Age-matched analysis of endometrial immune profile between frozen embryo transfer success and recurrent implantation failure cases with predictive model

Abstract

Successful implantation during frozen embryo transfer (FET) is influenced by multiple factors. However, the role of endometrial immune cells in predicting FET outcomes remains unclear, particularly in an age-independent context. This study aimed to develop a prediction model based on endometrial immune cell characteristics to predict FET outcomes regardless of patient age. We analyzed 78 patients undergoing FET (2021–2023), categorized into successful FET and Recurrent Implantation Failure (RIF) groups. After age-matching via Propensity Score Matching, endometrial immune cell profiles were compared between groups. Key predictors identified through Elastic Net analysis were used to develop a CatBoost prediction model. Before age matching, significant differences between groups were observed in age, neutrophil, monocyte, lymphocyte, and CD56^hiCD16^-NK levels. After age matching, only lymphocyte and B cell levels remained significantly different. The model using lymphocytes, CD56^hiCD16⁺ NK cells, and B cells as predictors achieved ROC AUC scores of 0.88 (age-matched test set) and 0.84 (full dataset), with 0.80 accuracy. CD56^hiCD16⁺ NK cells showed strong positive correlation with RIF. This study reveals a novel association between CD56^hiCD16⁺ NK cells and recurrent implantation failure. The age-independent prediction model demonstrates the potential of immune cell profiling for predicting FET outcomes and may guide personalized reproductive medicine approaches.