Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2026 Jul 21;40(8):e71026. doi: 10.1002/jbt.71026

Deciphering the Formulation‐Dependent Neurotoxicity of Irinotecan: An Integrated Pharmacovigilance and Mechanistic Study

Cheng Shen 1, Wei Ma 2, Jing Lu 3, Yongbo Tang 1, Wei Cao 1,✉
PMCID: PMC13387174  PMID: 42478934

ABSTRACT

Irinotecan is a key chemotherapeutic agent, but its neurotoxicity limits its utility. Observed differences in neurotoxicity between conventional and liposomal formulations are poorly characterized mechanistically. This study aimed to systematically compare their neurotoxicity profiles and investigate the underlying mechanisms, testing the hypothesis that the parent drug irinotecan contributes directly to neurotoxicity. We employed a multi‐dimensional strategy. Computational toxicology predicted neurotoxicity for irinotecan and its metabolite SN‐38. Disproportionality analysis of the FDA Adverse Event Reporting System (FAERS) compared real‐world neurotoxicity signals. Network pharmacology and molecular docking explored irinotecan's direct molecular targets and pathways. Multi‐platform computational prediction confirmed high neurotoxicity risk for both irinotecan and SN‐38. FAERS analysis revealed a distinct clinical profile: conventional irinotecan (C‐Irinotecan) was associated with both central (CNS) and peripheral nervous system (PNS) adverse events, whereas liposomal irinotecan (L‐Irinotecan) signals were confined to the periphery. This CNS‐specific disparity, combined with pharmacokinetic evidence, implicated the parent drug. Network analysis identified eight core neural targets (e.g., AKT1, EGFR) with high binding affinity to irinotecan. Pathway enrichment highlighted the MAPK/p38 cascade as a central mechanism. This integrated analysis demonstrates a formulation‐dependent neurotoxicity dichotomy for irinotecan. We propose and provide converging evidence for a novel paradigm: prototype irinotecan is a direct mediator of neurotoxicity, and its CNS access dictates central effects. These findings offer critical insights for formulation safety and neuroprotective strategies.

Keywords: drug‐induced neurotoxicity, FAERS database, irinotecan, molecular docking, network pharmacology, pharmacovigilance

Overall workflow of the study, from computational toxicology screening and FAERS‐based data mining to mechanistic prediction of neurotoxicity of irinotecan.

graphic file with name JBT-40-e71026-g005.jpg

1. Introduction

Irinotecan, a topoisomerase I inhibitor, is a cornerstone of chemotherapy regimens for metastatic colorectal, gastric, and pancreatic cancers [1]. Despite its efficacy, treatment is often limited by adverse effects, with neurotoxicity emerging as a clinically significant concern [2]. Notably, clinical reports and trials suggest a disparity in neurotoxicity profiles between the conventional formulation (C‐Irinotecan) and its liposomal counterpart (L‐Irinotecan) [3, 4]. While C‐Irinotecan is associated with both central nervous system (CNS) events (e.g., dysarthria) and peripheral neuropathy, L‐Irinotecan appears to carry a lower risk, particularly for CNS toxicity [5, 6, 7]. However, this observed difference remains anecdotal, lacking systematic large‐scale validation. More critically, the mechanistic basis for this potential formulation‐dependent neurotoxicity is entirely unexplored.

The prevailing paradigm attributes irinotecan's toxicities primarily to its potent active metabolite, SN‐38 [8, 9]. This metabolite‐centric view, however, fails to explain why two formulations delivering the same active moiety to systemic circulation would exhibit distinct neurological side effect patterns. This discrepancy points to a fundamental knowledge gap: could the parent drug, irinotecan itself, play a direct and underappreciated role in neurotoxicity, with its effect modulated by formulation‐dependent pharmacokinetics and tissue distribution? Given the rarity of neurotoxic events, this question is challenging to address, and prospective comparative randomized trials have not yet been reported.

To bridge this gap, we employed a multi‐tiered translational strategy. First, we conducted a computational toxicology screen to prospectively assess the neurotoxic potential of both irinotecan and SN‐38, establishing a predictive baseline. We then performed a comprehensive analysis of the FDA Adverse Event Reporting System (FAERS) to systematically compare the real‐world neurotoxicity signals of C‐Irinotecan and L‐Irinotecan, testing the clinical reality of the predicted toxicity and quantifying the suspected disparity [10]. Finally, to move beyond descriptive epidemiology and probe causality, we integrated network pharmacology and molecular docking. This mechanistic arm aimed to identify whether irinotecan possesses high‐affinity molecular targets within networks relevant to neurotoxicity, thereby providing a plausible biological foundation for a direct effect of the parent compound.

This study is the first to integrate predictive toxicology, large‐scale pharmacovigilance, and computational biology to dissect the neurotoxicity of irinotecan formulations. This study aims both to characterize the clinical risk profiles and to provide initial evidence for a hypothesis: irinotecan itself may play a key role in neurotoxicity, and its formulation‐dependent access to the CNS could underlie the differing neurological safety profiles of conventional and liposomal irinotecan.

2. Methods

2.1. Potential Toxicological Prediction of Irinotecan

Four computational toxicology platforms, namely ADMETlab 3.0, toxCSM, ProTox‐3.0 and NeuroTDPi, were employed to predict the potential toxicological endpoints of irinotecan. The canonical SMILES structure of irinotecan and SN‐38 was retrieved from the PubChem database and input into each platform to generate probability scores for toxicity prediction. All prediction results were normalized to a probability scale ranging from 0 to 1. According to the official documentation of ADMETlab 3.0, a probability score greater than 0.5 indicates a high‐risk toxicity prediction [11]. In the ProTox database, a higher probability score represents a higher confidence level in the predicted toxicity [12]. In this study, probability scores ranging from 0.5 to 1.0 across all platforms were regarded as indicators of potential toxicological concerns. A Toxicity risk prediction heatmap was generated using the Python programming language with the Seaborn statistical visualization library. The heatmap displays toxicity risk probabilities (0 = low risk, 1 = high risk).

