Estimating Postoperative Renal Function After Surgery for Nonmetastatic Renal Masses: A Systematic Review of Available Prediction Models

Editorial Comment on External validation of Yonsei nomogram predicting chronic kidney disease development after partial nephrectomy: An international, multicenter study

“…1 During the last decade, a variety of prediction models have been proposed, yet no "ideal" one has been endorsed by the guidelines. 2 Of interest, a recent systematic literature review retrieved seven existing models created for the prediction of postoperative renal function following PN. 2 The Yonsei nomogram is among these "magnificent seven."…”

“…2 Of interest, a recent systematic literature review retrieved seven existing models created for the prediction of postoperative renal function following PN. 2 The Yonsei nomogram is among these "magnificent seven." It has been published in 2018 and developed on a cohort of 700 Korean patients.…”

“…As such, resection and renorrhaphy techniques in expert hands could perform and impact differently on renal function. This latter issue introduces the awareness provided by the systematic review performed, 2 which found no predictive model including provider-related factors, even if the impact of surgeon/center volume on renal function is still a matter of debate. 5 Aside from the aforementioned issues, this study represents a step in the right direction: external validation and assessment of clinical utility using formal statistical techniques is the way toward the generalization of a model to more patients and scenarios.…”

To clamp or not to clamp? No turning back now

“…This procedure was powerful and ensured an unbiased estimate of the predictive performance of the model. Khene et al [1] also suggested the use of deep neural networks to model pT3a upstaging by citing Heller et al [5] who developed a deep learning model to automatically segment CT-scan images and compute R.E.N.A.L. nephrometry scores.…”

“…Finally, Khene et al [1] challenged the use of Shapley Additive exPlanations (SHAP) values to explain patientspecific predictions in case of feature dependence. First, the reference is inappropriate (Fryer et al [6] questioned the use of Shapley values for feature selection purpose, and not for local interpretability purpose).…”