Introduction

Doxorubicin (DOX) is an anthracycline antitumor drug. Since its approval by the U.S. Food and Drug Administration (FDA) in 1974, it has been widely used as a first-line treatment, either alone or in combination with other agents, for various solid and metastatic tumors1. However, its clinical application is limited by its high hydrophilicity, short half-life, low bioavailability, and large volume of distribution. These properties necessitate high doses, which can lead to severe side effects such as cardiotoxicity, extravasation, nephrotoxicity, and myelosuppression2.

DOX has been widely reported to induce dose-dependent, progressive, and potentially lethal myocardial damage. A large-scale study involving 4,018 patients reported an overall incidence of DOX-induced congestive heart failure of 2.2%3. Another retrospective study further established high cumulative dose and very young or old age as primary risk factors for this cardiotoxicity4. Furthermore, a prospective evaluation of DOX cardiotoxicity demonstrated a clear dose–response relationship: the incidence of Grade ≥ 3 cardiac dysfunction was 2% at a cumulative dose of < 450 mg/m2, 3% at 450- < 600 mg/m2, and 1.1% at ≥ 600 mg/m25. To enhance therapeutic efficacy and reduce organ toxicity, novel drug delivery systems (DDS) such as liposomes have been developed. These systems are designed to increase drug accumulation in tumor cells while minimizing exposure in normal tissues. Clinically available liposomal doxorubicin (Lip-DOX) formulations are broadly categorized into pegylated and non-pegylated types.

While Lip-DOX is designed to reduce drug exposure in normal tissues, the fundamental mechanism of DOX-induced cardiotoxicity (DIC) remains dose-dependent. Therefore, a comparative investigation into the specific characteristics of cardiotoxicity induced by conventional doxorubicin (Con-DOX) and Lip-DOX is warranted to more effectively guide the rational clinical application of these chemotherapeutic agents. The U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) offers a unique resource for investigating these associations at scale, enabling a comprehensive analysis of real-world, post-marketing drug safety data6. Although Fukuda et al. conducted a data mining analysis of this database to systematically compare the general adverse event profiles between the conventional and liposomal formulations of DOX7, a systematic exploration of their differences in cardiotoxicity was not performed. This study aims to address that gap.

Although previous studies have indicated that DOX may induce cardiotoxicity, systematic analyses focusing on the interactions between DOX and cardiotoxicity-related genes remain scarce. Furthermore, numerous studies have investigated the mechanisms of DIC. Kong et al. reviewed the mechanisms of DIC, focusing on free radical generation and related cell death pathways8. Other studies have shown that DIC also involves mitochondrial dysfunction, iron regulation disturbance, Ca2⁺ imbalance, autophagy dysregulation, nitric oxide release, inflammation, and apoptotic gene/protein alterations. Additionally, DOX downregulates DNA methyltransferase 1 (DNMT1) enzymatic activity, impairing DNA methylation3. Despite these advances, the complete mechanistic picture of DIC remains unclear, as most preclinical studies focus on isolated pathways rather than the disease’s multifaceted pathogenesis.Large-scale data analysis strategies, such as network pharmacology, offer a promising approach for comprehensively predicting drug action mechanisms and adverse reaction pathways. For example, Cheng et al. successfully used this method to investigate taxane-induced liver injury9.

In the present study, we leveraged the FAERS database to identify and compare cardiotoxicity events associated with Con-DOX and Lip-DOX formulations. Building on this pharmacovigilance analysis, we utilized network pharmacology to further explore the potential multifactorial mechanisms of DIC. This integrated approach was used to identify key molecular targets and mechanisms. Ultimately, our findings provide critical safety evidence to guide the clinical use of DOX, enhance the understanding of its cardiotoxic mechanisms, and improve medication safety for cancer patients.

Results

Descriptive analysis

A total of 56,321,150 adverse event (AE) reports were collected from the FAERS database from the first quarter of 2004 to the first quarter of 2025. From this dataset, 100,936 AE reports identified DOX as the primary suspected (PS) drug, involving 32,960 patients. These reports originated from six continents (excluding Antarctica), with the majority Submitted from North America, Europe, and Asia. Females accounted for a higher proportion of patients than Males. Regarding age distribution, the Majority of patients were aged 45 to 64 years, while those aged 65 years and older accounted for 20% of reports. Physicians submitted the highest number of reports (41.35%), whereas the proportion submitted by consumers was relatively low (5.16%). Geographically, the most reports were submitted from the U.S. The detailed characteristics of all reports are shown in Supplementary Table S1.

