taxanes is often associated with drug-induced liver injury (DILI) during cancer therapy4. However, the specific clinical manifestations and underlying mechanisms of DILI related to taxanes in cancer patients have not been well documented. DILI shows various clinical manifestations, varying from the increasement of hepatic transaminases to hepatic failure, and even result in patient mortality5. It is necessary to comprehensively understand the safety and associated risks of taxanes for the optimization of cancer treatment.

Current research on the adverse events of taxanes primarily derives from short-term clinical trials, which often do not capture rare but serious adverse events like DILI. There is also limited literature reporting on the real-world implications of taxanes associated liver injury. The Food and Drug Administration Adverse Event Reporting System (FAERS) provides a valuable opportunity to explore these associations on a broader scale, enabling researchers to examine post-market drug safety data more thoroughly6. Investigating these associations through large-scale data analysis could bridge existing gaps in our understanding of taxane-induced liver injury.

Although previous studies have indicated that taxanes may influence liver metabolism7,8, systematic analyses focusing on the interactions between taxanes and genes associated with liver injury remain scarce. Network pharmacology, an emerging bioinformatics analysis method developed in recent years, can be used to systematically and comprehensively predict the mechanisms of drug action9,10. In this study, we utilized the FAERS database to identify DILI events associated with the use of taxanes, thereby elucidating the relationship between taxanes and liver injury. Based on this, network pharmacology techniques were employed to explore the potential mechanisms of DILI associated with taxanes, preliminarily revealing the molecular targets and the similarities and differences in their mechanisms of inducing DILI. This research can provide valuable insights into the mechanisms of DILI associated with taxanes, contributing to the improvement of medication safety for cancer patients.

Results

Descriptive analysis

From the first quarter of 2004 to the first quarter of 2024, a total of 50,390,907 adverse event reports (AERs) were collected from the FAERS database. Of these, there were 156,120 records of adverse events (AEs) for the three target drugs classified as primary suspected (PS) drugs. The most of these reports originated from North America and certain European countries, with the specific distribution areas illustrated in Fig. 1A-C. Among these reports, there were 2,168 AERs related to DILI, including 1,196 reports (55.17%) for paclitaxel, 238 reports (10.98%) for docetaxel, and 734 reports (33.86%) for nab-paclitaxel. The annual distribution of these reports is depicted in Fig. 1D. Additionally, we can see that the number of female patients is greater than that of male patients from Table 1. Regarding age distribution, most patients are concentrated between 18 and 65 years old, but elderly individuals over 65 also account for nearly one-third, which requires attention. Most reporters are healthcare professionals, with reports from consumers being relatively low, all below 10%. The geographical distribution of reports reveals that liver injury associated with paclitaxel is most frequently documented in European countries such as France and Italy. In contrast, reports for docetaxel often lack specific origin information, while nab-paclitaxel reports are mainly from Japan and the United States. China shows a reporting rate of less than 10% despite its extensive population, which may be attributed to its national adverse reaction reporting system and the utilization of locally marketed paclitaxel formulations. The median time to onset of liver injury is 21 days for both paclitaxel and nab-paclitaxel, whereas docetaxel tends to cause liver injury earlier, usually within 11 days.

Fig. 1
figure 1

Geographical distribution of adverse events to taxanes and annual changes in the incidence of liver injury. (A–C) The geographical distribution maps of three taxanes. (D) The number of three taxanes inducing liver injury from the first quarter of 2004 to the first quarter of 2024 report.

Full size image
Table 1 Characteristics of reports on taxanes-associated dilis in the FAERS database (January 2004 to March 2024).
Full size table

Liver injury is a critical condition that can result in severe adverse consequences. In our study, we compiled and analyzed the outcomes of adverse events related to liver injury, as illustrated in Fig. 2. Our findings indicate that while docetaxel is associated with the lowest incidence of liver injury among the drugs studied, it showed a statistically higher mortality rate compared to paclitaxel (χ² = 19.2, p < 0.0001). Specifically, the mortality rate associated with liver injury was 25.21% for docetaxel, compared to 14.34% for paclitaxel and 18.01% for nab-paclitaxel. These observations suggest that liver injury outcomes may vary across taxanes, with docetaxel-associated cases warranting particular attention in clinical monitoring.

