基于LASSO回归的多发性骨髓瘤诊断模型的建立

Construction of multiple myeloma diagnostic model based on LASSO regression

目的利用常见的实验室指标构建多发性骨髓瘤(MM)的诊断模型,以提高现有检验指标的诊断效能。方法回顾性分析2015年1月至2022年1月在广东省中医院确诊的MM患者96例(病例组),随机选取85例健康体检者作为对照组。常见相关实验室指标17个:血清免疫球蛋白M(IgM)、血清免疫球蛋白G(IgG)、血清免疫球蛋白A(IgA)、血清β_(2)-微球蛋白(β_(2)-MG)、校正血清钙、血清清蛋白(ALB)、血清球蛋白(GLB)、清蛋白与球蛋白比值(A/G)、血清碱性磷酸酶(ALP)、血清肌酐(Cr)、白细胞计数(WBC)、中性粒细胞计数(NEUT)、淋巴细胞计数(LYMT)、中性粒细胞计数与淋巴细胞计数比值(NLR)、红细胞计数(RBC)、血红蛋白(Hb)、血小板计数(PLT)。对病例组和对照组两组的17个检验指标数据进行统计分析,通过LASSO回归算法筛选出系数不为0的检验指标,再进行Logistic回归分析以及受试者工作特征(ROC)曲线分析。结果通过LASSO回归,筛选出系数不为0的指标4个,分别为IgM、NLR、Hb和ALB。再对这4个指标进行二元多因素Logistic分析,结果显示回归模型为Y=18.008-4.329IgM+1.374NLR-0.067Hb-0.240 ALB。该回归模型曲线下面积为0.987,灵敏度为96.9%,特异度为95.3%。结论该诊断模型对MM的诊断具有重要价值。

Objective To construct a diagnostic model of multiple myeloma(MM)by using the common laboratory indicators so as to improve the diagnostic efficiency of existing detection indicators.Methods Ninety-six cases of MM definitely diagnosed in this hospital from January 2015 to January 2022(case group)were retrospectively analyzed,and 85 healthy subjects undergoing physical examination were randomly selected as the control group.Seventeen common laboratory indicators were as follows serum immunoglobulin M(IgM),serum immunoglobulin G(IgG),serum immunoglobulin A(IgA),serumβ_(2)-microglobulin(β_(2)-MG),corrected serum calcium,serum albumin(ALB),serum globulin(GLB),albumin to globulin ratio(A/G),serum alkaline phosphatase(ALP),serum creatinine(Cr),white blood cell count(WBC),neutrophil count(NEUT),lymphocyte count(LYMT),neutrophil count to lymphocyte count(NLR),red blood cell count(RBC),hemoglobin(Hb)and platelet count(PLT).The data of these 17 test indexes in the case group and control group were statistically analyzed.The detection indexes whose coefficient was not 0 were screened by LASSO regression algorithm,and then the Logistic regression analysis and the receiver operating characteristic(ROC)curve analysis were performed.Results Four indexes whose coefficient was not 0 were screened out by the LASSO regression,which were IgM,NLR,Hb and ALB.Then these 4 indexes conducted the binary multivariate Logistic analysis and the results showed that the regression model was Y=18.008-4.329IgM+1.374NLR-0.067 Hb-0.240ALB.The area under the curve of this regression model was 0.987,the sensitivity was 96.9%and the specificity was 95.3%.Conclusion This diagnostic model is of great value in the diagnosis of MM.