To specifically investigate the cardiotoxicity of the two DOX formulations, we extracted reports from the U.S. to ensure consistent and complete reporting information. This approach facilitated accurate drug identification and minimized potential bias. This analysis included 10,695 AE reports involving 4,186 patients treated with Con-DOX and 13,164 AE reports involving 3,941 patients treated with Lip-DOX. The detailed characteristics of these patients are shown in Table 1. For Con-DOX reports, the median patient age was 55 years. Females accounted for 49.9% of reports, compared to 31.9% for males (with the remainder unspecified). For Lip-DOX reports, the median patient age was 56 years. Females accounted for 42.63% of reports, compared to 21.25% for males (with the remainder unspecified). The reporting sources differed between formulations: Con-DOX reports were primarily submitted by pharmacists (35.28%) and physicians (37.30%), while Lip-DOX reports were predominantly submitted by physicians (49.30%). The median age was comparable between the two groups. With health professionals (e.g., pharmacists, physicians, and other health professionals) accounting for over 90% of submissions, the reports can be considered high quality.

Table 1 Demographic and reporting characteristics of Con-DOX and Lip-DOX AE reports in FAERS. (Con-DOX, conventional doxorubicin; Lip-DOX, liposomal doxorubicin; AE, adverse event).
Full size table

Signal detection associated with doxorubicin in the FAERS database

In this study, a total of 100,936 doxorubicin-related AE reports were included in the disproportionality analysis. Using the reporting odds ratio (ROR) method, we identified 1,272 positive signals covering 26 System Organ Classes (SOCs) after excluding the ‘Social circumstances’ SOC. Preferred Terms (PTs) with positive signals were ranked by ROR value. The top 30 PTs, based on the frequency of reports, are listed in Fig. 1; these signals spanned 12 SOCs. Among the top 30 PTs with the highest signal strength, the five PTs with the highest ROR values were ‘cardiotoxicity’ (ROR = 64.68), ‘febrile neutropenia’ (ROR = 33.29), ‘cardiomyopathy’ (ROR = 20.29), ‘mucosal inflammation’ (ROR = 19.48), and ‘ejection fraction decreased’ (ROR = 18.83). At the SOC level, the number of reports in the ‘Cardiac disorders’ SOC was Substantial, accounting for 6,199 reports (6.14%). This finding is consistent with the results reported in previous literature10.

Fig. 1
figure 1

The top 30 Preferred Terms (PTs) by reporting frequency of positive signals for doxorubicin.

Full size image

Signals of cardiotoxicity associated with Con-DOX and Lip-DOX in the FAERS database

AE signal detection revealed the distribution and number of signals for Con-DOX and Lip-DOX at the SOC level; the specific results are presented in Table 2. For Con-DOX, relevant signals were detected in 26 out of 27 SOCs, excluding ‘Social circumstances’. In contrast, no signals were detected for Lip-DOX in three SOCs: ‘Eye disorders’, ‘Psychiatric disorders’, and ‘Social circumstances’. In total, Con-DOX generated 429 positive signals, compared with 333 for Lip-DOX. Among all SOCs, ‘Infections and infestations’, ‘Neoplasms benign, malignant and unspecified (incl cysts and polyps)’, and ‘Cardiac disorders’ consistently ranked among the top three in terms of signal quantity for both formulations. For Con-DOX, the number (and proportion) of signals in these three SOCs was 77 (17.95%), 44 (10.26%), and 40 (9.32%), respectively. For Lip-DOX, the corresponding figures were 58 (17.42%), 59 (17.72%), and 27 (8.11%).