2.2. Dataset Acquisition 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 repository containing spontaneous AE reports submitted globally [13]. The FAERS database includes seven datasets: DEMO (demographic), DRUG (drug), REAC (reaction), OUTC (outcomes), RPSR (report source), THER (drug therapy dates), and INDI (indication information). We downloaded the American Standard Code for Information Interchange (ASCII) data files for the period from Q1 2004 to Q2 2025. Data for all reports mentioning irinotecan were extracted and merged from the relevant FAERS tables. Following FDA recommendations, duplicate reports were identified and removed using a standardized deduplication algorithm: For cases sharing an identical CASE ID, the report with the most recent FDA_DT (FDA receipt date) was retained. If multiple reports shared identical CASE ID and FDA_DT values, the report with the highest PRIMARY ID was retained. Irinotecan is commercially available in two main formulations: C‐Irinotecan and L‐Irinotecan. Reports were first categorized as C‐Irinotecan or L‐Irinotecan based on the drug name. For reports where this information was missing, the formulation was determined by cross‐referencing the reported manufacturer name.

2.3. AE Signal Detection and Data Analysis

In the field of pharmacovigilance, disproportionality analysis plays a fundamental role as a screening tool to identify potential relationships between a specific AE and a drug. Disproportionality analysis was performed through the application of the Reporting Odds Ratio (ROR) and Proportional Reporting Ratio (PRR) methods to identify potential associations between the drugs and AEs [14]. Signal thresholds were defined as follows: ROR signals required both a lower 95% confidence interval (CI) > 1.0 and at least 3 case reports. PRR signals required both PRR > 2.0 and χ 2 ≥ 4 and at least 3 case reports. All data processing and statistical analyses were performed using SAS software (version 9.4). The summary of major algorithms used for signal detection is shown in Table S1.

2.4. Identification of Common Targets for Irinotecan and Neurotoxicity

First, to identify potential molecular targets of Irinotecan, its Canonical SMILES notation was obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) [15] and imported into the SwissTargetPrediction (http://www.swisstargetprediction.ch/) [16]. The 2D structure of Irinotecan was obtained from PubChem for use in the PharmMapper server (http://lilab.ecust.edu.cn/) [17]. Second, to identify neurotoxicity‐related targets, the CTD (Comparative Toxicogenomics Database) (https://ctdbase.org/) [18] and the GeneCards database (https://www.genecards.org/) [19] and the OMIM (Online Mendelian Inheritance in Man) (https://www.omim.org/) [20] were queried using the keyword “neurotoxicity.” Finally, the potential Irinotecan targets were intersected with the neurotoxicity‐related targets using a Venn diagram (https://jvenn.toulouse.inrae.fr) [21] to identify overlapping genes, which were defined as the common targets for subsequent analysis. These shared targets in both the Irinotecan target dataset and the neurotoxicity target dataset reflect a statistical association between Irinotecan and its induced neurotoxicity at the molecular level.

2.5. PPI Network Construction

The common targets were imported into the STRING database (https://cn.string-db.org) to evaluate protein–protein interactions (PPIs) [22]. The search was limited to Homo sapiens with a minimum required interaction score of 0.7 (high confidence). The resulting PPI network was visualized and analyzed using Cytoscape software (version 3.7.2). The CytoHubba plugin was used to identify the top 10 hub genes based on EPC, Degree, MNC, and MCC centrality [23] Only the genes that appear in all four of the above algorithms are regarded as high‐confidence ones.

2.6. Molecular Docking

Molecular docking was performed between selected core proteins and compounds to evaluate binding affinities and interactions. The 3D structure of irinotecan was retrieved from the PubChem database, and the protein structures were obtained from the RCSB Protein Data Bank database. Molecular docking was performed using CB dock2 (http://183.56.231.194:8001/cb-dock2/index.php) [24]. Ligand–protein interactions between irinotecan and each protein were examined in PyMOL (3.1.6.1) and Discovery Studio Visualizer (2025.1.0), focusing on hydrogen bonding and hydrophobic interactions.

2.7. Functional and Pathway Enrichment Analysis

The common targets were imported into the Metascape platform (https://metascape.org/gp/) [25] for functional enrichment analysis. KEGG pathway and GO enrichment analyses were performed for the target gene list, selecting Homo sapiens as the species, with the significance threshold set at p < 0.05 and a minimum enrichment factor of 1.5 [26, 27]. These standard parameter settings help ensure the reliability and biological significance of the enrichment results, effectively validating the enrichment analysis process. The results of the enrichment analysis indicate that these disease‐specific pathways act through the shared targets, forming a functional link with irinotecan‐induced neurotoxicity (IIN), and then the results were visualized using the Bioinformatics Online Platform (https://www.bioinformatics.com.cn). Finally, an integrated network illustrating the relationships between Irinotecan, its common targets, and the enriched pathways was constructed using Cytoscape.

