Abstract
BACKGROUND: High-quality blastocysts with good developmental potential are crucial for successful pregnancy, yet the factors influencing their formation remain unclear in clinical practice, and no predictive model is available for early clinical use. Antinuclear antibodies (ANA), which are closely associated with autoimmune diseases but are also commonly found in the infertile patients, may significantly impact embryo quality. Therefore, evaluating the effect of ANA on blastocyst quality in infertile patients is of great importance, as it could facilitate the establishment of an early predictive model for the formation of high-quality blastocysts prior to oocyte retrieval. METHOD: This study aims to investigate the potential factors influencing high-quality blastocyst formation in infertile patients undergoing their first IVF/ICSI cycle with ANA titer screening, and to assess whether ANA titer impacts the formation and predictive significance of high-quality blastocysts. A total of 2,876 eligible infertility patients were randomly assigned to a training set (n = 1,725) and a validation set (n = 1,151) at a 6:4 ratio. Clinical characteristics of the optimal blastocyst development group (OBD group) and suboptimal blastocyst development group (SBD group) were compared to initially screen potential predictive variables (P < 0.1). Variables were further selected using RF, LASSO, and XGBoost machine learning methods, and those common to all three methods were identified as key predictors of optimal blastocyst development. A predictive model was then constructed using a multivariate logistic regression model, and its performance was evaluated through ROC and calibration curves in both the training and validation sets. Internal validation was performed using the Bootstrap method, and the Hosmer-Lemeshow test assessed the goodness-of-fit between predicted and actual outcomes. A nomogram was developed to visualize the model and facilitate early clinical application. Decision curve analysis (DCA) was performed to assess the clinical utility of the nomogram. RESULTS: There were no statistically significant differences in clinical characteristics between the training and validation sets after 6:4 randomization (P > 0.05). In the training set, female age, duration of infertility, baseline LH level, T level, FSH level, FSH/LH ratio, sperm concentration, normal morphology rate, male age, AMH level, AFC, infertility type, ANA titer, number of spontaneous abortions, and MCL were compared with the SBD group (P < 0.1) and considered as potential influencing factors. Five predictive variables closely related to OBD were identified using RF, LASSO, and XGBoost machine learning methods: female age, AMH level, AFC, ANA titer, and number of spontaneous abortions. The prediction model based on these variables achieved an AUC of 0.69 (95% CI: 0.66–0.71) in the training set and 0.70 (95% CI: 0.67–0.73) in the validation set, demonstrating good discriminative ability. The calibration curve for both the training set (P = 0.735) and validation set (P = 0.099) showed that the predicted probabilities closely aligned with the actual probabilities. The nomogram and DCA further enhance the clinical applicability of this model. CONCLUSION: A high-quality blastocyst prediction model was constructed using five easily obtainable clinical baseline characteristics, demonstrating clinically acceptable predictive ability. This model facilitates individualized diagnosis and treatment for infertility patients before embryo culture. Additionally, this study is the first to highlight the importance and predictive potential of ANA in high-quality blastocyst formation, underscoring the need for ANA-related screening and treatment in clinical practice. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s13048-025-01830-z.