Fig. 2
figure 2

Outcome of liver injury caused by three taxanes.

Full size image

Signal detection of DILI-related AEs in the FAERS database

Signal detection of the results revealed an association between taxanes and DILI. Paclitaxel (ROR = 2.35) and nab-paclitaxel (ROR = 3.14) were associated with an increased incidence of DILI, while docetaxel (ROR = 0.68) had no significant impact on the incidence of DILI.

The specific signal detection results for adverse events under certain Preferred Terminology (PT) conditions are shown in Table 2. Paclitaxel had positive signals in 14 PT terms, including 30 patients with pseudocirrhosis (ROR = 61.77), 41 patients with hepatic atrophy (ROR = 49.81), 99 patients with hypertransaminasaemia (ROR = 10.51), and immune-mediated hepatitis (ROR = 6.81), among others. Docetaxel had only 9 positive signals, mainly related to liver tenderness (ROR = 26.33), decreased bilirubin (ROR = 11.86), bilirubin conjugated increased (ROR = 3.38) and so on. Nap-paclitaxel, like paclitaxel, also showed 14 positive signals, including a similar pseudocirrhosis (ROR = 18.81), immune-mediated hepatitis (ROR = 13.07), and hypertransaminasaemia (ROR = 3.52), but nab-paclitaxel also has some abnormalities related to bilirubin, such as blood bilirubin abnormality (ROR = 4.60) and cholestatic jaundice (ROR = 2.60).

Table 2 Positive signal strength for liver injuries associated with taxanes based on PT levels of FAERS.
Full size table

Drug-liver injury-related gene interaction network analysis

A total of 116 potential targets for docetaxel and 122 potential targets for paclitaxel were predicted, linked to 1715 genes associated with liver injury (Supplementary Table S1-S3). As shown in Fig. 3, by intersecting these gene sets, 48 interactive target genes were identified that represent the overlap between the targets of docetaxel and the genes implicated in hepatic injury. Similarly, 53 interactive target genes were identified representing the overlap between the targets of paclitaxel and the liver injury-related genes. Hypergeometric testing demonstrated significant enrichment of overlapping targets between taxanes and liver injury genes. For docetaxel, 48 of its 116 predicted targets overlapped with liver injury genes (expected by chance = 10.27; P = 1.16 × 10⁻21; enrichment ratio = 4.67). Similarly, paclitaxel showed 53 overlaps among 122 targets (expected = 10.79; P = 5.62 × 10-26; enrichment ratio = 4.91), indicating strong non-random associations between taxane targets and hepatic injury mechanisms. Comparative analysis of the potential targets for liver injury induced by both drugs revealed that while paclitaxel and docetaxel share some common targets, there are also distinct targets unique to each. Specifically, there are 35 shared targets, with 13 exclusive targets for docetaxel and 18 exclusive targets for paclitaxel, as shown in Table 3. To elucidate the interrelationships among the potential targets, the Protein-Protein Interaction (PPI) was performed by using the STRING database (https://string-db.org). This facilitated the construction of a protein interaction network using Cytoscape 3.7.2 software, with the results showed in Fig. 4. For docetaxel, key targets (central nodes of the interaction) included EGFR, BCL2, HSP90AA1, MMP9, JAK2, HSP90AB1, IGF1R, KDR, MMP2, and MAPK1. For paclitaxel, key targets (central nodes of the interaction) included TNF, STAT3, EGFR, ERBB2, MMP9, HSP90AA1, MCL1, MMP2, IGF1R, and HSP90AB1. Topological analysis revealed their centrality within the network.

Fig. 3
figure 3

Potential targets of drug-induced liver injury. (A) Paclitaxel; (B) Docetaxe. The overlaps meant the potential targets.

Full size image
Fig. 4
figure 4

PPI network of targets associated with drug-induced liver injury. (A) Paclitaxel; (B) Docetaxe. The nodes with higher degree owned dark color.

