Comparative performance of predictive nomograms for pathological upstaging in clinical T1 renal masses: an independent external validation

BACKGROUND: Preoperative identification of pathological upstaging (≥ pT3a) in clinical T1 (cT1) renal masses is critical for surgical planning. We externally validated three established predictive models, Barra, Nocera, and Mei, to assess their discrimination, calibration, and clinical utility in a contemporary cohort. PATIENTS AND METHODS: We retrospectively reviewed patients undergoing surgery for cT1 renal masses at our institution. The primary endpoint was pathological upstaging to pT3a. Model performance was evaluated using the Area Under the Curve (AUC) for discrimination and the Brier Score for overall accuracy. Calibration was assessed using parametric and nonparametric Loess-smoothed curves. Clinical utility was quantified using Standardized Decision Curve Analysis. RESULTS: The institutional upstaging rate was 17.1% (19/111), and clear cell histology predominated (77.5%). Nocera (AUC = 0.736) and Barra (AUC = 0.733) had higher point estimates of discriminative ability than Mei (AUC = 0.677), but the differences were not significant. All models descriptively overestimated absolute risk. However, Mei demonstrated the highest absolute calibration stability, whereas the Nocera model showed significant global unreliability (U = 0.052, p = 0.021). The Barra model achieved the lowest overall error (Brier Score = 0.122) and the highest clinical net benefit within the decisive 8%-18% decision window, although the overall accuracy differences were not statistically significant. CONCLUSION: All nomograms are useful risk-stratification adjuncts, but wide confidence intervals and low event counts mean these findings are exploratory. No single model is definitively superior. Nocera and Barra are effective for relative risk-ranking, while Mei provides superior absolute precision for patient counseling.