Reply: Predicting sperm extraction in non-obstructive azoospermia patients: a machine-learning perspective

Based on our experience, and supported by the literature in the field, we sincerely doubt that the surgical approach to sperm retrieval as described by the authors resulted in such a high SSR that was reported.The authors performed only a small equatorial horizontal incision on the albuginea, with 8-10 tiny scissors biopsies from the extruded parenchyma, eventually followed by contralateral multiple conventional superficial biopsies: due to the high heterogeneity of testicular spermatogenesis in patients with NOA, in most cases, the testes hide only isolated, focal areas of residual spermatogenesis that could be hardly identified with superficial biopsies.Even the more invasive approach used by the authors by means of multiple testicular biopsies may fail to retrieve sperm in patients with NOA: however, such an approach might result in the loss of significant amounts of testicular tissue, with risks of testicular devascularization and subsequent testicular hypo-atrophy (Schlegel and Su, 1997) and/or hypogonadism.Management of NOA patients cannot prescind from the use of the best surgical treatment available: given the existent Level 1 evidence in well-performed meta-analyses (Deruyver et al., 2014;Bernie et al., 2015) as previously highlighted (Esteves et al., 2020) is it undoubtful that microdissection testicular sperm extraction provides optimized sperm retrieval results, especially in the hands of well-trained and experienced urologists, since it enables the identification at high magnification of 'dilated' tubules with preserved spermatogenesis, at least in 90% of cases (Caroppo et al., 2019).When a less-effective surgical treatment is used, even sophisticated machine-learning models cannot add to the counselling of patients with NOA.