Predict Cancer-specific Survival After Nephrectomy for Nonmetastatic Renal Cancer: A Deep Learning-Based Prognostic Model

Abstract

Background There are few analyses comparing radical nephrectomy with resection of the renal parenchyma only (RNRP) or radical nephrectomy that includes simultaneous resection of the parenchyma, affected perirenal fascia, perirenal fat, and ureter (RNPU) relative to partial nephrectomy (PN) for patients with nonmetastatic (M0) renal cell carcinoma (RCC) in terms of cancer-specific survival (CSS). This study aimed to evaluate the effect of different nephrectomy on the CSS of nonmetastatic RCC (nmRCC) and to identify the main beneficiaries of different nephrectomy. Methods The data was collected from the Surveillance, Epidemiology and End Results (SEER) database. Kaplan-Meier plots, and multivariable Cox regression models were used. Propensity score matching (PSM) was performed to reduce the effect of selection bias. A prognostic model for nmRCC patients after nephrectomy was established using the deep learning framework. Results Kaplan-Meier analysis after PSM showed that lymph node dissection (LND) was effective in patients after RNRP (HR = 0.41, 95%CI: 0.27–0.64, p < 0.0001). RNRP demonstrated less strongly association with CSS than was PN (HR = 0.49, 95%CI༚0.34–0.71, p < 0.0001). The established prognostic model showed that grade II stage I T1N0M0 patients were the primary beneficiary population of RN. Conclusions RN is more recommended than PN for grade II stage I T1N0M0 RCC patients. LND is necessary when performing RNRP.