胃肠道恶性肿瘤患者导管相关性血栓形成预测模型的构建与验证  

Construction and Validation of Predictive Model for Catheter-Related Thrombosis in Patients with Gastrointestinal Malignant Tumors

作　　者：刘海志 韩野[1] 石晓 匡玉庭[1] LIU Hai-zhi;HAN Ye;SHI Xiao;KUANG Yu-ting(Department of General Surgery,First Affiliated Hospital of Soochow University,Suzhou,Jiangsu,215006,China)

机构地区：[1]苏州大学附属第一医院普外科,江苏苏州215006

出　　处：《中国血液流变学杂志》2024年第4期572-579,587,共9页Chinese Journal of Hemorheology

摘　　要：目的在胃肠肿瘤患者中,构建导管相关性血栓形成(catheter-related thrombosis,CRT)预测模型,并进行内部验证。方法选取2020年6月—2022年6月苏州大学附属第一医院接收的胃癌、结直肠癌患者共400例,按照7:3的比例分为训练队列(n=280)和验证队列(n=120)。根据中心静脉置管是否有血栓,分为实验组(血栓组)和对照组(无血栓组)。观察对照组与实验组在临床指标、营养指标、血栓形成的发生率,经单因素分析、多因素Logistic回归,筛选出独立危险因素,构建预测风险的nomogram,绘制ROC曲线、校正曲线、决策分析曲线评估模型精度和判别能力。结果1.训练队列与内部验证队列的基线分析差异无统计学意义;2.经单因素分析(以P<0.3为标准),筛选出以下可能因素:年龄、发热时间、Caprini评分、NRS2002营养评分、凝血酶原时间(PT)、C反应蛋白(CRP)、凝血酶时间(TT)、活化部分凝血活酶时间(APTT)、国际标准化比值(INR)、D-二聚体、前白蛋白、高密度脂蛋白胆固醇(HDL)、心脏病史、脑卒中史;3.多因素Logistic回归分析结果表明,发热时间、Caprini评分、NRS2002营养评分、CRP、APTT、D-二聚体、前白蛋白、心脏病史差异具有统计学意义(P<0.05);4.在训练队列中,预测模型的AUC分别为0.928(95%CI为0.885~0.961)。结论1.胃肠道恶性肿瘤患者CRT的独立危险因素包括发热时间、Caprini评分、NRS2002营养评分、CRP、APTT、D-二聚体、前白蛋白、心脏病史;2.该风险预测模型对胃肠道恶性肿瘤患者CRT,具有较高准确性,相较于传统的Caprini评分患者可从中获益。Objective To construct a prediction model for catheter-related thrombosis(CRT)in patients with gastrointestinal tumors and perform internal validation.Methods 400 patients with gastric cancer and colorectal cancer admitted in First Affiliated Hospital of Soochow University from June 2020 to June 2022 were selected and divided into training cohort(n=280)and verification cohort(n=120)according to a ratio of 7:3.The patients were divided into experimental group(thrombus group)and control group(no thrombus group)according to whether there was thrombus in central vein catheterization.The clinical indicators,nutritional indicators and incidence of thrombosis of the control group and the experimental group were observed.Independent risk factors were screened out by univariate analysis and multivariate Logistic regression,and nomogram was constructed to predict risk.ROC curves,calibration curves and decision analysis curves were drawn to evaluate the accuracy and discriminant ability of the model.Results 1.No significant difference was found in baseline analysis between the training cohort and the internal verification cohort.2.Through single factor analysis(P<0.3 as the standard),the following possible factors were screened out:age,fever time,Caprini score,NRS2002 nutritional score,prothrombin time(PT),C-reactive protein(CRP),thrombin time(TT),activated partial thromboplastin time(APTT),international normalized ratio(INR),D-dimer,prealbumin,high density lipoprotein cholesterol(HDL),heart disease history,and stroke history.3.The results of multivariate Logistic regression analysis showed that fever time,Caprini score,NRS2002 nutritional score,CRP,APTT,D-dimer,prealbumin and heart history had statistical significance(P<0.05).4.In the training cohort,the AUC of the prediction model was 0.928(95%CI,0.885-0.961),respectively.Conclusion 1.Independent risk factors for CRT in patients with gastrointestinal malignance include fever time,Caprini score,NRS2002 nutritional score,CRP,APTT,D-dimer,prealbumin,and cardiac history.2.Thi

关 键 词：胃肠道恶性肿瘤 营养风险 中心静脉置管 导管相关性血栓形成 LOGISTIC回归 预测模型 

分 类 号：R735[医药卫生—肿瘤]