Predicting Recurrence After Surgical Resection for High-Risk Localized Renal Cell Carcinoma: A Radiomics Clinical...

Adjuvant immunotherapy for clear cell renal cell carcinoma (ccRCC) is controversial because of the absence of reliable biomarkers for identifying patients most likely to benefit. The aim of this study was to develop and validate a quantitative radiomics signature (RS) and a radiomics clinical model to identify patients at increased risk of recurrence after surgery among those eligible for adjuvant immunotherapy.

This retrospective study included patients with ccRCC who are at intermediate to high risk or high risk of recurrence after nephrectomy. Inclusion criteria were patients with baseline characteristics matching the KEYNOTE-564 criteria. Radiomics texture features were extracted from preoperative CT scans. Affinity propagation clustering and random survival forest algorithms were applied to construct the RS. A radiomics clinical model was developed using multivariable Cox regression. The primary end point was disease-free survival (DFS). Model performance was assessed using time-dependent and integrated AUCs (iAUCs) and compared with conventional prognostic models using decision curve analysis.

A total of 309 patients were included, split into training (247) and test (62) sets. From each patient, 1316 radiomics features were extracted. The RS achieved an iAUC of 0.78 in the training set and 0.72 in the test set. Multivariable analysis identified node status, vascular invasion, hemoglobin, and the RS as predictors of DFS (all P < .05). These factors formed the radiomics clinical model, which achieved an iAUC of 0.81 (95% CI: 0.76-0.85) in the training set and 0.78 (95% CI: 0.69-0.88) in the test set. Decision curve analysis demonstrated its superior clinical utility compared with conventional prognostic models.

Integrating radiomics with clinical factors improves DFS prediction in intermediate-to-high-risk or high-risk ccRCC. This model offers a tool for individualized risk assessment, potentially optimizing patient selection for adjuvant therapy.