2.8. Validation of Core Target Expression in Human Brain Tissues

The brain is the main component of the human nervous system, and its tissue expression can reflect the overall expression characteristics of the nervous system. To verify the nervous system specificity of the eight core hub targets, their mRNA and protein expression profiles were obtained from three authoritative public databases: Human Protein Atlas (HPA, https://www.proteinatlas.org/), GTEx (Genotype‐Tissue Expression, https://gtexportal.org/), and FANTOM5 (https://fantom.gsc.riken.jp/). Expression levels were determined following the official grading standards of each database, and data were summarized based on their relative abundance in key brain regions.

3. Results

3.1. Computational Toxicology Prediction

To prospectively evaluate the organ‐specific toxicity profiles of irinotecan and its active metabolite SN‐38, we employed a consensus strategy across four independent predictive platforms: ADMETlab 3.0, toxCSM, ProTox‐3.0, and NeuroTDPi. A probability score > 0.5 was defined as indicating a high risk of toxicity. Strikingly, all platforms consistently predicted a high risk of neurotoxicity for both compounds (Figure 1A,B). For irinotecan, neurotoxicity scores reached 0.993 (ADMETlab 3.0), 0.870 (ProTox‐3.0), and 0.960 (NeuroTDPi). Corresponding scores for SN‐38 were 0.965, 0.750, and 0.934, all exceeding the high‐risk threshold. This multi‐platform consensus provided a strong in silico rationale to hypothesize that neurotoxicity is an intrinsic adverse effect of irinotecan pharmacotherapy. Predictions for other organ toxicities (e.g., respiratory and hepatic) showed compound‐ and platform‐dependent variability.

Figure 1.

Figure 1

Open in a new tab

Multi‐platform computational toxicity prediction for irinotecan (A) and its active metabolite SN‐38 (B).

3.2. Real‐World Pharmacovigilance Data

To validate and characterize the predicted neurotoxicity in a clinical context, we analyzed reports from the FAERS. After standard data curation (Figure 2), we identified 39,928 reports for C‐Irinotecan and 4313 for L‐Irinotecan. Over 83% of reports for both formulations were submitted by healthcare professionals, indicating the reliability of the dataset. Detailed demographic and reporting characteristics are provided in Table S2.

Figure 2.

Figure 2

Open in a new tab

Flow chart of data extraction for FAERS database analysis.

Disproportionality analysis at the System Organ Class (SOC) level initially indicated that “Nervous system disorders” was a significant signal for C‐Irinotecan (ranking 5th, 7.06% of its signals) but was absent from the top signals for L‐Irinotecan (Table S3). A granular analysis using MedDRA terminology uncovered a fundamental qualitative difference. For C‐Irinotecan, 43 nervous system disorder signals spanned 11 High‐Level Group Terms (HLGTs), encompassing both peripheral nervous system (PNS) (e.g., Peripheral neuropathies) and CNS disorders (e.g., Encephalopathies, CNS vascular disorders). In stark contrast, L‐Irinotecan was associated with only six nervous system signals, which mapped exclusively to peripheral neuropathies or non‐specific neurological disorders; no HLGT terms specific to the CNS were identified (Table S4). This clear dissociation indicates that while both formulations exhibit neurotoxicity signals, clinically reported CNS involvement is uniquely associated with the conventional formulation.

The convergence of computational prediction and real‐world observation presents a key mechanistic question. To address this, we integrated pharmacokinetic evidence. Literature consistently reports that after administration of C‐Irinotecan, the prototype drug irinotecan achieves measurable concentrations in the cerebrospinal fluid, whereas its metabolite SN‐38 demonstrates significantly lower CNS penetration [28]. Considering that the liposomal formulation is engineered to alter biodistribution and may further restrict CNS access, the absence of CNS signals with L‐Irinotecan provides compelling indirect evidence. Therefore, we propose that the parent compound, irinotecan, is a primary mediator of the CNS‐specific neurotoxicity profile observed with the conventional formulation. This hypothesis guided our subsequent mechanistic investigation into the molecular targets of irinotecan itself.

3.3. Network Pharmacology Identifies Core Molecular Targets for IIN With Nervous System Expression Validation

To elucidate the molecular basis of IIN, we employed a network pharmacology approach. Potential targets of irinotecan were retrieved from PharmMapper and SwissTargetPrediction, yielding 378 unique targets. Neurotoxicity‐related targets were sourced from CTD, OMIM, and GeneCards, resulting in 2953 unique targets. A Venn diagram analysis identified 162 targets at the intersection, representing putative mediators of IIN (Figure 3A). A compound‐target‐disease network visualized these interactions (Figure 3B).

Figure 3.

Figure 3

Open in a new tab

Identification of irinotecan‐induced neurotoxicity (IIN) targets and Irinotecan‐targets‐neurotoxicity network construction: (A) Venn diagram showing the number of unique targets of Irinotecan (n = 378) and neurotoxicity (n = 2953), as well as overlapping targets (n = 162). (B) The network Irinotecan‐targets‐neurotoxicity.

The 162 overlapping targets were imported into the STRING database to construct a PPI network, which was then analyzed and visualized using Cytoscape software (version 3.7.2). The resulting PPI network contained 161 nodes and 624 edges. In the visualization (Figure 4A). Top 10 genes were separately screened by four topological algorithms (EPC, Degree, MNC, MCC) using the CytoHubba plugin embedded in Cytoscape software(Figure 4B). Afterward, the overlapping genes across the four algorithm‐derived gene sets were visualized and extracted via a Venn diagram, leading to the identification of 8 core targets. (ESR1, STAT1, AKT1, PIK3CA, ERBB2, EGFR, HSP90AB1, SRC) were finally screened out (Figure 4C). These genes are likely to be the core genes associated with IIN. To further verify the nervous system specificity of the eight core targets, we analyzed their expression in normal human brain tissues, since the cerebral cortex and hippocampus are core regions responsible for neural function, neuroinflammation, and chemotherapy‐induced neurotoxicity. Their expression profiles were validated using HPA, GTEx, and FANTOM5 databases. Results showed that AKT1, EGFR, PIK3CA, SRC, and STAT1 were highly expressed in the cerebral cortex and hippocampus, which are closely associated with neural function and neuroinflammation. ERBB2, ESR1, and HSP90AB1 exhibited moderate‐to‐high expression in the cerebral cortex and hippocampus, which are also highly relevant to nervous system function. All eight targets were stably expressed in the normal CNS, supporting their potential biological relevance to IIN (Figure S1).

