Machine learning studies for predicting 5-year renal cell cancer survival: a multicenter study.

Background: Renal cell carcinoma (RCC) is biologically heterogeneous, complicating prognosis and treatment. Our study aims to clarify prognostic factors and support evidence-based clinical decision-making in RCC management. Methods: This study analyzed clinical data of RCC patients diagnosed between 2010 and 2021 from the Surveillance, Epidemiology, and End Results database. Prognostic factors associated with patient outcomes were selected using least absolute shrinkage and selection operator regression and Boruta. Seven machine learning algorithms were employed to construct predictive models for 5-year survival in RCC patients. Shapley Additive Explanations (SHAP) analysis was utilized to visually interpret the model's predictions. Model performance was evaluated using the following metrics: accuracy, sensitivity, specificity, positive predictive value, negative predictive value, F1 score, area under the receiver operating characteristic curve (AUC), Brier score, and estimated calibration index (ECI). Results: In the realm of predicting 5-year survival for RCC patients, the gradient boosting machine (GBM) model achieved the best overall performance with an AUC of 0.841 [95% confidence interval (CI): 0.833-0.850], along with the lowest Brier score (0.127; 95% CI: 0.126-0.128) and ECI (0.005; 95% CI: 0.003-0.007). Furthermore, SHAP analysis highlighted key clinical factors influencing 5-year survival of RCC patients, reinforcing their prognostic significance in RCC. Conclusions: This study developed an accurate predictive model for 5-year survival in RCC patients using machine learning, which may offer valuable support to clinicians in making informed clinical decisions.