Insights into anthracycline-based chemotherapy and implications for genetic correlation, expression variation and long-term outcomes of heart disease among breast cancer survivors: Mendelian randomization, bioinformatic analyses, and a SEER-Cohort study

ABSTRACT BACKGROUND Prior cohort studies and observational research indicate that the probability of developing cardiovascular disease (CVD) may rise following a breast cancer (BC) diagnosis. However, it is uncertain whether this association is causal or simply due to confounding factors (treatment regimen, chemotherapeutic agents, etc.). Cardiotoxicity remains a significant adverse effect of chemotherapy, resulting in reduced acceptance and tolerance of treatment, as well as morbidity and death associated with cardiovascular disease. Nonetheless, hub immune-related genes linked to adriamycin-induced cardiotoxicity (AIC) in BC survivors have not been extensively investigated. Moreover, the prognosis for cardiovascular disease in BC women undergoing anthracycline-based treatment remains under further investigation. METHOD Pooled data on breast cancer and prevalent CVD were retrieved by genome-wide association studies (GWAS). In the random effects models, the principal estimates were conducted using inverse variance weighting (IVW) together with five additional methods to investigate the causal association between breast cancer and cardiovascular disease. We performed various bioinformatics methodologies to identify crucial innate immunity-related genes and their association with AIC in BC women through datasets from the GEO and TCGA databases. Our population-based retrospective study utilized the Surveillance, Epidemiology, and End Results (SEER) database sourced from the United States. The standardized mortality rate (SMR) for CVD in BC patients undergoing chemotherapy and other systemic therapies was estimated. We used a multifactorial logistics regression model to calculate the odds ratios (ORs) for cardiovascular mortality among BC women who died from CVD within one year after receiving chemotherapy and developed a nomogram model to visualize the relationship between clinicopathological characteristics and cardiovascular mortality. Ultimately, we constructed a Lasso regression and XGBoost-based machine learning model to assess the predictive efficacy of the established machine learning model against multifactorial logistic regression in forecasting cardiovascular mortality. RESULT AND CONCLUSION In the Mendelian randomized study, we did not detect a positive causal association between BC and CVD. We applied a bioinformatics methodology to discover variations of innate immune genes resulting from adriamycin-induced cardiotoxicity in BC patients. We observed that DEGs related to the innate immune system were significantly linked to the enrichment of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, encompassing immunological response, immunomodulation, mitochondrial autophagy, B-cell receptor signaling, chemokine signaling, MAPK signaling, lipid metabolism, and atherosclerosis. Furthermore, we found that adriamycin-induced alterations in innate immune-related genes play a crucial role in the PPI network. Certain immune-related genes were substantially correlated with poor survival outcomes in BC survivors and demonstrated diagnostic effectiveness in concurrent heart failure after chemotherapy. Women receiving chemotherapy-only treatment have a highest risk of developing fetal heart disease than those undergoing other systemic therapies within 1 year from a large population-based retrospective study. We constructed a Lasso-XGBoost machine Learning model to predict the risk of cardiovascular death following chemotherapy to demonstrate the great potential of machine learning and artificial intelligence to be applied in this domain.

BACKGROUND Prior cohort studies and observational research indicate that the probability of developing cardiovascular disease (CVD) may rise following a breast cancer (BC) diagnosis. However, it is uncertain whether this association is causal or simply due to confounding factors (treatment regimen, chemotherapeutic agents, etc.). Cardiotoxicity remains a significant adverse effect of chemotherapy, resulting in reduced acceptance and tolerance of treatment, as well as morbidity and death associated with cardiovascular disease. Nonetheless, hub immune-related genes linked to adriamycin-induced cardiotoxicity (AIC) in BC survivors have not been extensively investigated. Moreover, the prognosis for cardiovascular disease in BC women undergoing anthracycline-based treatment remains under further investigation. METHOD Pooled data on breast cancer and prevalent CVD were retrieved by genome-wide association studies (GWAS). In the random effects models, the principal estimates were conducted using inverse variance weighting (IVW) together with five additional methods to investigate the causal association between breast cancer and cardiovascular disease. We performed various bioinformatics methodologies to identify crucial innate immunity-related genes and their association with AIC in BC women through datasets from the GEO and TCGA databases. Our population-based retrospective study utilized the Surveillance, Epidemiology, and End Results (SEER) database sourced from the United States. The standardized mortality rate (SMR) for CVD in BC patients undergoing chemotherapy and other systemic therapies was estimated. We used a multifactorial logistics regression model to calculate the odds ratios (ORs) for cardiovascular mortality among BC women who died from CVD within one year after receiving chemotherapy and developed a nomogram model to visualize the relationship between clinicopathological characteristics and cardiovascular mortality. Ultimately, we constructed a Lasso regression and XGBoost-based machine learning model to assess the predictive efficacy of the established machine learning model against multifactorial logistic regression in forecasting cardiovascular mortality. RESULT AND CONCLUSION In the Mendelian randomized study, we did not detect a positive causal association between BC and CVD. We applied a bioinformatics methodology to discover variations of innate immune genes resulting from adriamycin-induced cardiotoxicity in BC patients. We observed that DEGs related to the innate immune system were significantly linked to the enrichment of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, encompassing immunological response, immunomodulation, mitochondrial autophagy, B-cell receptor signaling, chemokine signaling, MAPK signaling, lipid metabolism, and atherosclerosis. Furthermore, we found that adriamycin-induced alterations in innate immune-related genes play a crucial role in the PPI network. Certain immune-related genes were substantially correlated with poor survival outcomes in BC survivors and demonstrated diagnostic effectiveness in concurrent heart failure after chemotherapy. Women receiving chemotherapy-only treatment have a highest risk of developing fetal heart disease than those undergoing other systemic therapies within 1 year from a large population-based retrospective study. We constructed a Lasso-XGBoost machine Learning model to predict the risk of cardiovascular death following chemotherapy to demonstrate the great potential of machine learning and artificial intelligence to be applied in this domain