Table 2 Number and proportion of signals for Con-DOX and Lip-DOX at the SOC level in FAERS. (SOC, System Organ Classes; Con-DOX, conventional doxorubicin; Lip-DOX, liposomal doxorubicin.).
Full size table

Within the ‘Cardiac disorders’ SOC, Con-DOX reports included 307 females, 147 Males, and 103 reports with unrecorded gender. For Lip-DOX, the corresponding numbers were 204 females, 72 Males, and 156 reports with unrecorded gender. We conducted chi-square tests to explore the association between gender and the reporting of cardiac disorders. The analysis compared the proportion of male and female patients with cardiac disorder AEs to the gender distribution in all other AE reports for each drug. The results showed a significant association for both Con-DOX (χ2 = 9.82, p = 0.002) and Lip-DOX (χ2 = 7.73, p = 0.005). The odds ratios (OR) for females were 1.40 (95% CI: 1.13–1.72) for Con-DOX and 1.49 (95% CI: 1.12–1.98) for Lip-DOX.

Consistent with previous studies indicating that Lip-DOX is associated with fewer cardiac signals than Con-DOX, we further investigated the differences between the two formulations using Standardized MedDRA Queries (SMQ) related to cardiotoxicity. The results across 12 selected cardiac-related SMQs are presented in Table 3. Con-DOX showed positive signals in 10 SMQs, whereas Lip-DOX generated positive signals in only 2: ‘cardiomyopathy’ (SMQ code: 20,000,150) and ‘cardiac failure’ (SMQ code: 20,000,004). The ROR for cardiomyopathy was 34.07 (95% CI: 30.64–37.88) for Con-DOX and 18.39 (95% CI: 16.18–20.90) for Lip-DOX. The ROR for cardiac failure was 5.88 (95% CI: 5.22–6.63) for Con-DOX and 3.80 (95% CI: 3.33–4.34) for Lip-DOX. Notably, Con-DOX was associated with positive signals for cardiac arrhythmias and non-infectious myocarditis and pericarditis, for which no signals were detected for Lip-DOX. Overall, Lip-DOX demonstrated a narrower spectrum of cardiac-related AEs and lower signal strength compared to Con-DOX.

Table 3 Number of signals and ROR for Con-DOX and Lip-DOX at the SMQ level in FAERS. (ROR, reporting odds ratio; Con-DOX, conventional doxorubicin; Lip-DOX, Liposomal doxorubicin; SMQ, standardized medical queries; 95% CI, 95% confidence interval).
Full size table

We further analyzed the outcomes of cardiotoxicity-related AEs (Table 4). The rates of serious AEs within the ‘Cardiac disorders’ SOC were 98.74% for Con-DOX and 98.61% for Lip-DOX, a difference that was not statistically significant (χ2 = 0.03, p = 0.856). Among these serious cardiac AEs, the mortality rates were 26.21% for Con-DOX and 23.15% for Lip-DOX, which also did not represent a statistically significant difference (χ2 = 1.22, p = 0.270). These results suggest that while the incidence and spectrum of cardiotoxicity differ, the severity of outcomes—once cardiotoxicity occurs—may not be significantly associated with the DOX formulation. This underscores that DIC, regardless of formulation, leads to severe outcomes with high mortality rates, necessitating clinical prioritization, early identification, and proactive management.

Table 4 Outcomes of cardiotoxicity-related adverse event for Con-DOX and Lip-DOX in FAERS. (Con-DOX, conventional doxorubicin; Lip-DOX, liposomal doxorubicin).
Full size table

Identification of targets associated with DIC

To identify the molecular targets associated with DIC, we employed a multi-database screening approach. First, potential DOX-binding targets were predicted using online servers: 298 targets were obtained from PharmMapper, 113 from SuperPred, and 22 from SwissTarget Prediction. After removing duplicates, 418 unique targets were identified as DOX-associated targets. Next, cardiotoxicity-related targets were sourced from the Comparative Toxicogenomics Database (CTD) and the GeneCards database, yielding 879 and 481 targets, respectively. After merging these results and removing duplicate entries, 1,197 unique targets were designated as cardiotoxicity-associated targets. Finally, the DOX-associated targets and the cardiotoxicity-associated targets were compared to identify common elements. A Venn diagram analysis revealed 113 common targets at the intersection of these two sets (Fig. 2A). A compound-targets-disease network visualizing the relationship between DOX, these targets, and cardiotoxicity is illustrated in Fig. 2B. These results indicate that these 113 overlapping targets may mediate DIC.

Fig. 2
figure 2

Identification of doxorubicin-induced cardiotoxicity (DIC) targets and doxorubicin-targets-cardiotoxicity network construction: (A) Venn diagram showing the number of unique targets of doxorubicin (n = 418) and cardiotoxicity (n = 1,197), as well as overlapping targets (n = 113). (B) The network of doxorubicin-targets-cardiotoxicity.

