实验室多参数指标预测恶性肿瘤骨髓转移的模型建立与验证  

Establishment and validation of a laboratory-based multiparameter model for predicting bone marrow metastasis in malignant tumors

作　　者：李浩成 续薇[1] 杜忠华[2] 宋琳 刘丹[2] 邵慧慧 赵春贺 崔巍琦 曲林琳[1] Li Haocheng;Xu Wei;Du Zhonghua;Song Lin;Liu Dan;Shao Huihui;Zhao Chunhe;Cui Weiqi;Qu Linlin(Department of Laboratory Medicine,the First Hospital of Jilin University,Changchun130021,China;Department of Haematology,the First Hospital of Jilin University,Changchun130021,China)

机构地区：[1]吉林大学第一医院检验科,长春130021 [2]吉林大学第一医院血液科,长春130021

出　　处：《中华检验医学杂志》2024年第11期1248-1255,共8页Chinese Journal of Laboratory Medicine

基　　金：吉林省科技厅(20210101335JC)。

摘　　要：目的筛选实验室多参数指标,建立恶性肿瘤骨髓转移(BMM)预测模型并验证。方法采用病例对照研究,选取2018年3月至2024年3月在吉林大学第一医院收治的恶性肿瘤患者444例为研究对象,其中模型建立集243例,模型验证集201例;模型建立集又分为BMM阳性组(81例)和BMM阴性组(162例),模型验证集分阳性验证组(67例)和阴性验证组(134例)。收集患者一般临床信息(包括性别、年龄、临床诊断等),以及47项实验室指标(包括血常规、凝血常规、肝功能、肿瘤标志物、钾、钠、氯、钙离子检测、骨髓形态学检查、骨髓活检等)。将BMM作为结局事件,采用2组间比较筛选差异变量,采用Pearson相关性分析研究参数之间的相关性,采用多因素Logistic回归分析筛选BMM的风险因素,建立Logistic模型,采用受试者工作特征(ROC)曲线评价BMM预测模型的诊断效能。结果模型建立集中BMM阳性组与阴性组间存在差异的28项参数的Pearson相关性分析显示,平均血小板体积(MPV)、红细胞比容(HCT)、血红蛋白(HGB)、凝血酶原时间(PT)等17项参数的相关系数均≤0.6(P<0.05),进一步多因素Logistic回归分析显示,MPV、HCT、HGB、PT、红细胞分布宽度(RDW)、碱性磷酸酶(ALP)、血小板计数(PLT)、平均红细胞血红蛋白浓度(MCHC)、氯离子(Cl-)为恶性肿瘤发生BMM的危险因素[MPV(OR=9.929,95%CI 2.688~71.335)、HCT(OR=8.232,95%CI 6.223~9.841)、HGB(OR=4.300,95%CI 1.947~16.577)、PT(OR=3.738,95%CI 1.359~11.666)、RDW(OR=1.995,95%CI 1.275~3.807)、ALP(OR=1.025,95%CI 1.012~1.045)、PLT(OR=1.014,95%CI 1.002~1.031)、MCHC(OR=0.724,95%CI 0.523~0.880)和Cl-(OR=0.703,95%CI 0.472~0.967)]。模型建立集中,各风险因素联合应用预测恶性肿瘤BMM的AUC为0.943(95%CI 0.898~0.987,P<0.001),敏感度为86.3%,特异度为89.2%;模型验证集中,AUC为0.924(95%CI 0.854~0.960,P<0.001),敏感度为86.7%,特异度为83.8%。结论本研究建立并验证了BMM的多参数预测模型,有助Objective To establish and validate the prediction model for bone marrow metastasis(BMM)in malignant tumors by screening out laboratory multiparameters.Methods This case-control study collected 444 cases of malignant tumor patients who were hospitalized in the First Hospital of Jilin University from March 2018 to March 2024,including 243 cases for model establishment set and 201 cases for model validation set.The model establishment set was divided into BMM positive group(81 cases)and BMM negative group(162 cases),and the model validation set was divided into positive group(67 cases)and a negative group(134 cases).We collected patients′clinical information such as gender,age,clinical diagnosis,and results of 47 laboratory tests including routine blood analysis,coagulation,liver function,tumor markers,potassium,sodium,chloride,and calcium ion tests,bone marrow morphology,and bone marrow biopsy.BMM was taken as the outcome event,differencial variables were analyzed using inter group comparisons,the correlation among parameters was analyzed using Pearson correlation analysis,the risk factors for BMM were analyzed using multivariate conditional logistic regression analysis,to establish logistic model,followed by efficiency evaluation on BMM predictive model using receiver operating characteristic(ROC)curves.Results In the model establishment set,Pearson correlation analysis of 28 parameters that differed between the BMM positive and negative groups revealed that the correlation coefficients of 17 parameters,including mean platelet volume(MPV),hematocrit(HCT),hemoglobin(HGB),and prothrombin time(PT),were no more than 0.6(P<0.05).Further multivariate conditional logistic regression analysis demonstrated that MPV,HGB,HCT,PT,red cell distribution width(RDW),platelet count(PLT),alkaline phosphatase(ALP),chloride(Cl-),and mean erythrocyte hemoglobin concentration(MCHC)were the risk factors of BMM occurence in malignancy[MPV(OR=9.929,95%CI 2.688-71.335),HCT(OR=8.232,95%CI 6.223-9.841),HGB(OR=4.300,95%CI 1.947-16.577),PT(O

关 键 词：肿瘤 骨肿瘤 预测 

分 类 号：R73-37[医药卫生—肿瘤]