Novel predictors for livebirth delivery rate in patients with idiopathic non-obstructive azoospermia based on the clinical prediction model

OBJECTIVE: To build a prediction model for live birth delivery per intracytoplasmic sperm injection (ICSI) in iNOA patients by obtaining sperm by microdissection testicular sperm extraction (mTESE). METHODS: A retrospective cohort study of 377 couples with iNOA male partners treated with 519 mTESE-ICSI cycles was conducted from September 2013 to July 2021 at the Reproductive Medical Centre of Peking University Third Hospital. Following exclusions, 377 couples with iNOA male partners treated with 482 mTESE-ICSIs were included. A prediction model for live birth delivery per ICSI cycle was built by multivariable logistic regression and selected by 10-fold cross-validation. Discrimination was evaluated by c-statistics and calibration was evaluated by the calibration slope. RESULTS: The live birth delivery rate per mTESE-ICSI cycle was 39.21% (189/482) in these couples. The model identified that the presence of motile sperm during mTESE, bigger testes, higher endometrial thickness on the day of human chorionic gonadotrophin (hCG) administration (ET-hCG), and higher quality embryos are associated with higher live birth delivery success rates. The results of the model were exported based on 10-fold cross-validation. In addition, the area under the mean ROC curve was 0.71 ± 0.05 after 10-fold cross-validation, indicating that the prediction model had certain prediction precision. A calibration plot with an estimated intercept of -1.653 (95% CI: -13.403 to 10.096) and a slope of 1.043 (95% CI: 0.777 to 1.308) indicated that the model was well-calibrated. CONCLUSION: Our prediction model will provide valuable information about the chances of live birth delivery in couples with iNOA male partners who have a plan for mTESE-ICSI treatment. Therefore, it can improve and personalize counseling for the medical treatment of these patients.