多囊卵巢综合征不孕症患者体外受精-胚胎移植助孕未妊娠的影响因素分析

Analysis of influencing factors of IVF-ET assisted pregnancy in PCOS infertile patients

摘要:

目的 分析多囊卵巢综合征（PCOS）不孕症患者体外受精-胚胎移植（IVF-ET）助孕未妊娠的影响因素，并根据影响因素构建预测模型。方法 选取2023年1月—2024年8月吉林大学中日联谊医院生殖医学中心PCOS不孕症患者135例，均进行首次IVF-ET助孕周期并接受新鲜胚胎移植，根据妊娠结果分为未妊娠组和妊娠组。采用多因素一般Logistic回归分析及森林图筛选PCOS不孕症患者IVF-ET助孕未妊娠的影响因素，建立预测模型。绘制受试者工作特征（ROC）曲线分析预测模型对PCOS不孕症患者IVF-ET助孕未妊娠的评估价值，绘制校准曲线进行一致性验证。结果 135例PCOS不孕症患者中IVF-ET助孕未妊娠发生率为45.93%（62/135）。未妊娠组年龄、体质量指数（BMI）>24 kg/m2占比、胰岛素抵抗指数（HOMA-IR）、基础黄体生成素/卵泡刺激素（LH/FSH）水平均高于妊娠组，基础雌二醇（E2）水平、基础抗苗勒管激素（AMH）水平均低于妊娠组（P <0.05）。多因素一般Logistic回归分析结果显示：孕妇年龄大［O^R =2.955（95% CI：1.573，5.553）］、BMI高［O^R =2.549（95% CI：1.044，6.226）］、HOMA-IR水平高［O^R =2.195（95% CI：1.435，3.355）］、基础LH/FSH水平高［O^R =2.732（95% CI：1.099，6.793）］均是PCOS不孕症患者IVF-ET助孕未妊娠的独立危险因素（P <0.05）；基础E2水平高［O^R =0.650（95% CI：0.537，0.786）］、基础AMH水平高［O^R =0.223（95% CI：0.104，0.479）］均是PCOS不孕症患者IVF-ET助孕未妊娠的保护因素（P <0.05）。ROC曲线分析结果显示：年龄、BMI、HOMA-IR、基础E2、基础LH/FSH、基础AMH联合评估PCOS不孕症患者IVF-ET助孕未妊娠的诊断效能更高，AUC值为0.947（95% CI：0.895，0.978），敏感性为90.32%（95% CI：0.804，0.956），特异性为91.78%（95% CI：0.831，0.963）。校准曲线显示预测概率与实际观测概率吻合良好（Brier得分0.087，校准斜率0.894）。内部验证（Bootstrap法及十折交叉验证）提示模型具有较好的稳定性，过拟合风险较低。结论 PCOS不孕症患者IVF-ET助孕未妊娠的影响因素包括孕妇年龄、BMI，HOMA-IR、基础E2、基础LH/FSH、基础AMH，基于以上影响因素构建的预测模型不仅具有较高的预测效能，预测值与实际发生风险的一致性较高。

Abstract:

Objective To analyze the factors associated with non-pregnancy after in vitro fertilization-embryo transfer (IVF-ET) in infertile patients with polycystic ovary syndrome (PCOS), and build a prediction model based on the influencing factorss.Methods A total of 135 PCOS infertile patients from Reproductive Medicine Center of China-Japan Union Hospital from January 2023 to August 2024 were selected. All patients underwent their first IVF-ET cycle with fresh embryo transfer. According to the pregnancy results, they were divided into non-pregnant group and pregnancy group. Multivariate binary Logistic regression analysis and forest map were used to screen the influencing factors of IVF-ET assisted pregnancy in PCOS infertile patients and establish a prediction model. Receiver operating characteristic (ROC) curve was drawn to evaluate the model for predicting non-pregnancy after IVF-ET in infertile patients with PCOS, and the calibration curve was drawn for consistency verification.Results In this study, the incidence of non-pregnancy after IVF-ET in 135 PCOS infertility patients was 45.93% (62/135). The non-pregnancy group had significantly higher maternal age, proportion of BMI >24 kg/m2, HOMA-IR level, and basal LH/FSH level, but lower basal E2 level and basal AMH level compared with the pregnancy group (all P < 0.05). Multivariate logistic regression analysis showed that advanced maternal age [O^R = 2.955 (95% CI: 1.573, 5.553) ], high BMI [O^R = 2.549 (95% CI: 1.044, 6.226) ], high HOMA-IR level [O^R = 2.195 (95% CI: 1.435, 3.355) ], and high basal LH/FSH level [O^R = 2.732 (95% CI: 1.099, 6.793) ] were independent risk factors for non-pregnancy after IVF-ET in PCOS infertility patients (P < 0.05), whereas high basal E2 level [O^R = 0.650 (95% CI: 0.537, 0.786) ] and high basal AMH level [O^R = 0.223 (95% CI: 0.104, 0.479) ] were protective factors (P < 0.05). ROC curve analysis indicated that the combination of age, BMI, HOMA-IR, basal E2, basal LH/FSH and basal AMH had higher diagnostic efficacy for predicting non-pregnancy, with an AUC of 0.947 (95% CI: 0.895, 0.978), a sensitivity of 90.32% (95% CI: 0.804, 0.956) and a specificity of 91.78% (95% CI: 0.831, 0.963). The calibration curve showed good agreement between predicted probabilities and observed outcomes (Brier score 0.087, calibration slope 0.894). Internal validation (bootstrap and ten-fold cross-validation) indicated good model stability and low risk of overfitting.Conclusion Factors associated with non-pregnancy after IVF-ET in infertile patients with PCOS include maternal age, body mass index, HOMA-IR, basal E2, LH/FSH and AMH. The prediction model constructed based on the above influencing factorss not only has high predictive efficacy, but also has a high consistency between the predicted value and the actual risk.