Are There Useful CT Features to Differentiate Renal Cell Carcinoma From Lipid-Poor Renal Angiomyolipoma?

Genitourinary Imaging Original Research Yang et al. Imaging Differentiation of Benign and Malignant Renal Tumors Genitourinary Imaging Original Research Ching-Wei Yang 1,2 Shu-Huei Shen 3,4 Yen-Hwa Chang 1,2,3 Hsiao-Jen Chung 1,2,3 Jia-Hwia Wang 3,4 Alex TL Lin 1,2,3 Kuang-Kuo Chen 1,2,3 Yang CW, Shen SH, Chang YH, et al. Keywords: CT, fat-minimal angiomyolipoma, lipid-poor angiomyolipoma, renal cell carcinoma subtypes, renal cell tumor DOI:10.2214/AJR.12.10204 Received October 23, 2012; accepted after revision April 12, 2013. 1 Division of Urology, Department of Surgery, Taipei Veterans General Hospital, Taipei, Taiwan. 2 Yang-Ming Shu-Tien Urological Science Research Center, Taipei, Taiwan. 3 Department of Radiology, Taipei Veterans General Hospital, 201, Sec 2, Shih-Pai Rd, Taipei 112, Taiwan. Address correspondence to S. H. Shen (shshen@vghtpe.gov.tw). 4 National Yang Ming University School of Medicine, Taipei, Taiwan. AJR 2013; 201:1017 1028 0361 803X/13/2015 1017 American Roentgen Ray Society Are There Useful CT Features to Differentiate Renal Cell Carcinoma From Lipid-Poor Renal Angiomyolipoma? OBJECTIVE. This study was an attempt to identify key CT features that can potentially be used to differentiate between lipid-poor renal angiomyolipoma and renal cell carcinoma (RCC). MATERIALS AND METHODS. We conducted an analysis of patients who received nephrectomy or renal biopsy from 2002 to 2011 with suspected RCC. We included tumors smaller than 7 cm with a completed three-phase CT examination. A radiologist and a urology fellow, blinded to histopathologic diagnosis, recorded the imaging findings by consensus and compared the values for each parameter between lipid-poor angiomyolipoma, RCC subtypes, and RCC as a group. Multivariate logistic regression analysis was performed for each univariate significant feature. RESULTS. The sample in our study consisted of 132 patients with 135 renal tumors, including 51 men (age range, 26 84 years; mean age, 57 years) and 81 women (age range, 29 91 years; mean age, 57 years). These tumors included 33 lipid-poor angiomyolipomas, 54 clear-cell RCC, 31 chromophobe RCC, and 17 papillary RCC. Multivariate analysis revealed four significant parameters for differentiating RCC as a group from lipid-poor angiomyolipoma (angular interface, p = 0.023; hypodense rim, p = 0.045; homogeneity, p = 0.005; unenhanced attenuation > 38.5 HU, p < 0.001), five for clear-cell RCC, two for chromophobe RCC, and one for papillary RCC. Lipid-poor angiomyolipoma and clear-cell RCC showed early strong enhancement and a washout pattern, whereas chromophobe RCC and papillary RCC showed gradual enhancement over time. CONCLUSION. Specific CT features can potentially be used to differentiate lipid-poor renal angiomyolipoma from renal cell carcinoma. R enal angiomyolipoma is the most common type of benign renal tumor and typically consists of smooth muscle, blood vessels, and adipose tissue. The CT diagnosis of angiomyolipoma relies on the detection of areas of macroscopic fat, with negative attenuation measurements [1]. However, no fat can be visualized on CT scans in approximately 4.5% of angiomyolipoma. These tumors are referred to as lipid-poor renal angiomyolipomas [2, 3]. Patients with lipid-poor angiomyolipoma usually receive unnecessary surgery for suspected renal cell carcinoma (RCC) when the diagnosis is not specifically established prospectively. The introduction of MDCT has greatly improved the characterization and diagnosis of RCC [4]. Radiologists have investigated imaging techniques, including contrast enhancement patterns on CT [5], CT histogram analysis [6 8], and specific MRI techniques [9 11] that can potentially be used to differentiate lipid-poor angiomyolipoma from RCC. However, these methods usually involve complex analyses or are not prospectively reliable enough to differentiate an individual lipid-poor angiomyolipoma from RCC. There are several subtypes of RCC, and the three most common are clear cell, chromophobe, and papillary types. Each type has its own imaging features, clinical aggressiveness, and prognosis [12, 13]. Previous researchers have generally studied only clearcell RCC or analyzed all RCC together to compare it with lipid-poor angiomyolipoma. We therefore retrospectively compared the CT findings for lipid-poor angiomyolipoma with the three subtypes of RCC and all RCC as a group to identify the imaging features that can potentially be used to differentiate benign from malignant renal tumors. AJR:201, November 2013 1017

Yang et al. Materials and Methods Patient Selection Our retrospective study was approved by our hospital s institutional review board. We retrieved the data of patients who underwent nephrectomy or renal biopsy with the preoperative diagnosis of a potentially malignant renal mass from 2002 to 2011. We included tumors smaller than 7 cm with a completed three-phase CT examination but excluded cases of angiomyolipoma with detectable gross fat on unenhanced CT (i.e., the fat was prospectively unrecognized). We also excluded epithelioid angiomyolipoma; oncocytoma; cystic RCC (cystic lesion with mural nodule or wall thickening); ruptured, recurrent, or metastatic tumors; and end-stage renal disease. CT Examinations Three-phase CT examinations were performed on a 16-MDCT scanner (Somatom Sensation 16, Siemens Healthcare). A 100-mL volume of nonionic contrast agent (iopamidol, Iopamiro 370, Bracco) was administered by a power injector at a rate of 2 ml/s. The scanning protocol included data acquisition in three phases: the unenhanced phase, the corticomedullary phase (30-second delay after contrast injection), and the nephrographic phase (90-second delay after contrast injection). The scanning parameters were as follows: pitch, 1.5; x-ray tube voltage, 120 kv; and tube current, 210 240 ma. The slice thickness of axial and coronal images was 5 mm. Coronal multiplanar reconstruction imaging was routinely performed for both the corticomedullary phase and the nephrographic phase. Image Analysis A radiologist with more than 10 years of experience in genitourinary radiology and a urology fellow with 1 year of training in genitourinary radiology) in charge of our study were blinded to the pathologic results. They reviewed all selected preoperative images on a PACS monitor (> 3 megapixels) together in single session. Age, sex, and the following image parameters were recorded for each patient by consensus: laterality (right or left); location (exophytic, meaning > 50% outside renal parenchyma; intraparenchymal, meaning > 50% within renal parenchyma; or endophytic, meaning > 50% protruding into renal sinus); shape (round or not round; lobulated or elongated); cystic component (water density in unenhanced phase; nonenhanced in contrast-enhanced phase); calcified component; angular interface [14] (exophytic renal mass with tapering pyramidal interface with a definable apex within the parenchyma); hypodense rim (a low-density rim at the peripheral area of the tumor in unenhanced CT that contrasts with the adjacent normal kidney parenchyma); coexisting lipid-rich angiomyolipoma (lipid-rich angiomyolipoma, other macroscopic fat-containing renal mass identified in either kidney); and perinephric collateral vessels [13] (engorged vessel in perinephric area). These