基于IVF-ET助孕周期前夫妻双方临床特征预测妊娠结局的Logistic回归模型研究  

A logistic regression model for predicting pregnancy outcomes based on clinical characteristics of couples before assisted pregnancy cycle with IVF/ET

作　　者：温晓晓[1] 翟亚楠[2] 杨杰[1] 许思娟[1] 胡颋 汪辉 倪亚莉[1] WEN Xiao-xiao;ZHAI Ya-nan;YANG Jie;XU Si-juan;HU Ting;WANG Hui;NI Ya-li(Reproductive Medicine Center,Gansu Provincial Maternity and Child Health Hospital,Lanzhou 730050;Department of Radiology,the First Hospital of Lanzhou University,Lanzhou 730000)

机构地区：[1]甘肃省妇幼保健院/甘肃省中心医院生殖医学中心,兰州730050 [2]兰州大学第一医院放射科,兰州730000

出　　处：《生殖医学杂志》2025年第2期206-212,共7页Journal of Reproductive Medicine

基　　金：甘肃省科技计划项目-创新基地和人才计划(21JR7RA679);兰州市科技计划项目(2023-2-23)。

摘　　要：目的基于IVF-ET助孕患者进入周期前的临床特征,建立预测妊娠结局的Logistic回归模型。方法分析812对不孕不育夫妇的临床数据,采用Lasso回归筛选影响妊娠结局的自变量。选择纳入模型的变量包括不孕类型、周期次数、窦卵泡数、抗苗勒管激素(AMH)水平、女方体质量指数(BMI)、不孕年限、女性年龄、梗阻性无精子症、严重少弱精子症、子宫内膜炎、精索静脉曲张、基础FSH(bFSH)水平、输卵管积水、男性年龄等。样本随机分为训练集(85%)和测试集(15%)。模型训练采用五折交叉验证,每折从训练集中抽取138对样本作为验证集。采用Logistic回归方法进行分类,以预测妊娠结局;采用受试者工作特征(ROC)曲线评估模型的预测效能,临床决策曲线分析(DCA)评估模型的临床实用价值。结果样本总数812对,妊娠失败282对,妊娠成功530对。在验证集中,模型预测的曲线下面积(AUC)为0.587[95%CI(0.486,0.688)]。最终模型在测试集中表现出AUC为0.726[95%CI(0.627,0.826)],准确度达0.730。DCA结果显示,当阈值概率为50%~100%时,列线图模型可获得较高的净获益值。结论基于Logistic回归的IVF-ET妊娠结局预测模型表现出中等的预测效能,具备一定的临床应用潜力。Objective:To establish a logistic regression model predicting pregnancy outcomes based on pre-cycle clinical characteristics of patients with IVF-ET.Methods:The analysis of clinical data from 812 infertile couples was conducted using Lasso regression to identify variables associated with pregnancy outcomes.Variables selected for inclusion in the model comprised the infertility type,the cycle number,antral follicle count,anti-Müllerian hormone(AMH)level,female body mass index(BMI),duration of infertility,female age,obstructive azoospermia,severe oligoasthenospermia,endometritis,varicocele,basal FSH(bFSH)level,hydrosalpinx and male age.The sample was randomly divided into a training set(85%)and a test set(15%).Model training employed 5-fold cross-validation,with each fold drawing 138 couples from the training set as the validation set.A logistic regression was utilized for classification to predict pregnancy outcomes,and the predictive performance of the model was assessed using the receiver operating characteristic(ROC)curve.Decision curve analysis(DCA)curve was drawn to evaluate the clinical practicability of the model.Results:In the dataset,there were 282 couples with failed pregnancies and 530 couples with successful pregnancies.In the validation set,the model predicted an area under the curve(AUC)of 0.587[95%CI(0.486,0.688)].The final model demonstrated an AUC of 0.726[95%CI(0.627,0.826)]and an accuracy of 0.730 in the test set.Clinical DCA results indicated that the nomogram model could achieve a higher net benefit when the threshold probability ranged from 50%to 100%.Conclusions:The pregnancy outcome prediction model based on logistic regression demonstrates moderate predictive performance and holds potential for clinical application.

关 键 词：体外受精-胚胎移植 妊娠结局 预测模型 

分 类 号：R711.6[医药卫生—妇产科学]