Angiomyolipoma with minimal fat: differentiation from renal cell carcinoma at biphasic helical CT.

PURPOSE: To compare various computed tomographic (CT) features of angiomyolipoma (AML) with minimal fat with those of size-matched renal cell carcinoma (RCC). MATERIALS AND METHODS: Eighty-one patients (19 with AML with minimal fat [mean diameter, 2.8 cm; range, 1.5-4.5 cm] and 62 with RCC [mean diameter, 3.1 cm; range, 1.8-4.5 cm]) who had undergone biphasic CT (ie, CT with unenhanced, corticomedullary, and early excretory phase scanning) were evaluated. Two reviewers who were unaware of the diagnosis retrospectively recorded tumor attenuation on unenhanced scans, enhancement characteristics (ie, homogeneity of enhancement, amount of enhancement, enhancement pattern over time), tumor margin, location of tumor center, intratumoral calcification, perinephric changes, and patient age and sex. The predictive value of each CT finding was determined by using multivariate logistic regression analysis. RESULTS: Homogeneous enhancement (observed in 79% of AMLs vs 5% of RCCs; odds ratio, 37) and prolonged enhancement pattern (observed in 58% of AMLs vs 10% of RCCs; odds ratio, 42) were valuable predictors for differentiating AML with minimal fat from RCC at multivariate analysis (P <.05 for both). When both CT findings were used as a criterion for differentiating AML from RCC, positive and negative predictive values were 91% (10 of 11 tumors) and 87% (61 of 70 tumors), respectively. Fifty-three percent of AMLs versus 13% of RCCs showed high tumor attenuation on unenhanced scans (P =.04), whereas RCCs showed greater mean enhancement than AMLs (114 HU +/- 44 [SD] vs 73 HU +/- 30 in corticomedullary phase and 66 HU +/- 24 vs 49 HU +/- 20 in early excretory phase) and a male predominance (male-to-female ratio, 50:12 vs 8:11; P =.001). CONCLUSION: Biphasic helical CT may be useful in differentiating AML with minimal fat from RCC, with homogeneous tumor enhancement and prolonged enhancement pattern being the most valuable CT findings.