Figure 4.

Figure 4

Open in a new tab

PPI Network Construction: (A) Protein–protein interaction (PPI) analysis of the 162 overlapping targets was conducted using the STRING database and visualized in Cytoscape. (B) Top 10 hub genes from each of the four topological algorithms (EPC, Degree, MNC, MCC). Node color reflects the hub score of each gene, with red indicating the highest importance. (C) The overlapping targets of four algorithms by Venn diagram.

3.4. Molecular Docking Validates High‐Affinity Binding of Irinotecan to Core Targets

To investigate the direct interaction between irinotecan and the identified core targets, molecular docking was performed. The results confirmed thermodynamically favorable binding between irinotecan and all eight core proteins. The calculated binding energies were: ESR1 (−9.4 kcal/mol), STAT1 (−8.4 kcal/mol), AKT1 (−7.3 kcal/mol), PIK3CA (−11.8 kcal/mol), ERBB2 (−7.6 kcal/mol), EGFR (−10.7 kcal/mol), HSP90AB1 (−9.6 kcal/mol), and SRC (−8.4 kcal/mol) (Figure 5). All binding energies were substantially lower than the −5.0 kcal/mol threshold indicative of spontaneous binding, confirming strong and stable interactions between irinotecan and these key neurotoxicity‐associated proteins.

Figure 5.

Figure 5

Open in a new tab

Molecular docking of irinotecan with core targets.

3.5. Pathway Enrichment Analysis Highlights the MAPK Signaling Pathway as a Central Hub

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed on the 162 common targets. Biological processes were primarily enriched in “cellular response to nitrogen compound,” “regulation of MAPK cascade,” and “response to wounding.” Cellular components included “receptor complex” and “postsynapse.” Molecular functions were enriched for “protein tyrosine kinase activity” and “kinase binding” (Figure 6A).

Figure 6.

Figure 6

Open in a new tab

Enrichment analysis and Irinotecan‐target‐neurotoxicity‐pathway network construction: (A) Gene Ontology (GO) enrichment analysis. (B) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. (C) Irinotecan‐target‐neurotoxicity‐pathway network.

KEGG pathway analysis of the top 20 enriched pathways revealed involvement in critical processes such as “Pathways in cancer,” “Neuroactive ligand‐receptor interaction,” and “Transcriptional misregulation in cancer.” Most notably, the “MAPK signaling pathway” was among the most significantly enriched pathways (Figure 6B). An integrated Irinotecan‐Targets‐Neurotoxicity‐Pathway network was constructed to visualize these complex interactions, placing the MAPK cascade as a central signaling hub potentially downstream of irinotecan‐target binding (Figure 6C).

4. Discussion

This integrated study bridges computational prediction, large‐scale pharmacovigilance, and mechanistic modeling to advance the understanding of IIN. Moving beyond the established metabolite‐centric paradigm, we present a pivotal clinical observation: conventional and liposomal irinotecan exhibit distinct neurotoxicity profiles, with CNS involvement being uniquely associated with the conventional formulation. Converging this finding with pharmacokinetic evidence, we propose a central hypothesis: the parent drug, irinotecan itself, is a primary mediator of this formulation‐specific CNS toxicity. The following discussion synthesizes the multi‐dimensional evidence supporting this hypothesis, explores its potential molecular underpinnings, and considers its clinical implications.

FAERS database analysis is a key tool for head‐to‐head comparison of the AE profiles among formulations with the same active ingredient [29]. Evidence from real‐world studies comparing AEs between novel drug delivery systems (e.g., liposomes) and conventional formulations has been accumulating. The formulation‐dependent dichotomy in CNS toxicity points to the parent drug as a key culprit. Our analysis confirms neurotoxicity as a significant signal for both irinotecan formulations, aligning with consistent predictions from multiple computational platforms. However, the real‐world data reveal a critical qualitative divergence: C‐Irinotecan is linked to a broad spectrum of CNS disorders (e.g., encephalopathies, CNS vascular disorders), while L‐Irinotecan shows no such specific CNS signals. This makes the analysis of such data increasingly viable. This discrepancy cannot be explained by the shared metabolic pathway to SN‐38 and instead directs attention to the physicochemical and pharmacokinetic differences between the formulations. The smaller molecular size of free irinotecan facilitates its passage across the blood‐brain barrier, a process evidenced by studies detecting irinotecan, but not SN‐38, in the cerebrospinal fluid of non‐human primates [28]. The ability of exogenous substances to cross the blood‐brain barrier and induce neural lesions has been well documented in previous studies [30, 31]. In contrast, the large, stable liposome encapsulating L‐Irinotecan effectively restricts its CNS distribution [32]. This fundamental difference in biodistribution provides a direct mechanistic link: C‐Irinotecan delivers the parent compound to the CNS, while L‐Irinotecan does so minimally. Supporting this, in vitro evidence demonstrates that irinotecan, but not SN‐38, can directly inhibit acetylcholinesterase activity [33], indicating an intrinsic neuroactive potential of the parent molecule. Therefore, the observed clinical profile of CNS toxicity being exclusive to C‐Irinotecan is most parsimoniously explained by differential CNS exposure to irinotecan, which positions the parent drug as a direct neurotoxicant. Supporting this parent drug‐centric view, published studies have shown that irinotecan primarily causes central neurotoxicity, whereas peripheral neurotoxicity is rare [34]. Conversely, while an animal study reported metabolic changes in mouse brain tissue after high‐dose SN‐38 treatment [35], it only examined central brain tissue without assessing peripheral nerve samples, and the observed metabolic alterations remain insufficient to confirm definitive neurotoxic damage. Thus, whether SN‐38 induces peripheral neurotoxicity or exerts synergistic neurotoxic effects with irinotecan remains unclear, and relevant research is currently lacking. This uncertainty further justifies shifting the mechanistic focus from the metabolite to the parent compound—a direction we explore next through network pharmacology and molecular docking.