Full size image
Table 3 Common and differential targets associated with docetaxel- and paclitaxel-induced liver injury.
Full size table

To further understand the involvement of docetaxel- and paclitaxel-induced liver injury target genes in biological signaling pathways, KEGG pathway enrichment analysis was conducted. The top 20 pathways for comprehensive mapping were illustrated in Fig. 5. Comparative analysis of the KEGG pathways revealed 132 shared pathways, with 5 specific pathways for docetaxel, namely ‘Ovarian Steroidogenesis’, ‘Regulation of Lipolysis in Adipocytes’, ‘B Cell Receptor Signaling Pathway’, ‘Parathyroid Hormone Synthesis, Secretion and Action’, and ‘Alcoholism’ (Supplementary Table S4). The above findings indicate that while docetaxel is a derivative of paclitaxel, the targets and mechanisms involved in the induced liver injury are similar but not entirely identical.

Fig. 5
figure 5

KEGG pathway enrichment analysis. (A) Paclitaxel; (B) Docetaxel. Each bubble represents a specific pathway. The horizontal axis represents the extent of enrichment for each pathway, while the size of the bubbles indicates the number of genes enriched in the corresponding pathway. Color indicates significance, with a gradient from green to red representing decreasing q-values.

Full size image

Discussion

With the widespread use of taxanes in cancer treatment, achieving a balance between risks and benefits has become particularly crucial. Drug-induced liver injury, as a rare but serious adverse reaction of taxanes, poses a significant risk to patient safety and treatment efficacy. The underlying mechanisms of DILI remain inadequately understood, potentially involving direct hepatocyte damage, immune-mediated responses, or disruptions in metabolic pathways, all of which can influence the overall prognosis of cancer patients11,12,13. Additionally, the unpredictability of DILI, coupled with the lack of specific biomarkers and diagnostic criteria, complicates timely identification and management, highlighting the urgent need for enhanced monitoring strategies14. Our study aims to systematically assess the risk of liver injury induced by taxanes through comprehensive analysis of large pharmacovigilance databases and to reveal potential biological mechanisms by identifying key genes and signaling pathways. These findings will provide valuable insights for drug safety monitoring and offer important references for the safe administration of these drugs in clinical practice.

The descriptive analysis of liver injury events reveals important epidemiological characteristics. Firstly, we found that female patients dominate among those with DILI related to taxanes, which may be partly due to these drugs being recommended as first-line treatments in clinical guidelines for gynecological malignancies and breast cancer. On the other hand, it may also be related to gender-related physiological differences and hormonal level changes. Studies have shown that estrogen may alter drug metabolism pathways by affecting the expression of liver metabolic enzymes, thereby increasing the risk of liver injury15,16. Additionally, age-related factors should not be overlooked. In our study, patients aged 65 years or older accounted for approximately one-third of reported liver injury cases, which may partly reflect the higher incidence of cancer diagnoses in this population. However, older adults may exhibit distinct pharmacokinetic responses and are more likely to have comorbidities or receive polypharmacy regimens, potentially exacerbating hepatotoxicity risks17. The FAERS database has inherent limitations in comprehensively capturing these complexities due to potential underreporting of drug-drug interactions and incomplete medication histories, particularly among elderly patients with multiple chronic conditions. Therefore, while our findings suggest that individualized drug use and monitoring based on age and gender could mitigate liver injury risks, clinical decision-making should integrate these observations with patient-specific factors (e.g., comorbidities, concomitant medications) and biomarker surveillance.

