Personalized survival predictions in chromophobe renal cell carcinoma: development of a machine learning-based web...

Chromophobe renal cell carcinoma (ChRCC) is a rare subtype of renal cancer, characterized by distinct clinical and genetic features. Existing studies on ChRCC are limited, and there is a critical need to explore the prognostic factors and treatment outcomes in this patient population. We used machine learning (ML) to build prognostic models and developed the first predictive web-based tool for survival.

The SEER database (2000-2020) was used for this study's analysis. To identify the prognostic variables, we conducted Cox regression analysis and constructed prognostic models using five ML algorithms to predict the 5-year survival. A validation method incorporating the area under the curve (AUC) of the receiver operating characteristic (ROC) curve was used to validate the accuracy and reliability of ML models. We also performed Kaplan-Meier survival analysis.

Our study analyzed 10,700 patients with ChRCC and identified metastasis and tumor size as significant predictors of survival. Subtotal nephrectomy was associated with the highest survival rates. Chemotherapy and radiotherapy were infrequently used but were associated with worse survival outcomes, particularly in patients with metastasis. The developed ML models demonstrated high accuracy in predicting survival, and a web-based tool offered real-time survival predictions based on patient-specific data.

Our study identified key prognostic factors and developed a machine learning-based web tool for personalized survival predictions. Metastasis and tumor size are critical in determining patient outcomes, with subtotal nephrectomy showing the highest survival rate.