Prediction of renal mass aggressiveness using clinical and radiographic features: a global, multicentre prospective study.

Objective: To examine the ability of preoperative clinical characteristics to predict histological features of renal masses (RMs).

Patients And Methods: Data from consecutive patients with clinical stage I RMs treated surgically between 2010 and 2011 in the Clinical Research Office of Endourology Society (CROES) Renal Mass Registry were collected. Based on surgical histology, tumours were categorised as benign, low- or high-aggressiveness cancer. Multivariate logistic regression was used to estimate the probability of the histological group by clinical and radiographic features in the entire cohort and a subcohort of cT1a tumours. The performance of the models was studied by calibration, Nagelkerke's R(2) , and discrimination (area under the receiver operating characteristic curve).

Results: The study cohort included 2 224 patients with a clinical stage I RM, of which 1 367 (61%) were cT1a. Benign lesions were found in 369 (16.6%), low-aggressiveness tumours in 1 156 (52%) and high-aggressiveness tumours in 699 (31.4%). Male gender, smoking history, increased tumour size, and lower exophytic rate were associated with malignancy and high-aggressiveness features (all P < 0.05). Models developed based on these characteristics had the ability to discriminate benign from malignant (bootstrap corrected c-index of 0.64) and high-aggressiveness tumours from benign and low-aggressiveness tumours (bootstrap corrected c-index of 0.66). Similar results were achieved in the cT1a subgroup. The c-index of tumour diameter as a single predictor of malignancy and high-aggressiveness tumours in the entire cohort was 0.6 and 0.63, respectively.

Conclusion: Although older age, male gender, smoking history, increased tumour diameter, and reduced exophytic rate are associated with malignancy and high aggressiveness of clinical stage I RMs, models incorporating these characteristics have modest discriminating power, being only slightly better than the predictive ability of tumour size alone.