Network pharmacology and molecular docking delineate a plausible molecular network for irinotecan‐mediated neurotoxicity. To explore how irinotecan might exert direct neurotoxic effects within the CNS, we shifted focus from its metabolite to the parent compound's potential targets. Our analysis identified eight core proteins (AKT1, EGFR, SRC, ESR1, HSP90AB1, PIK3CA, ERBB2, STAT1) with high‐affinity binding to irinotecan. Crucially, this target set is enriched for proteins with established roles in CNS homeostasis and pathology. For instance, EGFR and SRC are key regulators of neuroinflammation and neuronal survival [36, 37]; ESR1 exerts neuroprotective effects, and its dysfunction promotes neuroinflammation [38]; STAT1 activation in microglia drives aberrant synaptic pruning and neurodevelopmental deficits [39]. The strong computational binding of irinotecan to these targets, particularly those like AKT1 and STAT1 implicated in glial activation and neuronal apoptosis, suggests a direct pathway through which irinotecan could disrupt neural cell function, independent of SN‐38.

The MAPK signaling pathway, particularly the p38 cascade, emerges as a central downstream effector. Enrichment analysis positioned the MAPK pathway as a key hub connecting the identified targets. Within this family, the p38 MAPK cascade is of particular relevance due to its well‐documented role as a stress‐sensitive mediator of neuronal apoptosis, neuroinflammation, and glial cell damage [40, 41, 42, 43]. Significantly, previous experimental studies have indicated that irinotecan can regulate p38 MAPK phosphorylation in a concentration‐ and time‐dependent manner. Given the well‐established role of p38 MAPK in mediating neuroinflammation and neuronal injury, our computational results suggest that the MAPK/p38 pathway may be critically involved in irinotecan‐induced neurotoxicity. This allows us to construct a coherent mechanistic model: upon reaching the CNS, irinotecan binds to and perturbs specific cellular targets (such as those identified herein), leading to the dysregulation of the p38 MAPK signaling cascade. This dysregulation, in turn, could trigger a cascade of events including microglial activation, pro‐inflammatory cytokine release, and ultimately, neuronal injury or death—hallmarks of the CNS toxicities observed in our FAERS analysis. Neuronal degeneration and myelin sheath damage are recognized as critical pathological features of neurotoxicity in various nervous system disorders [44, 45]. This model provides a specific molecular pathway through which irinotecan, distinct from SN‐38, could mediate its observed CNS‐selective toxicity.

This study has limitations inherent to pharmacovigilance and retrospective analysis. First, FAERS database reliance brings biases (e.g., underreported mild/moderate AEs), precluding causal links between irinotecan and neurotoxicity or AE incidence calculation. Second, C‐irinotecan (180 mg/m2) and L‐irinotecan (70 mg/m2) differ in indications (detailed in Table S5) and dosages. However, the disproportionality analysis of the FAERS database is inherently suitable for comparing safety signals across different drug formulations. This method has been widely adopted in numerous published pharmacovigilance studies, even when formulations differ in dosage and indications [46]. Our study identified qualitative differences: CNS toxicity was only associated with conventional irinotecan, rather than mere numerical disparities. Third, irinotecan's common use with other chemotherapeutics (e.g., 5‐fluorouracil, oxaliplatin) complicates AE attribution. Only reports naming it as the primary suspect were included. Finally, this study relies on computational and pharmacovigilance analyses without independent experimental validation. Although relevant published studies provide supportive evidence for the potential roles of blood–brain barrier penetration and p38 MAPK signaling, definitive confirmation of core targets and pathway activation still requires further in vitro and in vivo experiments. Future studies using cellular or animal models are warranted to verify these preliminary computational findings.

5. Conclusions

This study establishes a critical distinction in the neurotoxicity profiles of irinotecan formulations: conventional irinotecan is associated with combined central and PNS adverse events, while its liposomal counterpart demonstrates a predominantly peripheral toxicity pattern. Using real‐world pharmacovigilance data, pharmacokinetic evidence, and computational modeling, this study proposes a mechanistic hypothesis in which the parent drug irinotecan may directly contribute to neurotoxicity. Its differential access to the CNS could help explain the distinct neurotoxic profiles of different formulations. This hypothesis is further supported by the identification of high‐affinity neural targets (e.g., AKT1, EGFR) for irinotecan and the delineation of the MAPK/p38 pathway as a plausible downstream effector. Our findings refine the prevailing metabolite‐centric paradigm of irinotecan's toxicity and provide a translational framework for optimizing formulation safety and developing targeted neuroprotective strategies.

