LAUSR

To identify specific imaging and clinicopathological features of a rare potentially malignant epithelioid variant of renal lipid-poor angiomyolipoma (E-lpAML). A total of 20 patients with E-lpAML and 43 patients with other lpAML were retrospectively included. Multiphase computed tomography (CT) imaging features and clinicopathological findings were recorded. Independent predictors for E-lpAML were identified using multivariate logistic regression and were used to construct a diagnostic score for differentiation of E-lpAML from other lpAML. The E-lpAML group consisted of 6 men and 14 women (age median ± SD: 39.45 ± 15.70, range: 16.0–68.0 years). E-lpAML tended to appear as hyperdense mass lesions located at the renal sinus (n = 8, 40%) or at the renal cortex (n = 12, 60%), with a “fast-in and slow-out” enhancement pattern (n = 20, 100%), cystic degeneration (n = 18, 90%), “eyeball” sign (n = 11, 55%), and tumor neo-vasculature (n = 15, 75%) on CT. Multivariate logistic regression analysis showed that the independent predictors for diagnosing E-lpAML were cystic degeneration on CT imaging and CT value of the tumor in corticomedullary phase of enhancement. A predictive model was built with the two predictors, achieving an area under the curve (AUC) of 93.5% (95% confidence interval (95%CI): 84.3–98.2%) with a sensitivity of 95.0% (95%CI: 75.1–99.9%) and a specificity of 83.72% (95%CI: 69.3–93.2%). We identified specific CT imaging features and predictors that could contribute to the correct diagnosis of E-lpAML. Our findings should be helpful for clinical management of E-lpAML which could potentially be malignant and may require nephron-sparing surgery while other lpAML tumors which are benign require no intervention. • It is important to differentiate renal epithelioid lipid-poor angiomyolipoma (E-lpAML) from other lpAML because of differences in clinical management. • E-lpAML tumors tend to be large hyperdense tumors in the renal sinus with cystic degeneration and “fast-in and slow-out” pattern of enhancement. • Our CT imaging-based predictive model was robust in its performance for predicting E-lpAML from other lpAML tumors.