国际生殖健康/计划生育杂志 ›› 2023, Vol. 42 ›› Issue (5): 353-360.doi: 10.12280/gjszjk.20230109

• 论著 •    下一篇

冻融胚胎移植临床妊娠的影响因素分析及列线图预测模型构建

丁凯, 赵纯, 凌秀凤, 李欣()   

  1. 210004 南京医科大学附属妇产医院(南京市妇幼保健院)生殖医学中心
  • 收稿日期:2023-03-13 出版日期:2023-09-15 发布日期:2023-09-13
  • 通讯作者: 李欣 E-mail:lixin@njmu.edu.cn
  • 基金资助:
    国家自然科学基金(81971386);江苏省妇幼保健协会科研课题(FYX202204)

Analysis of Factors Affecting Clinical Pregnancy during Frozen-Thawed Embryo Transfer and Construction of Prediction Model of Nomogram

DING Kai, ZHAO Chun, LING Xiu-feng, LI Xin()   

  1. Reproductive Medicine Center, Maternity Hospital of Nanjing Medical University (Nanjing Maternal and Child Health Hospital), Nanjing 210004, China
  • Received:2023-03-13 Published:2023-09-15 Online:2023-09-13
  • Contact: LI Xin E-mail:lixin@njmu.edu.cn

摘要:

目的: 探讨行激素替代治疗(hormone replacement therapy,HRT)的冻融胚胎移植(frozen-thawed embryo transfer,FET)周期临床妊娠的影响因素,构建临床妊娠预测模型。方法: 回顾性分析2018年1月—2020年12月在南京医科大学附属妇产医院行HRT的2 107个周期FET患者的临床资料。按照7 ∶ 3的比例随机分为训练集(1 499例)和验证集(608例),并根据临床妊娠与否,分为临床妊娠组与未临床妊娠组,探讨各临床特征与临床妊娠结局的关系。利用最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)回归模型筛选影响因素,通过十折交叉验证法选择模型中的最优参数λ,并通过多因素Logistic回归分析构建列线图预测模型,采用校准曲线、受试者工作特征(receiver operating characteristic,ROC)曲线和决策曲线分析(decision curve analysis,DCA)对其效能进行评估。结果: ①2 107个周期中有1 354个周期(64.26%)获得临床妊娠。②LASSO回归模型筛选10个变量纳入模型,包括年龄、基础卵泡刺激素(FSH)、基础黄体生成素(LH)、基础抗苗勒管激素(AMH)、FET后14 d雌二醇(E2)及妊娠、流产、生产、人工流产和足月产次数。③调整混杂因素后,多因素Logistic回归分析发现年龄、基础LH、基础AMH和FET后14 d E2水平与临床妊娠相关(均P<0.05),并据此构建列线图模型。④该模型在训练集和验证集的ROC曲线下面积分别为0.662和0.683,并且校准曲线在训练集和验证集中的预测风险与实际结果之间具有良好一致性,表现出一定的区分度和良好的校准度。DCA结果表明,当训练集和验证集的阈值概率分别在1%~79%和1%~81%时,采用该列线图预测模型可以使患者的净收益提高。结论: 建立了预测FET患者临床妊娠的列线图模型,在一定程度上可以帮助临床医师在移植周期中或移植前采取个性化治疗措施提高临床妊娠率。

关键词: 生殖技术,辅助, 冻融胚胎移植, 妊娠, 影响因素分析, 列线图, 预测, 模型,统计学

Abstract:

Objective: To investigate the influencing factors of clinical pregnancy during the freeze-thawed embryo transfer (FET) cycle of hormone replacement therapy (HRT), and construct a clinical pregnancy prediction model. Methods: The clinical data of 2 107 cycles of the FET who underwent HRT at the Maternity Hospital of Nanjing Medical University from January 2018 to December 2020 were retrospectively analyzed. These cycles were randomly divided into the training set (1 499 cases) and the verification set (608 cases) according to the ratio of 7 ∶ 3. To explore the relationship between clinical characteristics and clinical pregnancy outcomes, they were divided into the clinical pregnancy group and the non-clinical pregnancy group. The least absolute shrinkage and selection operator (LASSO) regression model was used to screen the influencing factors, and the optimum parameter λ in the model was selected by the ten-fold cross-validation method. Multivariate Logistic regression analysis was used to construct a nomogram prediction model. The calibration curve, receiver operating characteristic (ROC) curve and decision curve analysis (DCA) were used to evaluate its effectiveness. Results: ① A total of 1 354 clinical pregnancy was in 2 107 cycles (64.26%). ②Ten variables selected by the LASSO regression model were included in the model, including age, basal FSH, basal LH, basal AMH, E2 at 14 days after FET, and the number of pregnancy, abortion, labor, induced abortion, and term birth. ③After adjusting for confounding factors, the multivariate Logistic regression analysis found that age, basal LH and AMH levels, and E2 levels at 14 days after FET were correlated with clinical pregnancy (both P<0.05), and a nomogram model was constructed accordingly. ④The areas under the ROC curve of the training set and the verification set were 0.662 and 0.683, respectively. The calibration curve showed a good consistency between the predicted risks and the actual results of the training set and the verification set, suggesting a certain degree of differentiation and good calibration degree. The DCA showed that when the threshold probabilities of the training set and the validation set were 1%-79% and 1%-81% respectively, the net benefit of patients was improved by using the column-line prediction model. Conclusions: A nomogram model was established to predict clinical pregnancy in the patients with FET, which can help clinicians to take the personalized treatment measure during or before FET so as to improve the clinical pregnancy rate to a certain extent.

Key words: Reproductive techniques, assisted, Frozen-thawed embryo transfer, Pregnancy, Root cause analysis, Nomograms, Forecasting, Models, statistical