{"title":"建立基于机器学习的多西他赛诱导的乳腺癌患者骨髓抑制风险预测模型及危险因素分析。","authors":"Hongya Ou, Zijan Tan, Anle Shen, Liting Yu, Yibo He, Ri Hong, Xiaoru Lin, Xiaofeng Shi, Konglang Xing, Xiaoli Song, Yonglin Liu, Lingli Zou, Junyu Li","doi":"10.1007/s11096-025-01989-x","DOIUrl":null,"url":null,"abstract":"<p><strong>Introduction: </strong>Breast cancer is the most prevalent malignancy in women worldwide. Docetaxel-based chemotherapy is commonly used for treatment, but its clinical application is often constrained by hematologic toxicity, particularly severe bone marrow suppression. The early identification of high-risk patients is essential to prevent complications and optimize therapeutic outcomes. Machine learning offers advanced capabilities for risk prediction by capturing complex patterns beyond those of traditional statistical models.</p><p><strong>Aim: </strong>This study aimed to identify the risk factors associated with bone marrow suppression in breast cancer patients receiving docetaxel-based chemotherapy, and to develop and validate predictive models using machine learning algorithms.</p><p><strong>Method: </strong>This retrospective case-control study included 119 patients with breast cancer treated with docetaxel-based chemotherapy at the Hainan Hospital of Shanghai Children's Medical Center, Shanghai Jiao Tong University School of Medicine, from January 2020 to December 2024. Patients were divided into bone marrow suppression (n = 57; WHO grades II-IV) and non-suppression (n = 62; grades 0-I) groups based on WHO toxicity criteria. Multivariate logistic regression was used to identify independent risk factors. Three prediction models, logistic regression, random forest, and AdaBoost, were constructed and evaluated. A five-fold cross-validation with 50 repetitions was performed to ensure model stability and reliability.</p><p><strong>Results: </strong>Multivariate analysis revealed that a low baseline lymphocyte count (OR = 4.95, 95% CI 1.62-17.0), decreased white blood cell (WBC) count (OR = 0.62, 95% CI 0.40-0.9), and reduced prealbumin level (OR = 0.98, 95% CI 0.97-0.99) were significantly associated with severe bone marrow suppression (all p < 0.05). Among the models, AdaBoost achieved the best overall performance (AUC = 0.81; specificity = 94%; accuracy = 77%). The random forest model showed the highest sensitivity (83%), while logistic regression was more interpretable but demonstrated a lower predictive ability (AUC = 0.69).</p><p><strong>Conclusion: </strong>Pretreatment lymphocyte count, WBC count, and prealbumin level are reliable predictors of docetaxel-induced bone marrow suppression. The AdaBoost model demonstrates high specificity (94%) in identifying low-risk patients, supporting accurate risk stratification and personalized care in breast cancer treatment.</p>","PeriodicalId":13828,"journal":{"name":"International Journal of Clinical Pharmacy","volume":" ","pages":""},"PeriodicalIF":3.2000,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Development of a machine learning-based risk prediction model and analysis of risk factors for docetaxel-induced bone marrow suppression in breast cancer patients.\",\"authors\":\"Hongya Ou, Zijan Tan, Anle Shen, Liting Yu, Yibo He, Ri Hong, Xiaoru Lin, Xiaofeng Shi, Konglang Xing, Xiaoli Song, Yonglin Liu, Lingli Zou, Junyu Li\",\"doi\":\"10.1007/s11096-025-01989-x\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Introduction: </strong>Breast cancer is the most prevalent malignancy in women worldwide. Docetaxel-based chemotherapy is commonly used for treatment, but its clinical application is often constrained by hematologic toxicity, particularly severe bone marrow suppression. The early identification of high-risk patients is essential to prevent complications and optimize therapeutic outcomes. Machine learning offers advanced capabilities for risk prediction by capturing complex patterns beyond those of traditional statistical models.</p><p><strong>Aim: </strong>This study aimed to identify the risk factors associated with bone marrow suppression in breast cancer patients receiving docetaxel-based chemotherapy, and to develop and validate predictive models using machine learning algorithms.</p><p><strong>Method: </strong>This retrospective case-control study included 119 patients with breast cancer treated with docetaxel-based chemotherapy at the Hainan Hospital of Shanghai Children's Medical Center, Shanghai Jiao Tong University School of Medicine, from January 2020 to December 2024. Patients were divided into bone marrow suppression (n = 57; WHO grades II-IV) and non-suppression (n = 62; grades 0-I) groups based on WHO toxicity criteria. Multivariate logistic regression was used to identify independent risk factors. Three prediction models, logistic regression, random forest, and AdaBoost, were constructed and evaluated. A five-fold cross-validation with 50 repetitions was performed to ensure model stability and reliability.</p><p><strong>Results: </strong>Multivariate analysis revealed that a low baseline lymphocyte count (OR = 4.95, 95% CI 1.62-17.0), decreased white blood cell (WBC) count (OR = 0.62, 95% CI 0.40-0.9), and reduced prealbumin level (OR = 0.98, 95% CI 0.97-0.99) were significantly associated with severe bone marrow suppression (all p < 0.05). Among the models, AdaBoost achieved the best overall performance (AUC = 0.81; specificity = 94%; accuracy = 77%). The random forest model showed the highest sensitivity (83%), while logistic regression was more interpretable but demonstrated a lower predictive ability (AUC = 0.69).