The signal detection results from the FAERS database indicate a significant association between paclitaxel and nab-paclitaxel with DILI, as evidenced by their ROR of 2.35 and 3.14. In contrast, docetaxel exhibited a lower DILI risk with an ROR of 0.68. Specifically, the forms of liver injury associated with paclitaxel and nab-paclitaxel include pseudocirrhosis, immune-mediated hepatitis and cholestasis, while docetaxel is primarily associated with abnormalities in liver enzymes and bilirubin levels. The study found that paclitaxel mainly relies on CYP2C8 and CYP3A4 for metabolism, whereas docetaxel primarily depends on CYP3A418. Different hepatic-drug enzymes affect the metabolic rate of drugs and the production of active metabolites, leading to oxidative stress and cellular damage in the liver4,19,20. Additionally, the longer half-life and slower clearance rate of paclitaxel may contribute to drug accumulation in the liver, which also increases the risk of liver injury. We also found that the median time to onset of liver injury for paclitaxel and nab-paclitaxel was 21 days, while docetaxel-related liver injury occurred more rapidly, typically within 11 days. This suggests that early and continuous monitoring of liver function is essential for patients undergoing docetaxel treatment. Interestingly, despite the lowest probability of liver injury, docetaxel-induced liver injury was associated with the highest mortality rate. These results indicate that although all three taxanes can cause liver injury, the underlying mechanisms may differ significantly.

Docetaxel, a semisynthetic derivative of the taxanes family, shares a similar structure with paclitaxel. However, our study revealed that docetaxel-induced liver injury is more severe and occurs earlier than that caused by paclitaxel, suggesting that the mechanisms underlying liver injury induced by these two drugs may not be entirely congruent. Through target prediction and the published literatures, BCL2, CNR2, and MAPK1 were identified as potential mediators of docetaxel-induced liver injury, which differ from those implicated in paclitaxel-induced liver injury. BCL2, known as the apoptosis regulator Bcl-2, typically suppresses apoptosis across various cell systems and is often downregulated in mouse liver tissue in response to LPS-induced hepatic injury21. CNR2, the cannabinoid receptor 2, is a heterotrimeric G protein-coupled receptor for the endocannabinoid 2-arachidonoylglycerol, which plays a role in mediating the inhibition of adenylate cyclase and may be involved in inflammatory responses. Activation of CNR2 has been shown to significantly ameliorate hepatic inflammation, impaired microcirculation, and fibrosis22. MAPK1, the mitogen-activated protein kinase 1, is a serine/threonine kinase that serves as a crucial component of the MAP kinase signal transduction pathway, with its activation often observed during chronic injury23,24. Previous research indicated that docetaxel could suppress the expression of BCL225 and promote the activation of MAPK126. Additionally, docetaxel may also increase the expression of CNR2, which enhances cellular sensitivity to the drug27. we propose that BCL2, CNR2, and MAPK1 are specific targets through which docetaxel mediates liver injury.

