Development of a predictive model for increasing sperm retrieval success by microdissection testicular sperm extraction in patients with nonobstructive azoospermia.

Abstract

Microdissection testicular sperm extraction (micro-TESE) is widely used to treat nonobstructive azoospermia. However, a good prediction model is required to anticipate a successful sperm retrieval rate before performing micro-TESE. This retrospective study analyzed the clinical records of 200 nonobstructive azoospermia patients between January 2021 and December 2021. The backward method was used to perform binary logistic regression analysis and identify factors that predicted a successful micro-TESE sperm retrieval. The prediction model was constructed using acquired regression coefficients, and its predictive performance was assessed using the receiver operating characteristic curve. In all, 67 patients (sperm retrieval rate: 33.5%) underwent successful micro-TESE. Follicle-stimulating hormone, anti-Müllerian hormone, and inhibin B levels varied significantly between patients who underwent successful and unsuccessful micro-TESE. Binary logistic regression analysis yielded the following six predictors: anti-Müllerian hormone (odds ratio [OR] = 0.902, 95% confidence interval [CI]: 0.821-0.990), inhibin B (OR = 1.012, 95% CI: 1.001-1.024), Klinefelter's syndrome (OR = 0.022, 95% CI: 0.002-0.243), Y chromosome microdeletion (OR = 0.050, 95% CI: 0.005-0.504), cryptorchidism with orchiopexy (OR = 0.085, 95% CI: 0.008-0.929), and idiopathic nonobstructive azoospermia (OR = 0.031, 95% CI: 0.003-0.277). The prediction model had an area under the curve of 0.720 (95% CI: 0.645-0.794), sensitivity of 65.7%, specificity of 72.2%, Youden index of 0.379, and cut-off value of 0.305 overall, indicating good predictive value and accuracy. This model can assist clinicians and nonobstructive azoospermia patients in decision-making and avoiding negative micro-TESE results.