Development and external validation of a predictive model for lung metastases in kidney cancer based on clinical and laboratory features.

Background: Lung metastases (LMs) are common and clinically significant in kidney cancer, contributing to poor prognosis and limited treatment response. Early detection of patients at high risk for LMs is crucial for developing personalized treatment strategies.

Methods: This retrospective study included 2652 patients with kidney cancer, pathologically confirmed, from two medical centers between 2000 and 2020. A training cohort (n = 1500) was used to develop a predictive model for LMs based on clinical, pathological, and laboratory variables. Multivariate logistic regression was applied to identify independent predictors, and a nomogram was constructed. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration plots, probability density functions (PDFs), and decision curve analysis (DCA). An external validation cohort (n = 1152) was used to assess generalizability.

Results: Family cancer history, advanced T stage, and N nodal stage were identified as independent predictors of LMs. The nomogram demonstrated excellent discrimination in the training cohort (AUC = 0.938) and moderate performance in the validation cohort (AUC = 0.883), outperforming individual variables. Calibration analysis showed good agreement between predicted and observed probabilities. Patients with LMs had significantly worse overall survival than those without (p < 0.001). DCA validated the nomogram's greater clinical value over treat-all or treat-none approaches.

Conclusion: This nomogram-based model offers an efficient and accessible tool for the early detection of kidney cancer patients at risk for LMs. It enables individualized risk stratification and may assist in optimizing clinical decision-making and improving patient outcomes. Further prospective validation is warranted.