基于LASSO回归分析构建不明原因复发性流产患者再次妊娠失败的风险预测模型与验证分析  

Construction and verification analysis of risk prediction model for re-pregnancy failure in patients with unexplained recurrent spontaneous abortion based on LASSO regression analysis

作　　者：张伟[1] 江雪娟 朱怡[2] 应翩[3] ZHANG Wei;JIANG Xuejuan;ZHU Yi;YING Pian(Department of Reproductive Medicine,Hangzhou Red Cross Hospital,Hangzhou,Zhejiang 310003,China;Department of Gynecology and Obstetrics,Hangzhou Red Cross Hospital,Hangzhou,Zhejiang 310003,China;Department of Gynecology and Obstetrics,Zhejiang Provincial Hospital of Chinese Medicine,Hangzhou,Zhejiang 310006,China)

机构地区：[1]杭州市红十字会医院生殖医学科,杭州310003 [2]杭州市红十字会医院妇产科,杭州310003 [3]浙江省中医院妇产科,杭州310006

出　　处：《重庆医学》2024年第24期3744-3751,共8页Chongqing Medical Journal

基　　金：浙江省自然科学基金项目(LY22H270001)。

摘　　要：目的基于最小绝对收缩和选择算子(LASSO)回归分析构建不明原因复发性流产(URSA)患者再次妊娠失败的相关因素,构建预测模型并验证。方法采取前瞻性研究,选择2021年2月至2024年2月杭州市红十字会医院收治的420例URSA患者作为研究对象,以7∶3比例进行简单随机抽样法,分为建模组与验证组。所有URSA患者均随访至终止妊娠,根据妊娠结局分为妊娠成功组、妊娠失败组,统计患者基线资料。采用LASSO回归筛选可能影响URSA患者再次妊娠失败的危险因素进行二元logistic回归分析,并依据相关因素构建列线图模型,根据回归结果构建列线图模型;采用Bootstrap内部验证法对列线图模型进行验证,以验证组数据对列线图模型进行外部验证。结果通过LASSO回归模型的三折交叉验证确定最佳惩罚项系数λ,在λ±s x处筛选,最终筛选出9个潜在的相关因素,分别为年龄、流产次数、自身免疫因素、空腹血糖受损、DNA断裂指数(DFI)、优质胚胎数、人绒毛膜促性腺激素(HCG)、黄体酮(P)、雌二醇(E_(2));二元logistic回归分析结果显示,年龄大、流产次数多、自身免疫因素异常、空腹血糖受损、DFI高是URSA患者再次妊娠失败的危险因素(OR=1.286、4.574、3.114、3.204、1.234,P<0.05);优质胚胎数多及血清HCG、P、E_(2)高表达是URSA患者再次妊娠失败的保护因素(OR=0.547、0.826、0.948、0.995,P<0.05);使用Bootstrap内部验证法对建模组、验证组列线图模型进行验证,一致性指数(C-index)值分别为0.959、0.968,两组的校正曲线均与理想曲线拟合反应良好,绘制ROC曲线结果显示:建模组AUC为0.959(95%CI:0.937~0.981,P<0.001),灵敏度为86.6%、特异度为95.7%、约登指数为0.823;验证组AUC为0.968(95%CI:0.949~0.988,P<0.001),灵敏度为88.5%、特异度为97.4%、Youden指数为0.859。结论年龄、流产次数、自身免疫因素、空腹血糖受损、DFI、优质胚胎数、HCG、P、E_(2)与UObjective To construct the related factors of re-pregnant failure in the patients with unexplained recurrent spontaneous abortion(URSA)based on the least absolute shrinkage and selection operator(LASSO)regression analysis,and to construct the prediction model and conduct the verification.Methods The prospective study was adopted to select 420 URSA patients admitted and treated in Hangzhou Municipal Red Cross Hospital from February 2021 to February 2024 as the research subjects.They were divided into modeling group and verification group by simple random sampling method with a ratio of 7∶3.All URSA patients were followed up until termination of pregnancy.They were divided into the pregnancy success group and pregnancy failure group based on the pregnancy outcome.The baseline data of the patients conducted the statistics.LASSO regression was used to screen the risk factors that might affect the re-pregnancy failure of URSA patients for conducting the binary logistic regression analysis,and the nomogram model was constructed according to the relevant factors and a nomogram model was constructed according to the regression results.The Bootstrap internal verification method was used to verify the nomogram model,and the verification group data were used to conduct the external verification on the nomogram model.Results The optimal penalty term coefficientλwas determined by the three-fold cross-validation of the LASSO regression model.Nine potential related factors were screened atλ±s x,which were the age,number of abortions,autoimmune factors,impaired fasting glucose,DFI,number of high-quality embryos,human chorionic gonadotropin(HCG),progesterone(P)and estradiol(E_(2)).The results of binary logistic regression analysis showed that elder,more abortions,abnormal autoimmune factors,impaired fasting glucose and high DFI were the risk factors for re-pregnancy failure in URSA patients(OR=1.286,4.574,3.114,3.204,1.234,P<0.05).The large number of high-quality embryos and high expression of serum HCG,P and E_(2) were the

关 键 词：不明原因复发性流产 再次妊娠失败 最小绝对收缩和选择算子回归 预测模型 危险因素 

分 类 号：R715.2[医药卫生—妇产科学]