Author Contributions

Cheng Shen: conceptualization, methodology, data curation, software, investigation, funding acquisition, project administration, writing – original draft, writing – review and editing, validation, formal analysis. Wei Ma: conceptualization, methodology, data curation, writing – original draft, formal analysis, writing – review and editing, investigation. Jing Lu: conceptualization, methodology, data curation, investigation, writing – original draft, writing – review and editing, formal analysis. Yongbo Tang: investigation, conceptualization, writing – review and editing, methodology, data curation. Wei Cao: conceptualization, methodology, writing – review and editing, data curation, formal analysis.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1

JBT-40-e71026-s006.pdf (696.4KB, pdf)

Supporting File 2

JBT-40-e71026-s001.docx (14KB, docx)

Supporting File 3

JBT-40-e71026-s003.docx (15KB, docx)

Supporting File 4

JBT-40-e71026-s004.docx (17.1KB, docx)

Supporting File 5

JBT-40-e71026-s002.xlsx (17.1KB, xlsx)

Supporting File 6

JBT-40-e71026-s005.xlsx (23.1KB, xlsx)

Acknowledgements

This study was performed using data from FAERS, provided by the FDA. The information, results, or interpretation of the current study does not represent any opinion of the FDA. This research was funded by the Natural Science Foundation of Hunan, grant number 2023JJ40342.

Data Availability Statement

The data that support the findings of this study are available in FAERS database at https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html. These data were derived from the following resources available in the public domain: ‐ FAERS database, https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