The mechanisms by which docetaxel induces liver injury appear to diverge from those of paclitaxel. The KEGG enrichment analysis of potential targets for docetaxel has identified the ‘Regulation of lipolysis in adipocytes (has04923)’pathway, which may represent a distinct pathway through which docetaxel exerts its hepatotoxic effects compared to paclitaxel. This pathway is central to lipid metabolism, facilitating the breakdown of triglycerides stored in adipose tissue into free fatty acids and glycerol. These molecules are subsequently utilized by the liver and other tissues for energy production28. Lipolysis is hormonally regulated, with insulin inhibiting the process and adipose-derived hormones such as adiponectin promoting it29. Three Key targets including the Insulin Receptor (INSR), Phosphoinositide 3-Kinase Catalytic Subunit Beta (PIK3CB), and Phosphoinositide 3-Kinase Catalytic Subunit Delta (PIK3CD) were hit in this pathway. INSR is a receptor tyrosine kinase which mediates the pleiotropic actions of insulin. When binding of insulin, INSR leaded to phosphorylation of several intracellular substrates and inhibits lipolysis30. While PIK3CB and PIK3CD are integral to the phosphoinositide signaling pathway that regulates lipid metabolism in response to hormonal and growth factor stimuli31. It has been reported that the INSR/PI3K/AKT pathway was activated during the progression of liver fibrosis32. The modulation of these targets by docetaxel within the ‘Regulation of lipolysis in adipocytes’ pathway may lead to an increase in circulating free fatty acids. Elevated levels of free fatty acids can accumulate in liver, resulting in steatosis, a condition characterized by the accumulation of fat within liver cells33. Persistent steatosis can evolve into non-alcoholic steatohepatitis (NASH), which is associated with liver inflammation and cell damage, thereby increasing the risk of liver injury34,35,36. Although no studies have yet confirmed that docetaxel can directly modulate INSR, our findings suggest that INSR may be a potential target of docetaxel, differing from paclitaxel, which does not appear to affect this target. Therefore, the ‘Regulation of lipolysis in adipocytes’ pathway, along with its key targets INSR, PIK3CB, and PIK3CD, are significantly implicated in the pathogenesis of DILI induced by docetaxel. In the PPI network analysis, edge betweenness is a key metric for evaluating the importance of edges, where higher values indicate critical ‘bridging’ or ‘bottleneck’ roles in mediating intermodular communication37. Further analysis of docetaxel-specific hepatotoxicity targets revealed that the interaction edges between BCL-2 and core components of the ‘Regulation of lipolysis in adipocytes (has04923)’ pathway (INSR and PIK3CD) exhibited higher edge betweenness scores (Supplementary Table S5). This topological feature suggests that BCL-2 serves as a pivotal bridge connecting the ‘apoptosis regulation module’ with the ‘lipid metabolism module’. Notably, hepatic lipid metabolism dysregulation-induced free fatty acid (FFA) accumulation has been established as a central driver of apoptosis in drug-induced liver injury38. Accumulating evidence suggests that valproic acid and tamoxifen have been reported to trigger drug-induced apoptosis and liver injury through dysregulation of hepatocyte lipid metabolism, mechanisms analogous to our findings with docetaxel39.These findings collectively suggest that docetaxel may disrupt hepatocyte lipid homeostasis through dysregulation of the ‘Regulation of lipolysis in adipocytes’ pathway, thereby triggering apoptosis cascades and ultimately contributing to liver injury. Furthermore, taxanes such as docetaxel and paclitaxel undergo hepatic metabolism primarily via cytochrome P450 enzymes, generating hydroxylated derivatives as major metabolites40. Although these metabolites exhibit lower cytotoxic activity compared to their parent drugs, docetaxel’s significantly longer systemic retention time (average half-life ~ 11 hours vs. paclitaxel’s ~ 5.8 hours) may lead to prolonged hepatic accumulation of both the parent compound and its derivatives40,41,42,43. This pharmacokinetic divergence could synergize with docetaxel’s unique target interactions, potentially amplifying lipid metabolic disturbances and subsequent hepatotoxicity—a hypothesis supported by our pathway analysis of adipocyte lipolysis regulation. Thus, gaining insights into these mechanisms is crucial for devising strategies to prevent or mitigate liver injury in patients undergoing docetaxel chemotherapy.

The limitations of this study should be interpreted within the context of pharmacovigilance methodologies. First, reliance on the FAERS database introduces inherent biases, including geographical reporting disparities and potential underreporting in vulnerable populations such as elderly patients with polypharmacy. The voluntary nature of adverse event reporting can result in underreporting or overreporting of specific drug-related injuries, potentially skewing the data. Second, while disproportionality analyses (ROR and BCPNN) provide valuable signal detection capabilities, their performance is sensitive to low reporting counts and arbitrary threshold selections, which may inflate false-positive signals for rare events or drugs with limited exposure44. Finally, the absence of laboratory-based experimental validation restricts our ability to elucidate the underlying biological mechanisms contributing to DILI associated with the identified drugs.

Conclusion

This pharmacovigilance study observed more frequent reports of associations between taxanes and DILI through disproportionality analysis of FAERS data. While docetaxel showed fewer reported DILI cases compared to other taxanes, these reports were more frequently associated with fatal outcomes and shorter time-to-onset, within the limitations of spontaneous reporting data. Network pharmacology predicted unique molecular targets and pathways for each taxane, suggesting different mechanisms of hepatotoxicity. These findings contribute to optimizing treatment regimens in clinical practice, enhancing patient safety, and improving the overall efficacy of taxanes-based chemotherapy regimens.

