A new machine-learning model to predict long-term renal function impairment after minimally invasive partial nephrectomy: the Fundacio Puigvert predictive model

PurposeTo provide a new model to predict long-term renal function impairment after partial nephrectomy (PN).MethodsData of consecutive patients who underwent minimally invasive PN from 2005 to 2022 were analyzed. A minimum of 12 months of follow-up was required. We relied on a machine-learning algorithm, namely classification and regression tree (CART), to identify the predictors and associated clusters of chronic kidney disease (CKD) stage migration during follow-up.Results568 patients underwent minimally invasive PN at our center. A total of 381 patients met our inclusion criteria. The median follow-up was 69 (IQR 38-99) months. A total of 103 (27%) patients experienced CKD stage migration at last follow-up. Progression of CKD stage after surgery, ACCI and baseline CKD stage were selected as the most informative risk factors to predict CKD progression, leading to the creation of four clusters. The progression of CKD stage rates for cluster #1 (no progression of CKD stage after surgery, baseline CKD stage 1-2, ACCI 1-4), #2 (no progression of CKD stage after surgery, baseline CKD stage 1-2, ACCI & GE; 5), #3 (no progression of CKD stage after surgery and baseline CKD stage 3-4-5) and #4 (progression of CKD stage after surgery) were 6.9%, 28.2%, 37.1%, and 69.6%, respectively. The c-index of the model was 0.75.ConclusionWe developed a new model to predict long-term renal function impairment after PN where the perioperative loss of renal function plays a pivotal role to predict lack of functional recovery. This model could help identify patients in whom functional follow-up should be intensified to minimize possible worsening factors of renal function.

Automated summary

This study aimed to provide a new model for predicting long-term renal function impairment after partial nephrectomy (PN). The researchers used a machine-learning algorithm to identify predictors and associated clusters of chronic kidney disease (CKD) stage migration. They found that the progression of CKD stage after surgery, ACCI and baseline CKD stage were the most informative risk factors for predicting CKD progression.