Objective: To examine the discriminant properties of the most contemporary version of the Tumour-Node-Metastasis (TNM) staging for renal cell carcinoma (RCC) sub-classification of T2 lesions according to a threshold size of 10 cm. Other thresholds were also assessed.
Patients and methods: Between 1988 and 2006, within the Surveillance, Epidemiology, and End Results database, patients with T2 N0-2 M0-1 RCC treated with a nephrectomy were abstracted. Tumour size was evaluated according to several thresholds: ≥8, ≥9, ≥10, ≥11, and ≥12 cm. Kaplan-Meier and life tables for cancer-specific mortality (CSM) were computed. Several Cox regression modes were fitted for prediction of CSM, using different thresholds. The predictive accuracy of various thresholds was compared using the area under the curve and methods of calibration.
Results: In all, 4963 patients were identified. Kaplan-Meier analyses showed statistically significant CSM-free survival differences between all examined thresholds. In multivariable Cox-regression models, all tested tumour size thresholds emerged as independent predictors of CSM. Of all thresholds, the values of 9 (0.55) and 11 cm (0.55) achieved the highest discrimination in univariable analysis, followed by 10 (0.539), 12 (0.539), and 8 cm (0.531). When the thresholds were combined with all other variables, the 11 cm (0.688) achieved the highest discrimination.
Conclusion: The discriminant properties of all examined thresholds showed very similar discriminant properties, which brings into questioning whether a dichotomization of pT2 tumours is really necessary.
© 2013 BJU International.