Full size image

PPI network construction

The 113 overlapping targets were imported into the STRING database to construct a protein–protein interaction (PPI) network, which was then analyzed and visualized using Cytoscape software (version 3.6). The resulting PPI network contained 112 nodes and 587 edges. In the visualization (Fig. 3A), node color represents the degree of connectivity, with red indicating highly connected hub proteins. The top 10 hub genes—TP53, EGFR, AKT1, SRC, HSP90AA1, STAT3, ESR1, HSP90AB1, NFKB1, and ERBB2—were identified within this network using the CytoHubba plugin by applying the Degree method of centrality analysis (Fig. 3B).

Fig. 3
figure 3

PPI Network Construction: (A) PPI analysis of the 113 overlapping targets was conducted using the STRING database and visualized in Cytoscape. (B) Top 10 hub targets identified by the CytoHubba plugin based on node degree. (PPI, protein–protein interaction).

Full size image

Enrichment analysis and DOX-Targets-Cardiotoxicity-Pathway network construction

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed on the 113 common targets using the Metascape database. The top 10 significantly enriched terms of GO-BP, GO-CC, and GO-MF and the top 20 KEGG (all with p < 0.05) were visualized as bubble charts.

Biological Process (BP) terms were primarily enriched in cellular response to hormone stimulus, cellular response to lipid, response to xenobiotic stimulus, positive regulation of cell migration, regulation of establishment of protein localization, response to molecule of bacterial origin, and response to oxidative stress. Cellular Component (CC) terms were significantly associated with the receptor complex, ficolin-1-rich granule lumen, membrane raft, transcription regulator complex, protein kinase complex, and glutamatergic synapse. Molecular Function (MF) terms were predominantly enriched for kinase binding, kinase activity, nuclear receptor activity, protein domain-specific binding, carboxylic acid binding, and kinase regulator activity (Fig. 4A).

Fig. 4
figure 4

Enrichment analysis and doxorubicin-target-cardiotoxicity-pathway network construction: (A) Gene Ontology (GO) enrichment analysis. (B) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. (C) Doxorubicin-target-cardiotoxicity-pathway network.

Full size image

The top 20 significantly enriched KEGG pathways are displayed in Fig. 4B. These pathways primarily included ‘Pathways in cancer’, ‘Fluid shear stress and atherosclerosis’, ‘Lipid and atherosclerosis’, ‘Proteoglycans in cancer’, ‘Chemical carcinogenesis—receptor activation’, ‘EGFR tyrosine kinase inhibitor resistance’, ‘HIF-1 signaling pathway’, ‘Diabetic cardiomyopathy’, ‘Thyroid hormone signaling pathway’, and ‘Insulin resistance’.

Finally, an integrated DOX-targets-cardiotoxicity-pathway network was constructed and visualized using Cytoscape to illustrate the complex interactions between the drug, its targets, the disease, and the involved pathways (Fig. 4C).

Discussion

DOX remains a core chemotherapeutic agent in oncology nearly 50 years after its introduction, though its clinical use is limited by significant cardiotoxicity. To enhance therapeutic efficacy and reduce toxicity, researchers have developed liposomal encapsulation strategies for DOX—an approach that has achieved significant success. Our research showed that compared with Con-DOX, Lip-DOX had fewer PT signals and lower ROR values for cardiotoxicity. This finding is consistent with numerous previous studies, including real-world analyses7,10, randomized controlled trials11, and meta-analyses12. Encapsulation of DOX within a liposome significantly alters its bioavailability, biodistribution, and, consequently, its biological activity. The reduced toxicity profile of the liposomal formulation is attributed to its prolonged retention within the circulatory system, thereby limiting distribution to off-target organs13. Among liposomal formulations, the most robust evidence for improved cardiac safety has been demonstrated for pegylated liposomal doxorubicin (PLD), which is associated with a significantly lower risk of cardiac events (P < 0.001)14. Furthermore, a critical finding was that once adverse cardiac events occurred, their severity and mortality were not significantly different between the two formulations. DOX has been widely reported to induce dose-dependent, progressive, and potentially lethal myocardial damage15,16. The prospective evaluation of DOX cardiotoxicity demonstrated a clear dose–response relationship5. As chemotherapy cycles progress and cumulative doses increase, Lip-DOX still carries the risk of inducing cardiotoxicity. Based on this, investigating the mechanism of DIC is essential for preventing such adverse reactions and has become a clinical challenge that urgently needs to be addressed. Network pharmacology, by integrating multi-omics data, can systematically reveal the potential molecular mechanisms of DIC; combining real-world data from the FAERS database with network pharmacology analyses enables cross-scale integration—from clinical observations to molecular mechanism elucidation. Our study not only provides a theoretical basis for the development of novel cardioprotective strategies but also offers scientific support for ensuring the safety of DOX in clinical practice.

