Endometrial receptivity characteristics in patients with repeated implantation failure: a study using LASSO regression and Bayesian generalized linear model analysis.

Background: Repeated implantation failure (RIF) is a key challenge in assisted reproductive technology (ART), and its mechanism is closely related to Endometrial receptivity (ER). Although the effects of some Endometrial receptivity parameters on pregnancy outcomes have been investigated, there is a lack of standardized assessment of the interaction of multidimensional parameters and systematic prediction models. Objective: This study aimed to investigate the key factors influencing Endometrial receptivity in patients with RIF using LASSO regression and Bayesian generalized linear models, with an emphasis on clinical interpretability and uncertainty quantification. Methods: This was a retrospective cohort study. A total of 506 women patients who underwent frozen-thawed embryo transfer (FET) at the Department of Reproductive Medicine from January 2023 to December 2024 were enrolled. All patient data and examination results were retrospectively collected from the electronic medical records system of our hospital. They were divided into a case group (RIF group, Results: The LASSO regression identified endometrial blood flow branches, endometrial arterial resistance index (RI), endometrial arterial pulsatility index (PI), endometrial arterial peak systolic velocity/end-diastolic velocity ratio (S/D), and endometrial peristaltic frequency as the most predictive variables. The Bayesian model indicated that increased endometrial blood flow branch count reduced the risk of RIF (OR = 0.1, 95% CI:0.06-0.17). Elevated endometrial arterial RI (OR = 1.21, 95% CI: 1.04-1.44) increased RIF risk, while decreased endometrial arterial PI (OR = 1.48, 95% CI: 1.06-2.10) showed an inverse association with RIF. Higher endometrial arterial peak systolic velocity/end-diastolic velocity ratio (OR = 3.63, 95% CI: 1.54-9.03) was positively correlated with RIF risk. Additionally, increased endometrial peristaltic frequency (OR = 1.93, 95% CI: 1.16-3.17) significantly elevated RIF risk. 10-fold cross-validation on the training set yielded a mean AUC of 0.904 ± 0.021. The model demonstrated excellent discriminatory ability (AUC = 0.911) and calibration performance (C-statistic = 0.94). Standardized net benefit analysis confirmed the clinical utility of the model across a wide range of risk thresholds (1:100 to 100:1). Conclusion: This study is the first to combine LASSO and Bayesian methods to construct a predictive model for RIF, highlighting the critical role of endometrial hemodynamic and peristaltic characteristics. The Bayesian framework offers uncertainty estimates that can guide personalized interventions. These findings provide new targets for precision diagnosis and treatment.