Abstract
OBJECTIVE: To identify factors associated with clinical pregnancy and live birth in frozen embryo transfer (FET) cycles among women with adenomyosis and to develop internally validated prediction models using readily available clinical, laboratory, and imaging indicators. METHODS: This retrospective cohort study included 673 FET cycles in women with imaging-confirmed adenomyosis treated at the Reproductive Center of Women and Children's Hospital of Ningbo University between June 2018 and December 2024. Cycles ending in transfer failure or biochemical pregnancy were classified as non-clinical pregnancy, whereas cycles resulting in miscarriage or live birth were classified as clinical pregnancy. Among cycles achieving clinical pregnancy, outcomes were further categorized as miscarriage or live birth. Candidate predictors comprised demographic and infertility characteristics, anti-Müllerian hormone (AMH), pictorial blood loss assessment chart (PBAC) score, hemoglobin, carbohydrate antigen 125 (CA125), uterine volume, dysmenorrhea severity, adenomyosis type, endometrial-myometrial junctional zone involvement, and pretreatment modality. Separate multivariable logistic regression models were developed to predict clinical pregnancy and live birth. Model discrimination, calibration, and clinical usefulness were evaluated using the area under the receiver operating characteristic curve (AUC), bootstrap calibration, and decision curve analysis. RESULTS: Of 673 cycles, 216 achieved clinical pregnancy (32.10%); 160 resulted in live birth (23.77%) and 56 in miscarriage. Independent predictors of clinical pregnancy included CA125, uterine volume, dysmenorrhea severity, adenomyosis type, JZ involvement, and pretreatment modality. Live birth was primarily predicted by CA125, uterine volume, JZ involvement, and pretreatment modality. Higher CA125, larger uterine volume, and JZ involvement were associated with miscarriage risk, whereas medical therapy was associated with improved live birth. These predictors were identified for risk prediction, the observed relationships should not be interpreted as causal treatment effects. The models showed good performance (AUC 0.830 for clinical pregnancy, 0.921 for live birth), with acceptable bootstrap calibration and net clinical benefit across a broad range of thresholds. Observed event rates increased across low, intermediate, and high predicted-probability strata. CONCLUSION: In adenomyosis patients undergoing FET, uterine volume, adenomyosis type, JZ involvement, dysmenorrhea severity, CA125, and pretreatment modality are key predictors of clinical pregnancy and live birth. The proposed models may support individualized risk stratification and clinical decision-making.