The descriptive analysis of DIC reveals important epidemiological characteristics. Age-related factors should not be overlooked.The role of age in DIC remains controversial. Some studies suggest that the risk of cardiotoxicity increases with patient age17,18. However, research on DIC in childhood and adolescent cancer indicates that younger age may also elevate this risk19,20. Investigating the effect of age on DIC is challenging because most publications focus exclusively on pediatric or adult populations, with few including both demographics21. Real-world data from pharmacovigilance databases like FAERS theoretically provide a robust method for exploring the relationship between age and DIC across all age groups. Su et al., for instance, reported through a FAERS analysis that age was a risk factor for cardiotoxicity10. However, upon reviewing the study data, we found that over 80% of the included patients had no age reported. We therefore Suggest this conclusion be interpreted with extreme caution. In our own study, even after applying strict inclusion criteria, age information was missing from 26.13% of Con-DOX reports and 54.25% of Lip-DOX reports. Consequently, we did not pursue a direct analysis of the relationship between age and DIC due to this substantial missing data. The role of sex in DIC also remains unclear. Thus, gender-related factors are equally important.Most animal studies suggest that male rodents are more susceptible to cardiotoxicity than females, who generally exhibit a protective effect22,23. However, findings regarding sex differences in human DIC are inconsistent with animal models and also vary between patient populations. Most studies in pediatric cancer patients identify female sex as a significant risk factor for DIC19,24, while some evidence in adult cancer patients suggests females may be protected25. After excluding reports with missing sex data, our study showed that female patients had a higher OR for cardiotoxicity at the SOC level in the general population. It is critical to note that our analysis focused solely on this broad association. Many other factors, such as age, comorbidities, and genetic polymorphisms, may interact with sex to influence the occurrence of cardiotoxicity. Future research should incorporate a more comprehensive set of clinically relevant variables to conduct in-depth multivariate analyses, aiming to fully elucidate the complex mechanisms underlying DOX-induced cardiotoxicity.

Building on the clinical observations of DIC, our network pharmacology analysis dissected its molecular underpinnings by identifying core targets and their interactive signaling pathways. We integrated data from the CTD, GeneCards, PharmMapper, SwissTargetPrediction, and SuperPred databases to systematically identify 113 potential targets associated with DIC. Using the STRING platform and Cytoscape software, we constructed a protein–protein interaction (PPI) network and identified 10 core targets—TP53, EGFR, AKT1, SRC, HSP90AA1, STAT3, ESR1, HSP90AB1, NFKB1, and ERBB2—that play critical roles in DIC. Existing literature supports the central role of these targets. Regarding TP53, McSweeney et al. found that the p53 protein is a key regulator of DOX-induced transcriptome changes that upregulate death receptors and activate the extrinsic apoptosis pathway26. Conversely, Li et al. revealed that p53’s regulation of the mitochondrial genome upon DOX exposure plays a protective role against cardiotoxicity, highlighting its complex function27. For EGFR, studies have found that deficiency of the key protein CHMP4C blocks the lysosomal degradation of EGFR, leading to its sustained activation and exacerbation of pressure overload-induced cardiac hypertrophy28. For AKT1, Ma et al. revealed that AKT1 deficiency exacerbates myocardial cell apoptosis and tissue injury29. For SRC, studies indicate that combined inhibition of SRC-mediated matrix mechanosensing and TGF-β signaling may serve as a therapeutic strategy for treating cardiovascular fibrosis30. Regarding HSP90AA1 and HSP90AB1 (members of the heat shock protein 90 family), Teresa et al. demonstrated that DOX exacerbates cardiotoxicity by binding to HSP90AA1, triggering mitochondrial dysfunction and ultimately leading to cardiomyocyte death in mice31. For STAT3, Zhang et al. demonstrated that during DIC, upregulation of miR-526b-3p leads to downregulation of STAT3, inhibiting VEGFA activity and exacerbating myocardial injury32. For ESR1, Zhang et al. reported that low ESR1 expression increases the risk of ascending aortic dilation and acute type A aortic dissection (ATAAD) by weakening vascular wall integrity33. For NFKB1, Luo et al. suggested that NFKB1 gene mutation leads to abnormal activation of the NF-κB pathway, which increases DRP1 expression, causes excessive mitochondrial fission and dysfunction, ultimately leading to increased cell apoptosis and exacerbating coronary artery disease34. For ERBB2, Aharonov et al. proposed that ERBB2 mediates a robust YAP mechanotransduction signal involving EMT-like features, which can promote cardiac regeneration35. These core targets form an interconnected regulatory network. For example, TP53 and AKT1 competitively modulate HIF-1α activity through the PI3K-AKT pathway36,37, while HSP90 stabilizes EGFR and ERBB2, thereby influencing their downstream signaling pathways38. This interconnected regulatory network forms the complex molecular basis underlying DIC.