  • 1. Bukowski K., Kciuk M., and Kontek R., “Mechanisms of Multidrug Resistance in Cancer Chemotherapy,” International Journal of Molecular Sciences 21, no. 9 (May 2020): 3233, 10.3390/ijms21093233. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Bailly C., “Irinotecan: 25 Years of Cancer Treatment,” Pharmacological Research 148 (October 2019): 104398, 10.1016/j.phrs.2019.104398. [DOI] [PubMed] [Google Scholar]
  • 3. Pérez‐Herrero E. and Fernández‐Medarde A., “Advanced Targeted Therapies in Cancer: Drug Nanocarriers, the Future of Chemotherapy,” European Journal of Pharmaceutics and Biopharmaceutics 93 (June 2015): 52–79, 10.1016/j.ejpb.2015.03.018. [DOI] [PubMed] [Google Scholar]
  • 4. Nussbaumer S., Bonnabry P., Veuthey J. L., and Fleury‐Souverain S., “Analysis of Anticancer Drugs: A Review,” Talanta 85, no. 5 (October 2011): 2265–2289, 10.1016/j.talanta.2011.08.034. [DOI] [PubMed] [Google Scholar]
  • 5. Conroy T., Bosset J. F., Etienne P. L., et al., “Neoadjuvant Chemotherapy with FOLFIRINOX and Preoperative Chemoradiotherapy for Patients with Locally Advanced Rectal Cancer (UNICANCER‐PRODIGE 23): a Multicentre, Randomised, Open‐Label, Phase 3 Trial,” The lancet oncology 22, no. 5 (May 2021): 702–715, 10.1016/S1470-2045(21)00079-6. [DOI] [PubMed] [Google Scholar]
  • 6. Loupakis F., Cremolini C., Masi G., et al., “Initial Therapy with FOLFOXIRI and Bevacizumab for Metastatic Colorectal Cancer,” New England Journal of Medicine 371, no. 17 (October 2014): 1609–1618, 10.1056/NEJMoa1403108. [DOI] [PubMed] [Google Scholar]
  • 7. Milano G., Innocenti F., and Minami H., “Liposomal Irinotecan (Onivyde): Exemplifying the Benefits of Nanotherapeutic Drugs,” Cancer science 113, no. 7 (July 2022): 2224–2231, 10.1111/cas.15377. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Zhu X., Huang Y., Ding J., Liu J., Cui C., and Han G, “Investigating the Impact of SN‐38 on Mouse Brain Metabolism Based on Metabolomics,” Drug Design, Development and Therapy 18 (2024): 2435–2447, 10.2147/DDDT.S457698. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Gagné J. F., Montminy V., Belanger P., Journault K., Gaucher G., and Guillemette C., “Common Human UGT1A Polymorphisms and the Altered Metabolism of Irinotecan Active Metabolite 7‐ethyl‐10‐hydroxycamptothecin (SN‐38),” Molecular Pharmacology 62, no. 3 (September 2002): 608–617, 10.1124/mol.62.3.608. [DOI] [PubMed] [Google Scholar]
  • 10. Morris R., Ali R., and Cheng F., “Drug Repurposing Using FDA Adverse Event Reporting System (FAERS) Database,” Current Drug Targets 25, no. 7 (2024): 454–464, 10.2174/0113894501290296240327081624. [DOI] [PubMed] [Google Scholar]
  • 11. He J., Zhu X., Xu K., Li Y., and Zhou J., “Network Toxicological and Molecular Docking to Investigate the Mechanisms of Toxicity of Agricultural Chemical Thiabendazole,” Chemosphere 363 (September 2024): 142711, 10.1016/j.chemosphere.2024.142711. [DOI] [PubMed] [Google Scholar]
  • 12. Banerjee P., Eckert A. O., Schrey A. K., and Preissner R., “Protox‐Ii: A Webserver for the Prediction of Toxicity of Chemicals,” Nucleic Acids Research 46, no. W1 (July 2018): W257–W263, 10.1093/nar/gky318. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Liu W. H., Hu H. M., Li C., et al., “Real‐World Study of Adverse Events Associated With Triptan Use in Migraine Treatment Based on the US Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) Database,” The Journal of Headache and Pain 25, no. 1 (November 2024): 206, 10.1186/s10194-024-01913-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Noguchi Y., Tachi T., and Teramachi H., “Detection Algorithms and Attentive Points of Safety Signal Using Spontaneous Reporting Systems as a Clinical Data Source,” Briefings in Bioinformatics 22, no. 6 (November 2021): bbab347, 10.1093/bib/bbab347. [DOI] [PubMed] [Google Scholar]
  • 15. Kim S., Chen J., Cheng T., et al., “PubChem 2025 Update,” Nucleic Acids Research 53, no. D1 (January 2025): D1516–D1525, 10.1093/nar/gkae1059. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Daina A. and Zoete V., “Testing the Predictive Power of Reverse Screening to Infer Drug Targets, with the Help of Machine Learning,” Communications Chemistry 7, no. 1 (May 2024): 105, 10.1038/s42004-024-01179-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Wang X., Shen Y., Wang S., et al., “PharmMapper 2017 Update: A Web Server for Potential Drug Target Identification with a Comprehensive Target Pharmacophore Database,” Nucleic Acids Research 45, no. W1 (July 2017): W356–W360, 10.1093/nar/gkx374. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Davis A. P., Grondin C. J., Lennon‐Hopkins K., et al., “The Comparative Toxicogenomics Database's 10th Year Anniversary: Update 2015,” Nucleic Acids Research 43, no. Database issue (January 2015): D914–D920, 10.1093/nar/gku935. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Rappaport N., Twik M., Plaschkes I., et al., “MalaCards: An Amalgamated Human Disease Compendium With Diverse Clinical and Genetic Annotation and Structured Search,” Nucleic Acids Research 45, no. D1 (January 2017): D877–D887, 10.1093/nar/gkw1012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Amberger J. S., Bocchini C. A., Schiettecatte F., Scott A. F., and Hamosh A., “OMIM.org: Online Mendelian Inheritance in Man (OMIM), an Online Catalog of Human Genes and Genetic Disorders,” Nucleic Acids Research 43, no. Database issue (January 2015): D789–D798, 10.1093/nar/gku1205. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Heberle H., Meirelles G. V., da Silva F. R., Telles G. P., and Minghim R., “Interactivenn: A Web‐Based Tool for the Analysis of Sets Through Venn Diagrams,” BMC Bioinformatics 16, no. 1 (May 2015): 169, 10.1186/s12859-015-0611-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Szklarczyk D., Gable A. L., Lyon D., et al., “STRING v11: Protein‐Protein Association Networks with Increased Coverage, Supporting Functional Discovery in Genome‐Wide Experimental Datasets,” Nucleic Acids Research 47, no. D1 (January 2019): D607–D613, 10.1093/nar/gky1131. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Shannon P., Markiel A., Ozier O., et al., “Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks,” Genome Research 13, no. 11 (November 2003): 2498–2504, 10.1101/gr.1239303. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Liu Y., Yang X., Gan J., Chen S., Xiao Z. X., and Cao Y., “CB‐Dock2: Improved Protein‐Ligand Blind Docking by Integrating Cavity Detection, Docking and Homologous Template Fitting,” Nucleic Acids Research 50, no. W1 (July 2022): W159–W164, 10.1093/nar/gkac394. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Zhou Y., Zhou B., Pache L., et al., “Metascape Provides a Biologist‐Oriented Resource for the Analysis of Systems‐Level Datasets,” Nature Communications 10, no. 1 (April 2019): 1523, 10.1038/s41467-019-09234-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Kanehisa M., Sato Y., Kawashima M., Furumichi M., and Tanabe M., “Kegg as a Reference Resource for Gene and Protein Annotation,” Nucleic Acids Research 44, no. D1 (January 2016): D457–D462, 10.1093/nar/gkv1070. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Ogata H., Goto S., Sato K., Fujibuchi W., Bono H., and Kanehisa M., “KEGG: Kyoto Encyclopedia of Genes and Genomes,” Nucleic Acids Research 27, no. 1 (January 1999): 29–34, 10.1093/nar/27.1.29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Blaney S. M., Takimoto C., Murry D. J., et al., “Plasma and Cerebrospinal Fluid Pharmacokinetics of 9‐aminocamptothecin (9‐AC), Irinotecan (CPT‐11), and SN‐38 in Nonhuman Primates,” Cancer Chemotherapy and Pharmacology 41, no. 6 (1998): 464–468, 10.1007/s002800050768. [DOI] [PubMed] [Google Scholar]
