Machine learning predictor to investigate treatment modalities and overall survival in HER2+ patients with early-stage breast cancer

利用机器学习预测器研究HER2阳性早期乳腺癌患者的治疗方式和总生存期

阅读:1

Abstract

PURPOSE: This study aimed to explore the impact of treatment modalities on the survival outcomes of HER2-positive patients with early-stage invasive ductal breast cancer. METHODS: Hierarchical clustering analysis was used to identify distinct subgroups based on treatment modalities. Comparative analysis between the clusters identified significant treatment-related variables. Cox regression analysis was performed to construct a survival prediction model and nomogram incorporating these variables. Random Survival Forest (RSF) and SHapley Additive exPlanations (SHAP) analysis were employed for further validation and interpretation. RESULTS: A total of 9569 patients with early-stage HER2+ invasive ductal breast cancer were included, and five treatment-related clusters were identified using hierarchical clustering. Post-clustering analysis revealed that survival outcomes were influenced by various treatment factors, including time length from diagnosis to treatment, surgery approach, response to neoadjuvant therapy, combination with radiation, chemotherapy and/or systemic therapy, and treatment sequence. A prediction model and nomogram were developed, demonstrating good discriminatory ability and excellent predictive performance at 3-, 5-, and 8-years. CONCLUSIONS: The study highlighted the importance of an aggressive and comprehensive treatment approach for patients with early-stage HER2-positive breast cancer. It emphasized the multifaceted nature of treatment outcomes and the need to consider multiple treatment factors beyond surgery alone. The developed survival prediction model provided valuable insights into the contribution of different treatment modalities to survival outcomes.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。