specific CT findings are illustrated in Figure 1. The tumor attenuation in the unenhanced phase was measured. After administering the contrast agent, we assessed the following features: enhancement pattern (homogeneous or heterogeneous), tumor attenuation in the corticomedullary phase and the nephrographic phase, amount of tumor enhancement in the corticomedullary phase and the nephrographic phase (unenhanced phase attenuation subtracted from corticomedullary phase / nephrographic phase attenuation), and enhancement over time (corticomedullary phase attenuation subtracted from nephrographic phase attenuation). The two physicians assessed the homogeneity of tumor enhancement by visual inspection until reaching consensus. To measure the attenuation, the radiologist selected three regions of interest (ROIs) for each lesion and recorded their mean values. Each ROI was placed over an enhancing solid area within the tumor, which had to be at least 20 mm 2 [15] and consistent in location and size on images obtained during all three scanning phases. If a tumor showed heterogeneous attenuation, we measured the area with the greatest attenuation. Statistical Analysis To identify which CT features could potentially be used to differentiate between lipid-poor angiomyolipoma, RCC subtypes, and all RCC as a group, we looked for statistically significant morphologic parameters and enhancement pattern differences. We compared the values for each parameter using the chi-square test and Fisher exact test for categoric variables, and the Student t test and Mann-Whitney U test for continuous variables. We performed all analyses with SPSS (version 18.0) and considered a p value of less than 0.05 to be statistically significant. The imaging parameters that emerged as significant differentiators in univariate analysis were used in multivariate logistic regression analysis with stepwise methods. For significant variables, we calculated the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy. Results In our study, 568 patients with RCC and 111 patients with angiomyolipoma underwent nephrectomy, and six patients with angiomyolipoma underwent renal biopsy. The study sample included 102 RCC and 33 lipid-poor angiomyolipoma patients who met the selection criteria with complete imaging examinations. Of the 102 RCC patients, 54 (53%) had clear-cell RCC, 31 (30%) had chromophobe RCC, and 17 (17%) had papillary RCC. Table 1 presents a summary of the results of the univariate analysis comparing the clinical and imaging parameters between lipid-poor angiomyolipoma, the three RCC subtypes, and all RCC. Age and Sex The lipid-poor angiomyolipoma group had a significantly younger age and female predominance compared with the clear-cell RCC group, whereas the lipid-poor angiomyolipoma and chromophobe RCC groups exhibited no significant differences in either age or sex. When compared with all RCC, the female predominance was significant, but the age difference was insignificant. Morphologic CT Image Parameters The mean tumor size (± SD) was 2.8 ± 1.3 cm (range, 1.0 5.8 cm) for lipid-poor angiomyolipoma and 3.7 ± 1.4 cm for all RCC (range, 1.5 6.7 cm) (p = 0.001). The selected cases of clear-cell RCC showed comparable sizes (3.1 ± 0.9 cm) without significant difference with lipid-poor angiomyolipoma (p = 0.165). The chromophobe RCC tumors had a mean diameter of 4.3 cm, and papillary RCC had a mean diameter of 4.7 cm, both of which were larger than that of the lipid-poor angiomyolipoma group (both p < 0.001). Compared with the clear-cell RCC group, the lipid-poor angiomyolipoma group exhibited more exophytic location (67% vs 39%, p = 0.031), angular interface (39% vs 2%, p < 0.001), hypodense