#0847

Computed Tomography Radiomics-Derived Nomogram for Predicting Early Renal Function Decline after Partial Nephrectomy in Renal Cell Carcinoma: A Multicenter Retrospective Development/Validation Study

X. Wu¹, S. Chen¹, N. ², X. Xue¹

¹The First Affiliated Hospital of Fujian Medical University, Fuzhou, China
²The First Affiliated Hospital of Fujian Medical University, Department of Urology, Fuzhou, China

Introduction:

Partial nephrectomy, as the standard surgical treatment for early-stage renal cell carcinoma, aims not only to remove renal tumors but also to preserve renal function. Consequently, early identification of renal function impairment is crucial for preventing further deterioration of renal function. Few investigations have explored the relationship between CT imaging features, radiomic characteristics, and postoperative renal function changes in renal tumor patients.

Material and methods:

Clinicopathologic characteristics and intraoperative parameters of 1019 patients with RCC who underwent partial nephrectomy from three different hospitals were retrospectively collected and analyzed. The LASSO was used to determine the radiomics features, and the results were compared, and univariate and multivariate logistic regression analyses were performed to identify six predictive factors affecting the postoperative decline in renal function, and based on these factors, a preoperative median line chart was constructed and its performance was evaluated by ROC curve analysis, and DeLong's test was used to compare the AUC values of different models. The calibration of the column plots was evaluated using the Hosmer-Lemeshow test and calibration curves. Their clinical utility was evaluated using decision curve analysis (DCA).

Results:

The LASSO algorithm and logistic regression analysis were used to screen the radiomics features associated with postoperative renal function changes and construct three signatures. ROC analysis showed that the combined radiomics signature had significantly higher predictive performance after both internal and external validation. Logistic regression analysis showed that age, diabetes, pre-eGFR, RENAL score, ischemia time and the Radiomics signature were independent predictors of early postoperative decline in renal function after PN in patients with RCC. Preoperative nomogram were constructed in the training group on the basis of five preoperative factors (AUC: 0.952, 95% CI, 0.916-0.960), and preoperative-intraoperative nomogram were constructed on the basis of six preoperative-intraoperative factors (AUC: 0.962, 95% CI, 0.941-0.977), which were internally and externally validated, and Hosmer-Lemeshow goodness-of-fit tests showed that the nomogram was well-corrected. DCA showed good clinical utility of the nomogram.