AUTHOR=Zhang Lei , Liu Xiaofang , Bai Jiaojiao , Wang Yushan 

TITLE=Building a predictive model for pregnancy outcomes in ART patients based on AMH, FORT, HCG day EMT, and clinical characteristics

JOURNAL=Frontiers in Medicine

VOLUME=Volume 12 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2025.1699477

DOI=10.3389/fmed.2025.1699477

ISSN=2296-858X

ABSTRACT=ObjectiveTo establish a predictive model for pregnancy outcomes in patients undergoing assisted reproductive technology (ART) treatment using human chorionic gonadotropin (HCG) day endometrial thickness (EMT), follicle output rate (FORT), serum anti-Müllerian hormone (AMH), and clinical characteristics.MethodsA retrospective study was conducted on 200 patients who underwent ART treatment at our hospital from August 2019 to August 2024. All patient data were obtained from the electronic medical record system, and patients were divided into a clinical pregnancy group and a non-pregnancy group based on whether they achieved pregnancy. Clinical-related data were collected and compared. Two machine learning methods, Lasso regression and Extreme Gradient Boosting (XGBoost), were used for “overlap coverage” screening of risk factors. The XGBoost model was implemented using the “xgboost” package in R, with the following hyperparameters: max_depth = 6, learning_rate = 0.3, n_estimators = 100, objective = “binary:logistic.” These parameters were selected based on common practices for binary classification tasks and were not extensively tuned due to the sample size limitations. Multivariate logistic analysis was performed to identify risk factors for pregnancy outcomes in ART patients. A regression model was established using R software for prediction and validation.ResultsA total of 200 ART patients were included in this study. According to medical record system records, 109 patients did not achieve pregnancy, accounting for 54.50%, and were included in the non-pregnancy group. 91 patients achieved clinical pregnancy, accounting for 45.50%, and were included in the clinical pregnancy group. There were significant statistical differences between the non-pregnant group and the clinically pregnant group in terms of age, body mass index, number of oocytes retrieved, HCG day EMT, AMH, and FORT (p < 0.05); Using LASSO regression and XGBoost machine learning, different risk factors were identified. Through the “overlap coverage” method, six common risk factors for pregnancy outcomes in ART patients were identified and included in the logistic regression model. The results indicated that age and body mass index were both risk factors influencing pregnancy outcomes in ART patients (OR = 1.196, 1.777, p < 0.05); the number of retrieved oocytes, HCG day EMT, AMH, and FORT were protective factors influencing pregnancy outcomes in ART patients (OR = 0.366, 0.382, 0.182, 0.862, p < 0.05); Based on the results of the logistic regression analysis, a risk prediction model for pregnancy outcomes in ART patients was constructed using a nomogram. The ROC curve showed an AUC value of 0.911, with a 95% confidence interval (CI) of 0.871 to 0.951. The model was validated using the Bootstrap method with 1,000 repeated samples, yielding a Cox-Snell R2 of 0.629, Nagelkerke R2 was 0.471, Brier Score was 0.117, model fit p-value was 0.240, and the statistic was 10.369. This indicates that the model’s predictive results align well with the actual occurrence of pregnancy outcomes in patients, with the clinical decision curve generally exceeding the two extreme curves. This suggests that the factors included in the nomogram provide a high net benefit for predicting pregnancy outcomes in ART-treated patients.ConclusionThere are numerous factors influencing pregnancy outcomes in ART patients, primarily including age, body mass index, number of oocytes retrieved, HCG day EMT, AMH, and FORT. The risk prediction nomogram model constructed based on these factors has certain predictive value for pregnancy outcomes in ART patients. Clinicians should identify high-risk populations early and optimize treatment plans accordingly.