rim (27% vs 2%, p = 0.001), and coexisting lipid-rich angiomyolipoma (27% vs 2%, p = 0.001) and less round shape (52% vs 83%, p = 0.001) (Figs. 2 and 3). None of the lipid-poor angiomyolipomas had calcification or a peripheral collateral vessel, but 6% of clear-cell RCCs displayed calcification and a quarter of the clear-cell RCCs had collateral vessels, indicating neovascularity. After contrast administration, the majority of the lipid-poor angiomyolipomas showed a homogeneous enhancement pattern (97%), and the majority of clear-cell RCCs showed heterogeneous enhancement (87%, p < 0.001) (Fig. 4). Compared with chromophobe RCC, lipidpoor angiomyolipomas were also more likely to display an exophytic location (p = 0.026), 1018 AJR:201, November 2013

Imaging Differentiation of Benign and Malignant Renal Tumors Table 1: Comparison of Clinical and CT Features Between Lipid-Poor Angiomyolipoma, Renal Cell Carcinoma (RCC) Subgroups, and All RCC Lipid-Poor Parameter Angiomyolipoma (n = 33) Clear-Cell RCC (n = 54) p Chromophobe RCC (n= 31) p Papillary RCC (n = 17) p All RCC (n = 102) p Mean age ± SD (y) 52.9 ± 12.8 60.1 ± 15.3 0.028 a 52.9 ± 12.1 0.999 a 63.2 ± 14.0 0.01 b 58.4 ± 14.6 0.055 a Mean tumor size ± SD (cm) 2.8 ± 1.3 3.1 ± 0.9 0.165 a 4.3 ± 1.5 0.001 a 4.7 ± 1.8 0.001 b 3.7 ± 1.4 0.001 b Female sex 20 (60.6) 14 (25.9) 0.002 13 (41.9) 0.132 5 (29.4) 0.072 32 (31.4) 0.004 Right laterality 16 (48.5) 27 (50.0) 1.0 16 (51.6) 0.802 11 (64.7) 0.372 54 (52.9) 0.692 Location Exophytic 22 (66.7) 21 (38.9) 0.031 c 11 (35.4) 0.026 c 8 (47.1) 0.307 c 4 (47.1) 0.016 c Intraparenchymal 10 (30.3) 26 (48.1) 14 (45.2) 8 (47.1) 40 (39.2) Endophytic 1(3.0) 7 (13.0) 6 (19.4) 1 (5.8) 12 (13.7) Shape Round 17 (51.5) 45 (83.3) 0.001 18 (58.1) 0.599 9 (52.9) 1.0 72 (70.6) 0.44 Nonround 16 (48.5) 9 (16.7) 13 (41.9) 8 (47.1) 30 (29.4) Cystic component 5 (15.2) 7 (13.0) 0.774 3 (9.7) 0.508 3 (17.6) 1.00 13 (12.7) 0.770 Calcified component 0 (0) 3 (5.6) 0.168 c 2 (6.5) 0.231 c 4 (23.5) 0.01 c 9 (8.8) 0.077 c Angular interface 13 (39.4) 1 (1.9) < 0.001 2 (6.5) 0.002 2 (11.8) 0.056 5 (4.9) < 0.001 Hypodense rim 9 (27.3) 1 (1.9) 0.001 c 1 (3) 0.002 c 0 (0) 0.02 c 2 (2.0) < 0.001 c Coexisting lipid-rich angiomyolipoma 9 (27.3) 1 (1.9) 0.001 c 2 (6.5) 0.027 c 2 (11.8) 0.292 c 5 (4.9) 0.01 c Perinephric collateral vessels 0 (0) 14 (25.9) 0.01 9 (29.0) 0.01 2 (11.8) 0.111 25 (24.5) 0.001 Enhancement pattern Homogeneous 30 (96.8) 7 (13.0) 0.001 c 27 (87.1) 0.354 13 (76.5) 0.047 c 47 (46.1) < 0.001 c Heterogeneous 1 (3.2) 47 (87.0) 4 (12.9) 4 (23.5) 55 (53.9) Note Data in parentheses are percentages. a Student t test. b Mann Whitney test. c Fisher exact test. angular interface (p = 0.002), hypodense rim (p = 0.002), and coexisting lipid-rich angiomyolipoma (p = 0.027). Perinephric collateral vessels were found in one third of the chromophobe RCC group (p = 0.01) (Fig. 5). However, unlike clear-cell RCC, nearly 40% of the chromophobe RCCs had a nonround shape, and 87% of them showed a homogeneous enhancement pattern (p = 0.354 compared with lipid-poor angiomyolipoma). The papillary RCCs did not differ significantly from the lipid-poor angiomyolipomas in location and shape. The papillary RCCs were more likely to exhibit calcification, with 24% of cases doing so (p = 0.01) (Fig. 6). Although the papillary RCCs were less likely than the lipid-poor angiomyolipomas to display coexisting lipid-rich angiomyolipomas (12%) and more likely to have collateral vessels (12%), these features did not differ significantly between the two groups. Although 77% of the papillary RCCs showed a homogeneous enhancement pattern, there was a still greater chance of such pattern being seen in lipid-poor angiomyolipomas