Purpose: To determine the extent to which radiologists and urologists can predict histology using multiphasic CT imaging.
Methods: Patients with a preoperative multiphasic CT undergoing surgery for a renal mass were identified between 2003 and 2013. Tumors >10 cm, locally advanced or metastatic disease, and patients managed by reviewers were excluded. A survey and deidentified scans were provided to reviewers. Sensitivity and accuracy in predicting histology was calculated for each reviewer. Correlation was assessed by the Fleiss kappa coefficient. Multivariable logistic regression determined factors associated with predictive accuracy for final pathology.
Results: There were 120 patients who met criteria. Mean tumor size was 3.3 cm; there were 102 (85%) that were malignant, and 73% of these were clear-cell renal-cell carcinoma (RCC). The most common benign histology was angiomyolipoma (n=10, 56%) followed by oncocytoma (n=5, 28%). Correlation among reviewers was statistically fair for predicting malignant (κ=0.25) and final pathology (κ=0.22). Sensitivity for predicting malignant masses was 90%. Reviewers accurately predicted malignant pathology in 82% of cases and predicted final pathology in 58% of cases. Adjusted for size, scan type, and reviewer, clear-cell RCC vs benign histology was associated with 21 times increased odds of accurate pathologic identification (P<0.001).
Conclusions: Urologists and radiologists were able to accurately identify malignant histology in 82% of cases, although sensitivity for malignant histology was 90%. Developing a preoperative nomogram for identification of clear-cell RCC may be feasible and should be further explored.