关键词: 异常受精, 体外受精, 预测模型, 正常受精, 列线图, 女方因素

Abstract: BACKGROUND: Reducing the rate of abnormal fertilization is an effective approach to improving the efficacy of in vitro fertilization and reducing patients’ financial strain. However, the current research on abnormal fertilization has focused on exploring the types of prokaryotic nuclei and their generation mechanisms, as well as analyzing embryos formed by abnormal fertilization, chromosomal ploidy and utilization value. There is a lack of clinical prediction models for abnormal fertilization based on retrospective studies.
OBJECTIVE: To construct a nomogram model to predict abnormal female factors in in vitro fertilization.
METHODS: A total of 5 075 patients undergoing treatment for conventional in vitro fertilization at Nanxishan Hospital of Guangxi Zhuang Autonomous Region from March 2017 to March 2022 were retrospectively analyzed. The male confounders were calibrated on a 1:1 propensity score with a match tolerance of 0.02, and 1 672 cases were successfully matched. According to the Vienna Consensus, patients with ≥ 60% normal fertilization capacity were included in the normal fertilization group (n=836) and those with < 60% normal fertilization capacity were included in the abnormal fertilization group (n=836). The model and validation groups were obtained by random sampling at a ratio of 7:3. Factors related to the occurrence of abnormal fertilization following conventional in vitro fertilization in the model group were screened using univariate analysis and the best matching factors were selected using the Least Absolute Shrinkage and Selection Operator (LASSO) and included in a multifactorial forward stepwise Logistic regression to identify their independent influencing factors and plot a nomogram. Finally, the prediction model was validated for discrimination, accuracy and clinical application efficacy using receiver operating characteristic curves, calibration curves, clinical decision curves and clinical impact curves. 
RESULTS AND CONCLUSION: The univariate analysis indicated the factors influencing the occurrence of abnormal fertilization were age, controlled ovarian hyperstimulation protocol, number of assisted pregnancies, years of infertility, infertility factors, anti-mullerian hormone, sinus follicle count, basal luteinizing hormone, luteinizing hormone concentration on the human chorionic gonadotropin day, and estradiol level on human chorionic gonadotropin injection day (P < 0.05). LASSO regression further identified the best matching factors, including age, microstimulation protocol, number of assisted pregnancies, years of infertility, anti-mullerian hormone, luteinizing hormone level on human chorionic gonadotropin injection day, and estradiol level on human chorionic gonadotropin injection day (P < 0.05). Multifactorial forward stepwise Logistic regression results showed that age, microstimulation protocol, number of assisted conceptions, years of infertility, anti-mullerian hormone, and estradiol level on human chorionic gonadotropin injection day were independent influencing factors for the occurrence of abnormal fertilization following conventional in vitro fertilization. The receiver operating characteristic curves showed an area under the curve of 0.761 (0.746, 0.777) for the model group and 0.767 (0.733, 0.801) for the validation group, indicating that the model has good discrimination. The mean absolute error of the calibration curve was 0.044, and the Hosmer-Lemeshow test indicated that there was no significant difference between the predicted probability of abnormal fertilization and the actual probability of abnormal fertilization (P > 0.05), indicating the prediction model has good consistency and accuracy. The clinical decision curves and clinical impact curves showed that the model and validation groups had the maximum net clinical benefit at valve probability values of 0.00-0.52 and 0.00-0.48, respectively, and there was a good clinical application efficacy in this valve probability range. To conclude, the nomogram model has good discrimination and accuracy as well as clinical application efficacy for predicting the occurrence of abnormal fertilization in women undergoing conventional in vitro fertilization based on age, microstimulation protocol, number of assisted conceptions, years of infertility, anti-mullerian hormone, and estradiol level on human chorionic gonadotropin injection day.

Key words: abnormal fertilization, in vitro fertilization, prediction model, normal fertilization, nomogram, female factor