and statistical analysis confirmed the significance (p = 0.047). Comparing lipid-poor angiomyolipoma and all RCC as a group showed that exophytic location (p = 0.016), angular interface (p < 0.001), hypodense rim (p < 0.001), coexisting lipid-rich angiomyolipoma (p = 0.01), lack of perinephric collaterals (p = 0.001), and homogeneous enhancement pattern (p < 0.001) were significant predictors for lipidpoor angiomyolipoma. Multiphasic CT Enhancement Pattern Table 2 and Figure 7 show a comparison of three-phase enhancement between lipidpoor angiomyolipomas and RCC subtypes. The average tumor attenuation in the unenhanced phase of lipid-poor angiomyolipoma was significantly higher than that of any of the RCC subgroups and all RCC (p < 0.001 for all four groups). The threshold for each RCC group was determined by reviewing the results for receiver operating characteristic (ROC) curve analysis, which were as follows: 37.5 HU for clear-cell RCC (area under the curve [AUC] = 0.902); 39.5 HU for chromophobe RCC (AUC = 0.854); 39.5 HU for papillary RCC (AUC = 0.816), and 38.5 for all RCC (AUC = 0.873). The enhancement pattern over time was similar in clear-cell RCC and lipid-poor angiomyolipoma, with prominent enhancement in the corticomedullary phase and a slight washout in the nephrographic phase. Tumor attenuation in the nephrographic phase was significantly lower for lipid-poor angiomyolipoma than for clear-cell RCC (p = 0.026). The average amount of tumor enhancement from the unenhanced phase to the corticomedullary phase was 68 HU in lipid-poor angiomyolipoma and 86 HU in clear-cell RCC, with the difference not statistically significant (p = 0.066). The average tumor enhancement over time (nephrographic phase- AJR:201, November 2013 1019

Yang et al. Table 2: Comparison of CT Enhancement Pattern Among Lipid-Poor Angiomyolipoma, Renal Cell Carcinoma (RCC) Subgroups, and All RCC CT Enhancement Lipid-Poor Angiomyolipoma Clear-Cell RCC p Chromophobe RCC p Papillary RCC p All RCC p Tumor attenuation (HU) Unenhanced phase 43 ± 6 30 ± 8 < 0.001 33 ± 7 < 0.001 35 ± 7 < 0.001 32 ± 8 < 0.001 Corticomedullary phase 111 ± 26 118 ± 43 0.426 85 ± 29 0.001 52 ± 22 < 0.001 98 ± 44 0.154 Nephrographic phase 97 ± 21 113 ± 34 0.026 96 ± 32 0.83 62 ± 17 < 0.001 100 ± 36 0.686 Amount of tumor enhancement a (HU) Corticomedullary phase minus 68 ± 26 86 ± 44 0.066 51 ± 26 0.02 17 ± 16 < 0.001 66 ± 44 0.773 unenhanced phase Nephrographic phase minus 53 ± 19 83 ± 34 0.001 59 ± 36 0.509 26 ± 14 < 0.001 67 ± 38 0.069 unenhanced phase Tumor enhancement over time b (HU) Nephrographic phase minus corticomedullary phase 14 ± 21 3 ± 38 0.170 11 ± 38 0.006 6 ± 16 0.006 2 ± 36 0.034 Note Data are mean attenuation values ± SD. Statistical method was independent Student t test. a Enhanced attenuation minus unenhanced phase. b Nephrographic phase minus corticomedullary phase. Table 3: Multivariate Logistic Regression Analyses of CT Parameters Differentiating Lipid-Poor Angiomyolipoma, Renal Cell Carcinoma (RCC) Subgroups, and All RCC Parameters Favoring Lipid-Poor Angiomyolipoma Clear-Cell RCC p Chromophobe RCC p Papillary RCC p All RCC p Female sex 5.2 (2 18) 0.01 Angular interface 21.8 (2 237) 0.011 12.0 (1 102) 0.023 Hypodense rim 14.4 (1 196) 0.045 Coexisting lipid-rich 23.3 (2 244) 0.009 8.7 (1 70) 0.042 angiomyolipoma Homogeneity 146.2 (13 1715) < 0.001 38.9 (3 501) 0.005 High attenuation a 73.4 (6 919) 0.001 22.3 (5 105) < 0.001 41.9 (6 291) < 0.001 39.4 (8 205) < 0.001 Note Data are odds ratio with 95% CI in parentheses. Dash indicates not significant. a Unenhanced CT with threshold attenuation: clear cell type = 37.5 HU, chromophobe type = 39.5 HU, papillary type = 39.5 HU, all