The classic mechanisms of DIC are mainly oxidative stress and inflammation. For oxidative stress, DIC is strongly associated with oxidative stress, primarily mediated through several key pathways. The Nrf2/Keap1/ARE pathway is a central regulator: under normal conditions, Nrf2 is sequestered by Keap1 and degraded, but DOX disrupts this interaction, impairing the activation of antioxidant genes (e.g., SOD, HO-1) by inhibiting Nrf2 and upregulating Keap139. The Sirt1/p66Shc pathway is also involved: DOX downregulates Sirt1, a deacetylase that normally suppresses p66Shc, leading to increased p66Shc-mediated mitochondrial ROS production40. Additionally, DOX activates NADPH oxidases (NOX2 and NOX4), which generate superoxide anions and hydrogen peroxide, triggering MAPK-dependent apoptosis41,42. Iron signaling further contributes: DOX forms complexes with Fe2⁺, promoting hydroxyl radical generation via the Fenton reaction and inhibiting GPX4, leading to lipid peroxidation and ferroptosis43,44. For inflammation, it plays a critical role in DIC through specific signaling cascades. The NLRP3/caspase-1 pathway is activated by DOX, leading to the assembly of the NLRP3 inflammasome, activation of caspase-1, and cleavage of pro-inflammatory cytokines (IL-1β, IL-18), which induces pyroptosis45. The HMGB1/TLR4/MAPKs pathway is also pivotal: DOX triggers the release of HMGB1, which activates TLR4 and downstream MAPK signaling, promoting the secretion of TNF-α and IL-1β46. Furthermore, the mTOR/TFEB/NF-κB pathway is involved: DOX inhibits TFEB nuclear translocation via mTOR, enhancing NF-κB activation and exacerbating inflammatory cytokine production47.

