Use of predictive models for assessing perioperative outcomes of surgical treatment of localized renal parenchymal tumors: a critical review

Magomedov Ts.G., Sirota E.S.

Institute of Urology and Human Reproductive Health, Sechenov University, Moscow, Russia

The aim of this review was to critically evaluate the applicability of predictive models for planning nephron-sparing surgery in patients with localized renal parenchymal tumors. A narrative review of publications from 2020–2025 indexed in PubMed/MEDLINE, as well as earlier key studies, was performed. Nephrometry scores, regression models, machine learning, radiomics, and deep learning based on computed tomography data were analyzed. Nephrometry scores remain valuable as standardized tools for describing anatomical complexity but have limited ability to predict individual complications and composite outcomes. Multimodal models combining clinical and imaging features appear promising. A limited set of preoperative variables is often sufficient to predict postoperative renal function. Surgeon experience and the learning curve remain important factors but are frequently not incorporated into predictive models. Clinical implementation requires external validation, assessment of model calibration and clinical utility, while maintaining the leading role of the physician in clinical decision-making.

Keywords

About the Authors

Corresponding author: Ts.G. Magomedov – Postgraduate Student, Institute of Urology and Human Reproductive Health, Sechenov University, Moscow, Russia; e-mail: Sechenova052023@gmail.com

Similar Articles