RCCs = 38.5 HU. corticomedullary phase) was 13.7 HU in lipid-poor angiomyolipoma and 2.6 HU in clear-cell RCC (p = 0.170). Compared with lipid-poor angiomyolipoma, chromophobe RCC showed significantly lower density in the corticomedullary phase (p = 0.001), whereas papillary RCC showed the statistically significantly lowest density in both the corticomedullary phase and the nephrographic phase (both p < 0.001). The amount of tumor enhancement in the corticomedullary phase of lipid-poor angiomyolipoma was significantly higher than that in both the chromophobe RCC (p < 0.02) and papillary RCC (p < 0.001) groups. Unlike the lipid-poor angiomyolipoma and clear-cell RCC patterns, the average tumor enhancement over time (nephrographic phase-corticomedullary phase) increased 11 HU in chromophobe RCC and 6 HU in papillary RCC (both p = 0.006 compared with lipid-poor angiomyolipoma). When comparing lipid-poor angiomyolipoma and all RCC, the only significant enhancing parameter was the tumor enhancement over time (p = 0.034). Multivariate Analysis The tumor attenuation evident in unenhanced scans was dichotomized into suggestive of lipid-poor angiomyolipoma or suggestive of RCC. The threshold was determined by the ROC curve, which was 37.5 HU for clear-cell RCC, 39.5 HU for chromophobe RCC, 39.5 HU for papillary RCC, and 38.5 HU for all RCC. We did not include parameters with a case number of zero in the analysis group in the multivariate analysis; this step excluded calcified components and perinephric collaterals for lipid-poor angiomyolipoma and hypodense rim in the papillary RCC subgroup. We excluded the contrast enhancement measurements from multivariate analysis because these results were usually influenced by subjective factors, including cardiac and renal function. Table 3 lists the significant predictors and their odds ratios (ORs). We found five significant predictors for clear-cell RCC, two for chromophobe RCC, one for papillary RCC, and four for all RCC. Unenhanced high attenuation was the only parameter that consistently and significantly differentiated between lipid-poor angiomyolipoma and all RCC groups. For significant variables, Table 1020 AJR:201, November 2013

Imaging Differentiation of Benign and Malignant Renal Tumors Table 4: CT Performance for Diagnosis of Lipid-Poor Angiomyolipoma Performance Value Clear-Cell RCC Chromophobe RCC Papillary RCC All RCC Female Sex Sensitivity 60.6 60.6 60.6 60.6 Specificity 74.1 58.1 70.6 68.6 PPV 58.8 60.6 80.0 38.5 NPV 75.5 58.1 48.0 84.3 Angular interface Sensitivity 39.4 39.4 39.4 39.4 Specificity 98.1 93.5 88.2 95.1 PPV 92.9 86.7 86.7 72.2 NPV 72.6 59.2 42.9 82.9 Hypodense rim Sensitivity 27.3 27.3 27.3 27.3 Specificity 98.1 96.8 100.0 98.0 PPV 90.0 90.0 100.0 81.8 NPV 68.8 55.6 41.5 80.6 Coexisting lipid-rich angiomyolipoma Sensitivity 27.3 27.3 27.3 27.3 Specificity 98.1 93.5 88.2 95.1 PPV 90.0 81.8 81.8 64.3 NPV 68.8 54.7 38.5 80.2 Homogeneity Sensitivity 96.8 96.8 96.8 96.8 Specificity 87.0 12.9 23.5 53.9 PPV 81.1 52.6 69.8 39.0 NPV 97.9 80.0 80.0 98.2 High attenuation a Sensitivity 83.9 83.9 83.9 83.9 Specificity 83.3 80.6 82.4 82.4 PPV 74.3 81.3 89.7 59.1 NPV 90.0 83.3 73.7 94.4 Note All values are percentages. RCC = renal cell carcinoma. a Unenhanced CT with threshold attenuation: clear cell type = 37.5 HU, chromophobe type = 39.5 HU, papillary type = 39.5 HU, all RCCs = 38.5 HU. 4 presents the sensitivity, specificity, PPV, NPV, and overall accuracy. We did not include the calcified component and perinephric collaterals in the multivariate analysis, but they had high NPV for excluding lipidpoor angiomyolipoma. Discussion The detection of renal carcinoma is increasing at a rate of 2 3% per year