In our KEGG enrichment analysis, we focused primarily on the HIF-1 signaling pathway. The HIF (hypoxia-inducible factor) signaling pathway plays a key role in cellular responses to hypoxic environments and is widely involved in physiological and pathological processes such as angiogenesis, energy metabolism, cell survival, and apoptosis36,48. Under physiological conditions, the HIF signaling pathway maintains basal activity, participating in the regulation of normal cardiac angiogenesis and energy metabolism. Moderate activation of this pathway can be cardioprotective; for example, it promotes angiogenesis in ischemic myocardium and improves myocardial blood supply36,49. However, under pathological conditions such as myocardial ischemia and heart failure, local cardiac hypoxia leads to excessive activation of the HIF signaling pathway, which may contribute to cardiotoxicity by promoting abnormal vascular remodeling and myocardial fibrosis49. Our prior analysis identified TP53 and AKT1 as playing important roles in this pathway. Regarding cardiotoxicity, the TP53 gene has a dual role. When the heart is damaged, normal TP53 expression can eliminate damaged cells by inhibiting proliferation and promoting apoptosis, a protective mechanism. For example, some strategies aim to enhance the protective effect of HIF-1α and reduce cardiotoxicity by inhibiting p53 activity50. However, excessive TP53 activation may lead to excessive cardiomyocyte apoptosis, exacerbating heart damage. For instance, in myocardial ischemia–reperfusion injury models, p53 aggravates myocardial injury by inhibiting HIF-1α-mediated angiogenesis and the expression of cell survival-related genes50. Similarly, AKT1 has a dual role in the heart. During myocardial ischemia, activation of the PI3K-AKT signaling cascade enhances HIF-1α expression and activity, thereby promoting neovascularization, facilitating energy metabolism adaptation, and mitigating injury36. However, excessive activation of the AKT1-HIF signaling pathway may lead to abnormal angiogenesis and cellular metabolic disorders, instead exacerbating cardiotoxicity36. DOX directly influences this interplay. It induces oxidative stress and DNA damage that activate TP53, upregulate p53 protein expression, and subsequently inhibit HIF-1α activity, thereby impairing HIF-mediated cardioprotective mechanisms. Concurrently, DOX may inhibit the PI3K-AKT signaling pathway, reducing AKT1 activity and the stabilization and nuclear translocation of HIF-1α, which further exacerbates cardiotoxicity. Thus, while activation of the HIF signaling pathway may represent an adaptive response of cardiomyocytes to DIC, this response is often insufficient to fully counteract the toxic effects of DOX37. These findings not only deepen our understanding of the complex interplay underlying DIC but also suggest potential multi-target intervention strategies for its clinical prevention and treatment.

The limitations of this study should be interpreted within the context of pharmacovigilance methodologies. First, reliance on the FAERS database inherently introduces biases, including widespread underreporting of adverse events (particularly for mild/moderate cases), indication bias (e.g., preferential use of Lip-DOX in patients with pre-existing cardiac risk factors), geographical reporting disparities, differences in concomitant medications, and variations in reporters’ professional backgrounds. Therefore, when analyzing cardiotoxicity across DOX formulations, we restricted patient regions to mitigate geographical bias; however, this approach cannot resolve fundamental limitations like underreporting or indication bias. Second, although ROR provides valuable signal detection capabilities, its performance is sensitive to low reporting counts and arbitrary threshold selection, which may lead to the exaggeration of false-positive signals for rare events or drugs with limited exposure51. Finally, the lack of laboratory-based experimental validation limits our ability to causally elucidate the biological mechanisms of DIC. Given the limitations of the retrospective analysis in this study, future multi-center, large-sample prospective cohort studies could be conducted to further validate the differences in cardiotoxicity between various DOX formulations and the related mechanisms, thereby providing more reliable evidence-based support for clinical medication decisions.

Conclusion

Through disproportionality analysis of data from the FAERS, we found that in terms of both signal frequency and intensity, DIC is a severe adverse reaction that should not be overlooked in clinical practice. Both formulations pose a risk of causing cardiac disorders; however, Lip-DOX is associated with fewer cardiac adverse event signals. Notably, there was no significant difference between the two formulations in terms of the incidence of serious adverse events or mortality rates attributable to cardiotoxicity. In addition, our gender-based analysis of DIC revealed that females have a higher risk of developing DIC compared to Males. Through network pharmacology, we identified 10 core targets and several key signaling pathways closely associated with DIC, especially the HIF-1 signaling pathway, and revealed the multi-target, multi-pathway synergistic regulatory mechanism underlying DIC. These study findings provide valuable insights and references for drug safety monitoring and clinical medication safety. The discovery of new pathways offers novel ideas for reducing the toxicity of DOX and conducting research on related drugs. However, due to the inherent limitations of this study, our conclusions require further validation through prospective investigations.

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 repository containing spontaneous adverse event reports submitted globally. We downloaded the American Standard Code for Information Interchange (ASCII) data files for the period from the first quarter of 2004 to the first quarter of 2025, comprising seven structured data tables: 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, and PRIMARYID. Data for all reports mentioning DOX were extracted and merged from the relevant FAERS tables. The variables of interest included case ID (CASEID), primary ID (PRIMARYID), indications (INDI), suspected drugs (DRUG), adverse events (REAC), outcome events (OUTC), reporter country, reporter type, patient sex, patient age, report date (FDA_DT), therapy start date (START_DT), and event date (EVENT_DT). Following FDA recommendations, duplicate reports were identified and removed using a standardized deduplication algorithm: For cases sharing an identical CASEID, the report with the most recent FDA_DT (FDA receipt date) was retained. If multiple reports shared identical CASEID and FDA_DT values, the report with the highest PRIMARYID was retained. The data processing procedures for the FAERS database and network pharmacology, as well as the connection between these two parts of data, are illustrated in the Supplementary Figure S1.

