预测含高致吐风险铂类药物化疗致恶心呕吐人工智能模型的建立与评价  

Establishment and evaluation of an artificial intelligence model for predicting nausea and vomiting caused by platinum‑based chemotherapy with high emetic risk

作　　者：张靖悦 杨翀[2] 蓝高爽[1] 孙银娟 张琳琳[3] 袁恒杰[1] Zhang Jingyue;Yang Chong;Lan Gaoshuang;Sun Yinjuan;Zhang Linlin;Yuan Hengjie(Department of Pharmacy,Tianjin Medical University General Hospital,Tianjin 300052,China;Department of Pharmacy,Tianjin Huanhu Hospital,Tianjin 300350,China;Department of Medical Oncology,Tianjin Medical University General Hospital,Tianjin 300052,China)

机构地区：[1]天津医科大学总医院药剂科,天津300052 [2]天津市环湖医院,天津300350 [3]天津医科大学总医院肿瘤科,天津300052

出　　处：《药物不良反应杂志》2023年第10期577-583,共7页Adverse Drug Reactions Journal

基　　金：国家自然科学基金青年项目(81102447)。

摘　　要：目的基于人工智能技术建立肿瘤患者接受高致吐风险铂类药物化疗诱发恶心呕吐(CINV)的风险预测模型,为止吐方案的选择提供依据。方法收集2018年1月至2022年12月在天津医科大学总医院肿瘤科登记的接受顺铂或卡铂[卡铂血药浓度-时间曲线下面积(AUC)≥4]肿瘤患者的临床信息,包括性别、年龄、饮酒史、孕吐史、化疗周期、患者对发生恶心呕吐的预期、化疗药物、止吐方案、院外止吐治疗、化疗前夜睡眠小于7 h、上周期发生CINV情况、肌酐清除率(Ccr)等。对数据进行预处理后随机分为训练集和测试集,训练集用于构建预测模型,测试集用于评价模型预测效能。分别采用梯度提升树(GBDT)、随机森林(RF)和逻辑回归(LR)3种算法建立预测模型,并对模型性能进行评价。评价指标包括准确度、灵敏度、召回率、F1值(灵敏度和召回率的调和平均数)和受试者工作特征曲线下面积(AUROC)。最后应用Shapley加法解释(SHAP)方法对有预测意义的临床特征进行可解释性分析。结果本研究共纳入患者698例,男性439例(62.9%);中位年龄64(21,84)岁;共接受1654个周期化疗。化疗方案含顺铂者364例,化疗864个周期;含卡铂且AUC≥4者361例,化疗790个周期。止吐方案选择神经激肽-1受体拮抗剂(NK-1 RA)、5-羟色胺3受体拮抗剂(5-HT3 RA)和地塞米松者的治疗周期数为1347个,选择5-HT3 RA和地塞米松者的周期数为307个。Spearman相关性分析结果显示肿瘤患者特征变量之间相关性不强,均可用于模型建立。GBDT最优超参数nestimators=500,maxdepth=9;RF最优超参数maxdepth=5;LR最优超参数penalty=L2。根据最优超参数训练数据分别建立GBDT、RF和LR 3种预测模型。GBDT模型准确度为0.903,灵敏度为0.882,召回率为0.903,F1值为0.883,AUROC为0.778±0.036(95%CI:0.739~0.814);RF模型的准确度为0.885,灵敏度为0.861,召回率为0.885,F1值为0.870,AUROC为0.679±0.041(95%CI:0.63Objective To provide a basis for the selection of antiemetic regimen by establishing an artificial intelligence model for predicting chemotherapy⁃induced nausea and vomiting(CINV)in cancer patients receiving platinum⁃based chemotherapy with high emetic risk.Methods The clinical informa⁃tion on cancer patients who received cisplatin or carboplatin with area under the blood concentration⁃time curve(AUC)≥4 and registered in the Department of Oncology,Tianjin Medical University General Hospital from January 2018 to December 2022 was collected,including gender,age,history of alcohol consumption,history of vomiting in pregnancy,chemotherapy cycle,patient expects to have CINV,chemotherapeutic agents,antiemetic regimen,out⁃of⁃hospital antiemetic treatment,sleep of less than 7 hours on the night before chemotherapy,occurrence of CINV in the previous cycle,and creatinine clearance(Ccr).After pre⁃processing,the data were randomly divided into the training set and the test set.The training set was used to con⁃struct the prediction model,and the test set was used to evaluate the prediction efficiency of the model.Three algorithms,gradient boosting decision tree(GBDT),random forest(RF),and logistic regression(LR),were used to build a prediction model and evaluate the model performance,respectively.The evaluation metrics included accuracy,sensitivity,recall,F1 value(the reconciled mean of sensitivity and recall),and area under the receiver operating characteristic curve(AUROC).Finally,Shapley Additive exPlanation(SHAP)was applied to analyze the interpretability of the clinical features with predictive significance.Results A total of 698 patients,439 males(62.9%)with a median age of 64(21,84)years,were included in this study and received a total of 1654 cycles of chemotherapy.The chemotherapy regimen contained cisplatin in 364 cases with 864 cycles of chemotherapy,and carboplatin with AUC≥4 in 361 cases with 790 cycles of chemo⁃therapy.The number of treatment cycles in which neurokinin⁃1 receptor antagoni

关 键 词：人工智能 顺铂 卡铂 预测模型 化疗引起的恶心呕吐 肿瘤 

分 类 号：R730.53[医药卫生—肿瘤]