To externally validate the previously published Mayo clinic model for the prediction of early (<30 days) postoperative renal failure, which relies solely on preoperative estimated glomerular filtration rate (eGFR) and develop a novel model for the prediction of long-term (>30 days) renal function after partial nephrectomy (PN) and radical nephrectomy (RN), including patient factors and nephrometry scores.
Retrospective, single-center cohort study on patients who underwent PN or RN for a unilateral renal tumor between 2003 and 2019 with a preoperative eGFR of at least 15 ml/min/1.73m2. Early postoperative renal failure was defined as eGFR <15 ml/min/1.73 m2 or receipt of dialysis within 30 days. We determined the area under the receiver operating characteristics curve (AUC) to assess the Mayo clinic model's discriminative power. We used hierarchical linear mixed models with backward selection of candidate variables to develop a prediction model for long-term eGFR following PN and RN, separately. Their predictive ability was quantified using the marginal and conditional R2GLMM and an internal validation.
We included 421 patients (7,548 eGFR observations) who underwent PN and 271 patients (6,530 eGFR observations) who underwent RN. The Mayo clinic model for prediction of early postoperative renal failure following PN and RN showed an AUC of 0.816 (95% CI 0.718-0.920) and 0.825 (95% CI 0.688-0.962), respectively. In multivariable models, long-term eGFR following PN was associated with age, diabetes, the presence of a solitary kidney, tumor diameter and preoperative eGFR, while long-term eGFR following RN was associated with age, body mass index, RENAL nephrometry score and preoperative eGFR. Marginal and conditional R2GLMM were 0.591 and 0.855 for the PN model, and 0.363 and 0.849 for the RN model, respectively.
The Mayo clinic model for short-term renal failure prediction showed good accuracy on external validation. Our long-term eGFR prediction models depend mostly on host factors as opposed to tumor complexity and can aid in decision-making when considering PN vs. RN.
Urologic oncology. 2022 Nov 10 [Epub ahead of print]
Eduard Roussel, Annouschka Laenen, Bimal Bhindi, Anouk De Dobbeleer, Arthur Vander Stichele, Lien Verbeke, Ben Van Cleynenbreugel, Ben Sprangers, Benoit Beuselinck, Hendrik Van Poppel, Steven Joniau, Maarten Albersen
Department of Urology, University Hospitals Leuven, Leuven, Belgium., Institute for Biostatistics and Statistical Bioinformatics, Leuven, Belgium., Section of Urology, Department of Surgery, University of Calgary, Calgary, Alberta, Canada., Department of Microbiology, Immunology and Transplantation, Laboratory of Molecular Immunology (Rega Institute for Medical Research), KU Leuven, Leuven, Belgium; Department of Nephrology, University Hospitals Leuven, Leuven, Belgium., Department of Nephrology, University Hospitals Leuven, Leuven, Belgium; Department of General Medical Oncology, University Hospitals Leuven, Leuven, Belgium., Department of Urology, University Hospitals Leuven, Leuven, Belgium. Electronic address: .