</p><p><strong>Conclusion: </strong>Pretreatment lymphocyte count, WBC count, and prealbumin level are reliable predictors of docetaxel-induced bone marrow suppression. The AdaBoost model demonstrates high specificity (94%) in identifying low-risk patients, supporting accurate risk stratification and personalized care in breast cancer treatment.</p>\",\"PeriodicalId\":13828,\"journal\":{\"name\":\"International Journal of Clinical Pharmacy\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":3.2000,\"publicationDate\":\"2025-08-26\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"International Journal of Clinical Pharmacy\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://doi.org/10.1007/s11096-025-01989-x\",\"RegionNum\":4,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"PHARMACOLOGY & PHARMACY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Clinical Pharmacy","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1007/s11096-025-01989-x","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"PHARMACOLOGY & PHARMACY","Score":null,"Total":0}
引用次数: 0
摘要
乳腺癌是世界范围内最常见的女性恶性肿瘤。以多西他赛为基础的化疗是常用的治疗方法,但其临床应用往往受到血液学毒性的限制,特别是严重的骨髓抑制。早期识别高危患者对于预防并发症和优化治疗结果至关重要。机器学习通过捕捉超越传统统计模型的复杂模式,为风险预测提供了先进的能力。目的:本研究旨在确定与接受多西他赛化疗的乳腺癌患者骨髓抑制相关的危险因素,并利用机器学习算法开发和验证预测模型。方法:本回顾性病例对照研究纳入2020年1月至2024年12月上海交通大学医学院附属上海儿童医学中心海南医院接受多西他赛化疗的119例乳腺癌患者。根据WHO毒性标准,将患者分为骨髓抑制组(n = 57, WHO分级II-IV)和非抑制组(n = 62, WHO分级0-I)。采用多因素logistic回归确定独立危险因素。构建并评价了logistic回归、随机森林和AdaBoost三种预测模型。进行了50次重复的五重交叉验证,以确保模型的稳定性和可靠性。结果:多因素分析显示,较低的基线淋巴细胞计数(OR = 4.95, 95% CI 1.62-17.0)、白细胞(WBC)计数降低(OR = 0.62, 95% CI 0.40-0.9)和前白蛋白水平降低(OR = 0.98, 95% CI 0.97-0.99)与严重的骨髓抑制显著相关(均为p)结论:预处理淋巴细胞计数、WBC计数和前白蛋白水平是多西他赛诱导的骨髓抑制的可靠预测因子。AdaBoost模型在识别低风险患者方面显示出高特异性(94%),支持准确的乳腺癌治疗风险分层和个性化护理。
Development of a machine learning-based risk prediction model and analysis of risk factors for docetaxel-induced bone marrow suppression in breast cancer patients.
Introduction: Breast cancer is the most prevalent malignancy in women worldwide. Docetaxel-based chemotherapy is commonly used for treatment, but its clinical application is often constrained by hematologic toxicity, particularly severe bone marrow suppression. The early identification of high-risk patients is essential to prevent complications and optimize therapeutic outcomes. Machine learning offers advanced capabilities for risk prediction by capturing complex patterns beyond those of traditional statistical models.
Aim: This study aimed to identify the risk factors associated with bone marrow suppression in breast cancer patients receiving docetaxel-based chemotherapy, and to develop and validate predictive models using machine learning algorithms.
Method: This retrospective case-control study included 119 patients with breast cancer treated with docetaxel-based chemotherapy at the Hainan Hospital of Shanghai Children's Medical Center, Shanghai Jiao Tong University School of Medicine, from January 2020 to December 2024. Patients were divided into bone marrow suppression (n = 57; WHO grades II-IV) and non-suppression (n = 62; grades 0-I) groups based on WHO toxicity criteria. Multivariate logistic regression was used to identify independent risk factors. Three prediction models, logistic regression, random forest, and AdaBoost, were constructed and evaluated. A five-fold cross-validation with 50 repetitions was performed to ensure model stability and reliability.
Results: Multivariate analysis revealed that a low baseline lymphocyte count (OR = 4.95, 95% CI 1.62-17.0), decreased white blood cell (WBC) count (OR = 0.62, 95% CI 0.40-0.9), and reduced prealbumin level (OR = 0.98, 95% CI 0.97-0.99) were significantly associated with severe bone marrow suppression (all p < 0.05). Among the models, AdaBoost achieved the best overall performance (AUC = 0.81; specificity = 94%; accuracy = 77%). The random forest model showed the highest sensitivity (83%), while logistic regression was more interpretable but demonstrated a lower predictive ability (AUC = 0.69).
Conclusion: Pretreatment lymphocyte count, WBC count, and prealbumin level are reliable predictors of docetaxel-induced bone marrow suppression. The AdaBoost model demonstrates high specificity (94%) in identifying low-risk patients, supporting accurate risk stratification and personalized care in breast cancer treatment.
期刊介绍:
The International Journal of Clinical Pharmacy (IJCP) offers a platform for articles on research in Clinical Pharmacy, Pharmaceutical Care and related practice-oriented subjects in the pharmaceutical sciences.
IJCP is a bi-monthly, international, peer-reviewed journal that publishes original research data, new ideas and discussions on pharmacotherapy and outcome research, clinical pharmacy, pharmacoepidemiology, pharmacoeconomics, the clinical use of medicines, medical devices and laboratory tests, information on medicines and medical devices information, pharmacy services research, medication management, other clinical aspects of pharmacy.
IJCP publishes original Research articles, Review articles , Short research reports, Commentaries, book reviews, and Letters to the Editor.
International Journal of Clinical Pharmacy is affiliated with the European Society of Clinical Pharmacy (ESCP). ESCP promotes practice and research in Clinical Pharmacy, especially in Europe. The general aim of the society is to advance education, practice and research in Clinical Pharmacy .
Until 2010 the journal was called Pharmacy World & Science.