worldwide because of improved screening and imaging techniques. However, mortality from the disease has not increased proportionately [16]. An increasing number of cases have involved small renal masses that were identified during a screening imaging examination [17]. These lesions are usually presumed to be RCC before surgery or local ablation, and 10 17% of these lesions turn out to be benign after surgery [18 20]. A relatively recent study [21] reported that 22% (81 of 376) of small renal masses that had been presumed to be RCC were benign after partial nephrectomy, and angiomyolipoma accounted for nearly half of this group of benign lesions (35 of 81 patients). The traditional diagnosis of angiomyolipoma depended on detecting gross intratumoral fat on CT. However, the amount of fat in an angiomyolipoma is variable. Bosniak et al. [22] described the importance of small amounts of fat in diagnosing angiomyolipoma. Some angiomyolipomas the so-called lipid-poor angiomyolipomas consist of predominantly muscular and vascular components and have no gross fat. This makes it difficult to differentiate them from RCC on preoperative imaging, even with newer-generation CT scanners [2, 3, 5]. On pathologic examination, lipid-poor angiomyolipoma may display a nonhomogeneous distribution of fat, generally defined as less than 25% of fat per high-power field [4, 23]. MRI is sensitive to detecting minimal fat. Kim et al. [9] showed that double-echo chemicalshift MRI could show the minimal intratumoral fat, achieving high accuracy in differentiating angiomyolipoma from other renal neoplasms. However, Hindman et al. [11] showed that clear-cell RCC and lipid-poor angiomyolipoma could not be distinguished on the basis of signal intensity or tumor-tospleen ratio for standard MR images. This may be because of the stricter patient inclusion (small tumor size and pathologically proven fat content of 25% or less) and ROI selection of this study. It is not clear if additional imaging signs can reliably be used to differentiate benign from malignant small renal masses. Several researchers have attempted to detect microscopic fat content within renal masses using CT histogram analysis for pixel measurement. This method yields variable and occasionally conflicting results [4, 6 8, 24]. The use of differing parameters or techniques also generates inconsistent and nonreproducible results. These techniques include different thresholds, ROI sizes, methods of ROI selection, slice thickness, and types of CT scanners. In addition, histogram analysis [6] is not widely available and is not necessarily applicable to general radiology practice. Our study looked for specific morphologic parameters that are easy to identify but may be potentially overlooked in diagnosing lipid-poor angiomyolipoma. Verma et al. [14] first described the angular interface as a feature of benign renal lesions, which could be used to differentiate them prospectively from RCC. They described the appearance of an exophytic renal AJR:201, November 2013 1021

Yang et al. mass that has a tapering almost pyramidal interface with a definable apex within the parenchyma. This feature appeared in 79% of benign complex exophytic masses and 76% of angiomyolipomas [14]. However, this has not been specifically described in lipid-poor angiomyolipoma, to our knowledge. Our results show that 39% of the lipid-poor angiomyolipomas displayed an angular interface, a proportion that is significantly higher than that found in any of the three RCC subgroups and all RCCs. Multivariate analysis showed that the angular interface consistently and significantly differed between clear-cell RCC (p = 0.011) and all RCCs (p = 0.023). These findings