RESUMO
AIM: To develop a predictive model to discriminate renal oncocytoma (RO) from chromophobe renal carcinoma (chRCC) using multiphase computed tomography (CT). MATERIALS AND METHODS: Two hundred and five cases of renal tumours were analysed retrospectively regarding attenuation values during four CT phases, in addition to age, size, and sex. Then, logistic analysis was applied and a nomogram model developed to predict the most significant variables that can be used to differentiate between both tumour types. The cases were histopathologically proven as 81 cases of RO and 124 cases of chRCC. RESULTS: There was no association between the sex of the patient and the tumour types (p=0.41); however, there was a significant positive association between RO and the age of the patient (odds ratio 1.05; 95% confidence interval 1.02-1.08; p=0.001)) and a significant negative association between tumour size and RO (odds ratio 0.81; 95% confidence interval 73-90; p<0.001). There was a significant difference between tumour types in the contrast-enhanced phases. Logistic regression showed that absolute arterial enhancement (AAE) and absolute venous enhancement (AVE) are the most significant predictors for discriminating between tumour types. Combining these variables, size, AAE, and AVE were the best classifiers to discriminate between tumour types with an area under the curve of 0.90. A nomogram model was developed using these variables to predict RO probability in different case scenarios. CONCLUSION: The nomogram can predict the probability of RO from chRCC by using the best predictors, size, AAE, and AVE, with high accuracy.