Adverse event and drug identification

Cardiotoxicity-related adverse events were identified using the following Standardized MedDRA Queries (SMQ): Cardiomyopathy (SMQ code: 20,000,150), Cardiac failure (SMQ code: 20,000,004), Shock-associated circulatory or cardiac conditions (excluding torsade de pointes/QT prolongation) (SMQ code: 20,000,067), Supraventricular tachyarrhythmias (SMQ code: 20,000,057), Ventricular tachyarrhythmias (SMQ code: 20,000,047), Other ischaemic heart disease (SMQ code: 20,000,168), Cardiac arrhythmia terms, nonspecific (SMQ code: 20,000,162), Ventricular tachyarrhythmias (SMQ code: 20,000,058), Supraventricular tachyarrhythmias (SMQ code: 20,000,059), Torsade de pointes/QT prolongation (SMQ code: 20,000,001), Torsade de pointes, shock-associated conditions (SMQ code: 20,000,068), Noninfectious myocarditis/pericarditis (SMQ code: 20,000,239), Arrhythmia-related investigations, signs and symptoms (SMQ code: 20,000,051).

DOX is commercially available in two main formulations: Con-DOX and Lip-DOX. However, as FAERS is a spontaneous reporting system (SRS), data quality varies by country. Many reports lack the critical information (e.g., generic drug name, formulation type) required to accurately distinguish between Con-DOX and Lip-DOX. Analyzing all DOX-related reports without reliable formulation classification would introduce significant misclassification bias. Therefore, to ensure accurate formulation identification, the analysis was restricted to reports where the reporter country was the United States. Reports were first categorized as Con-DOX or Lip-DOX based on the generic name and formulation type fields. For reports where this information was missing, the formulation was determined by cross-referencing the reported manufacturer name with official FDA drug labeling databases.

Signal detection and data analysis

This study utilized disproportionality analysis, a standard pharmacovigilance approach for detecting potential associations between drugs and AEs. ROR was used to quantify the relative reporting frequency of a specific AE for the target drug compared to all other drugs in the database. Signal thresholds were defined as follows: ROR signals required both a lower 95% confidence interval (CI) > 1.0 and at least 3 case reports51. The chi-square test required a p-value < 0.05. 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 Supplementary Table S2.

Identification of common targets for DOX and cardiotoxicity

First, to identify potential molecular targets of DOX, its Canonical SMILES notation was obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and imported into the SwissTargetPrediction (http://www.swisstargetprediction.ch/) and SuperPred servers (http://prediction.charite.de/). The 2D structure of DOX was obtained from PubChem for use in the PharmMapper server (http://lilab.ecust.edu.cn/). Second, to identify cardiotoxicity-related targets, the CTD (Comparative Toxicogenomics Database) (https://ctdbase.org/) and the GeneCards database (https://www.genecards.org/) were queried using the keyword“cardiotoxicity”. Finally, the potential DOX targets were intersected with the cardiotoxicity-related targets using a Venn diagram (https://jvenn.toulouse.inrae.fr) to identify overlapping genes, which were defined as the common targets for subsequent analysis52,53,54. These shared targets in both the DOX target dataset and the cardiotoxicity target dataset reflect a statistical association between DOX and its induced cardiotoxicity at the molecular level.

Protein–protein interaction (PPI) network construction

The common targets were imported into the STRING database (https://cn.string-db.org) to evaluate protein–protein interactions55. 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 3.6). The CytoHubba plugin was used to identify the top hub genes based on degree centrality.

Functional and pathway enrichment analysis

The common targets were imported into the Metascape platform (https://metascape.org/gp/) for functional enrichment analysis. KEGG pathway56,57 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. 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 through the shared targets, forming a functional link with DIC, and then the results were visualized using the Bioinformatics online platform (https://www.bioinformatics.com.cn). Finally, an integrated network illustrating the relationships between DOX, its common targets, and the enriched pathways was constructed using Cytoscape58,59.