  • 29. Su H., Jia J., Mao Y., Zhu R., and Li Z., “A Real‐World Analysis of FDA Adverse Event Reporting System (FAERS) Events for Liposomal and Conventional Doxorubicins,” Scientific Reports 14, no. 1 (March 2024): 5095, 10.1038/s41598-024-55185-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Ding J., Sun B., Gao Y., et al., “APOE4 Exacerbates Cerebral Tau Pathology Through Cholesterol‐Induced Degradation of Phosphatase in Atherosclerosis,” CNS Neuroscience & Therapeutics 31, no. 8 (August 2025): e70536, 10.1111/cns.70536. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Ding J., Sun B., Gao Y., et al., “Evidence for Chromium Crosses Blood Brain Barrier from the Hypothalamus in Chromium Mice Model,” Ecotoxicology and Environmental Safety 273 (March 2024): 116179, 10.1016/j.ecoenv.2024.116179. [DOI] [PubMed] [Google Scholar]
  • 32. Maeda H., “Tumor‐Selective Delivery of Macromolecular Drugs via the EPR Effect: Background and Future Prospects,” Bioconjugate Chemistry 21, no. 5 (May 2010): 797–802, 10.1021/bc100070g. [DOI] [PubMed] [Google Scholar]
  • 33. Blandizzi C., Danesi R., De Paolis B., et al., “Cholinergic Toxic Syndrome by the Anticancer Drug Irinotecan: Acetylcholinesterase Does Not Play a Major Role,” Clinical Pharmacology & Therapeutics 71, no. 4 (April 2002): 263–271, 10.1067/mcp.2002.121909. [DOI] [PubMed] [Google Scholar]
  • 34. Was H., Borkowska A., Bagues A., et al., “Mechanisms of Chemotherapy‐Induced Neurotoxicity,” Frontiers in Pharmacology 13 (2022): 750507, 10.3389/fphar.2022.750507. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Zhu X., Huang Y., Ding J., Liu J., Cui C., and Han G., “Investigating the Impact of SN‐38 on Mouse Brain Metabolism Based on Metabolomics,” Drug Design, Development and Therapy 18 (2024): 2435–2447. 10. 2147/DDDT.S457698. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Huang H. J., Tseng Y. J., Lee I. J., Lo Y. L., and Lin A. M. Y., “Involvement of EGFR‐AKT Signaling in Hemin‐Induced Neurotoxicity,” Experimental Biology and Medicine 250 (2025): 10554, 10.3389/ebm.2025.10554. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Iqbal Hossain M., Hoque A., Lessene G., et al., “Dual Role of Src Kinase in Governing Neuronal Survival,” Brain Research 1594 (January 2015): 1–14, 10.1016/j.brainres.2014.10.040. [DOI] [PubMed] [Google Scholar]
  • 38. Liu J., Yuan S., Niu X., Kelleher R., and Sheridan H., “ESR1 Dysfunction Triggers Neuroinflammation as a Critical Upstream Causative Factor of the Alzheimer's Disease Process,” Aging 14, no. 21 (November 2022): 8595–8614, 10.18632/aging.204359. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Zhong H., Xue R., Han Y., et al., “S‐Ketamine Exposure in Early Postnatal Period Induces Social Deficit Mediated by Excessive Microglial Synaptic Pruning,” Molecular Psychiatry 30, no. 8 (August 2025): 3615–3631, 10.1038/s41380-025-02949-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Chen X., Ning Y., Wang B., et al., “HET0016 Inhibits Neuronal Pyroptosis in the Immature Brain Post‐TBI via the p38 MAPK Signaling Pathway,” Neuropharmacology 239 (November 2023): 109687, 10.1016/j.neuropharm.2023.109687. [DOI] [PubMed] [Google Scholar]
  • 41. Zhu Z., Ge M., Li C., et al., “Effects of p38 MAPK Signaling Pathway on Cognitive Function and Recovery of Neuronal Function After Hypoxic‐Ischemic Brain Injury in Newborn Rats,” Journal of Clinical Neuroscience 78 (August 2020): 365–370, 10.1016/j.jocn.2020.04.085. [DOI] [PubMed] [Google Scholar]
  • 42. Wang C., Li Y., Yi Y., et al., “Hippocampal microRNA‐26a‐3p Deficit Contributes to Neuroinflammation and Behavioral Disorders via p38 MAPK Signaling Pathway in Rats,” Journal of Neuroinflammation 19, no. 1 (November 2022): 283, 10.1186/s12974-022-02645-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43. Wang S., Lin Q., Wang Z., and Pan X., “Ropivacaine Induces Neurotoxicity by Activating MAPK/p38 Signal to Upregulate Fas Expression in Neurogliocyte,” Neuroscience Letters 706 (July 2019): 7–11, 10.1016/j.neulet.2019.04.048. [DOI] [PubMed] [Google Scholar]
  • 44. Ding J., Huang J., Xia B., et al., “Transfer of α‐synuclein from Neurons to Oligodendrocytes Triggers Myelin Sheath Destruction in Methamphetamine Administration Mice,” Toxicology Letters 352 (November 2021): 34–45, 10.1016/j.toxlet.2021.09.005. [DOI] [PubMed] [Google Scholar]
  • 45. Ding J., Wu J., Hou X., et al., “α‐synuclein‐lack Expression Rescues Methamphetamine‐Induced Mossy Fiber Degeneration in Dorsal Hippocampal CA3,” Neurotoxicology 101 (March 2024): 36–45, 10.1016/j.neuro.2024.01.005. [DOI] [PubMed] [Google Scholar]
  • 46. Lu J., Shen C., Ma W., Lin Z., and Gao W., “FAERS Based Disproportionality Analysis and Network Pharmacology Investigation of Doxorubicin Associated Drug Induced Cardiotoxicity,” Scientific Reports 15, no. 1 (September 2025): 33278, 10.1038/s41598-025-18383-2. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting File 1

JBT-40-e71026-s006.pdf (696.4KB, pdf)

Supporting File 2

JBT-40-e71026-s001.docx (14KB, docx)

Supporting File 3

JBT-40-e71026-s003.docx (15KB, docx)

Supporting File 4

JBT-40-e71026-s004.docx (17.1KB, docx)

Supporting File 5

JBT-40-e71026-s002.xlsx (17.1KB, xlsx)

Supporting File 6

JBT-40-e71026-s005.xlsx (23.1KB, xlsx)

Data Availability Statement

The data that support the findings of this study are available in FAERS database at https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html. These data were derived from the following resources available in the public domain: ‐ FAERS database, https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html The data that support the findings of this study are available from the corresponding author upon reasonable request.

Articles from Journal of Biochemical and Molecular Toxicology are provided here courtesy of Wiley

RESOURCES