Materials and methods

Data source and processing

This study utilized data from the FAERS database (https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html), a publicly accessible database containing spontaneous adverse event reports submitted globally. We downloaded the report files of the American Standard Code for Information Interchange (ASCII) from the FAERS database for the period from January 2004 to March 2024, comprising seven structured subfiles: DEMO (demographic), DRUG (drug), REAC (reaction), OUTC (outcomes), RPSR (report source), THER (drug therapy dates), and INDI (indication information).

Duplicate reports were removed following FDA guidelines, prioritizing entries by CASEID, FDA_DT (receipt date), and PRIMARYID. And drug names in the data table were standardized using the Medex_UIMA 1.3.7 system. In this study, three representative taxanes were selected, including paclitaxel, docetaxel, and nab-paclitaxel. We limited the search to Primary Suspect (PS) drugs and extracted reports that suspected these three drugs as the main drugs related to Adverse Drug Events (ADEs). These reports included, but not limited to report date, patient age and gender and reporter. Adverse events were classified via Medical Dictionary for Regulatory Activities (MedDRA v26.1), identifying 277 Preferred Terms (PTs) through the standardized query ‘Drug-Related Hepatic Disorders - Comprehensive Search’ (SMQ ID: 20000006), which served to filter clinically significant drug-related liver injury signals. The workflow was shown in Fig. 6.

Fig. 6
figure 6

Flow chart of data extraction. A detailed description of the data extraction process for drug-induced liver injury (DILI) adverse events for taxanes in the FAERS.

Full size image

Signal detection and data analysis

This study utilized disproportionality analysis methods, a pharmacovigilance approach commonly employed to detect signals between drugs and adverse events. Two complementary algorithms were implemented: (1) the Reporting Odds Ratio (ROR), which quantifies the relative reporting frequency of a specific adverse event for the target drug compared to all other medications in the database, (2) the Information Component (IC), a Bayesian data mining approach derived from Bayesian Confidence Propagation Neural Networks (BCPNN)44. Signal thresholds were defined as follows: ROR signals required both a lower 95% confidence interval (CI) > 1.0 and ≥ 3 case reports, while IC signals required IC025 > 0 (lower 95% CI threshold). All data processing and statistical analyses were performed using R software (version 4.3.3) with faersR package (version 0.0.0.90003)45.

Screening targets of taxanes associated with liver injury

The SwissTargetPrediction online tool (http://www.swisstargetprediction.ch/index.php) and DrugCentral 2023 ( https://drugcentral.org/) were utilized to predict the potential targets of taxanes (docetaxel and paclitaxel)46. Disease targets associated with liver injury were collected from open-source databases, including the GeneCards database (https://www.genecards.org), DisGenet database (https://www.disgenet.org/home), and OMIM database (https://omim.org). Subsequently, the targets of taxanes and liver injury were identified by using the online Venny tool (https://bioinfogp.cnb.csic.es/tools/venny/index.html). The hypergeometric test was used to assess if the overlap between the drug targets and liver injury genes was significant compared to random. The background number of human protein-coding genes was set as 19,37047. The overlapping targets were considered as the potential targets of taxanes associated with liver injury.

PPI analysis and KEGG enrichment analysis

The PPI network was constructed utilizing the STRING online tool (https://cn.string-db.org/ (version 12.0))48. The main PPI network parameter settings are as follows: organism was set as Home sapiens, the minimum required interaction score was set as medium confidence (0.400), max number of interactors to show was set as ‘none /query proteins only’. Subsequent analysis of the network was conducted using Network Analyzer, a plugin integrated within the Cytoscape environment (version 3.7.2). The degree values of the targets in PPI network, as well as the edge betweenness between different targets, were both calculated using the plugin (Network Analyzer) in Cytoscape. The top 10 targets, as determined by degree values within the PPI network, were designated as core targets49. KEGG pathway enrichment analysis was executed through the Metascape platform (https://metascape.org/gp/index.html#/main/step1)50, with the visualization of the top 20 enriched KEGG pathways. This analytical approach was employed to elucidate the potential mechanisms underlying liver injury induced by taxanes.