likely reflect the soft and flexible consistency of lipid-poor angiomyolipomas. Although the sensitivity of this feature was not high (0.39), the specificity was high (0.98). A hypodense rim is another morphologic feature in lipid-poor angiomyolipoma that has not been previously reported in the literature, to our knowledge. This term describes a low-density rim at the peripheral area of the tumor on unenhanced CT, which contrasts with the adjacent normal kidney parenchyma. A low-density area at the periphery of a tumor is not a typical location for tumor necrosis or cystic change. In one patient who underwent both CT and MRI examinations, CT showed a hypodense rim at the interface of lipid-poor angiomyolipoma and kidney, and the opposed-phase MRI showed a sharp dark line at the corresponding location, confirming its nature as fat. Our results reveal that 27.3% of the lipid-poor angiomyolipoma show this feature, which may suggest that fat at the periphery of a tumor is usually subtle and therefore requires special attention for identification on a preoperative CT. An argument may arise concerning the usefulness of this feature in differentiating benign and malignant renal tumors because intratumoral fat may be detected in a minority of clear-cell RCC [25 27]. However, our results showed that hypodense rim sign rarely appeared in the RCC groups and had high specificity and PPV for the diagnosis of lipid-poor angiomyolipoma. The multiplicity of angiomyolipoma provides another pointer for diagnosis. The multiplicity of angiomyolipoma is usually emphasized when it occurs in patients with tuberous sclerosis [28]. We found that in approximately one third of lipid-poor angiomyolipoma, coexisting lipid-rich angiomyolipoma could be identified in the ipsilateral or contralateral kidney, representing a significant predictor of lipid-poor angiomyolipoma. However, caution should be taken when evaluating these circumstances. There is also an increased incidence of RCC in tuberous sclerosis patients [29]. There are also syndromes in which RCC can be multifocal (e.g., Birt-Hogg-Dubé and Von Hippel- Lindau syndrome [30]). Perinephric engorged vessels are a wellknown sign of malignant lesions; this condition indicates neovascularity and is commonly found in large tumors [13]. Our study showed that no lipid-poor angiomyolipoma displayed peripheral collaterals, although abundant vascular tissue was present within the tumor. In contrast, perinephric collateral vessels appeared relatively frequently in size-comparable RCC subgroups. It yielded high NPV (100%) for the diagnosis of lipid-poor angiomyolipoma. High tumor attenuation in the unenhanced phase has been reported as a finding specific to lipid-poor angiomyolipoma [3, 4, 31]. Previous research has shown that papillary RCC has a higher attenuation in the unenhanced phase, whereas clear-cell RCC and chromophobe RCC have lower unenhanced phase attenuation [13]. Because both lipid-poor angiomyolipoma and papillary RCC have high attenuation in unenhanced CT, differentiating between these two conditions may be challenging [31]. However, this parameter still showed a strong significance after the stepwise multivariate analysis of all RCC groups. Using the cutoff values generated from the ROC curve, the results achieved at least 80% for both sensitivity and specificity for lipid-poor angiomyolipoma compared with the three RCC subtypes and all RCCs. Most lipid-poor angiomyolipomas showed a homogeneous enhancement pattern that is consistent with the findings of Kim et al. [5]. Our study also found that most of the clear-cell RCC showed a heterogeneous enhancement pattern. In contrast, the majority of chromophobe RCC and papillary RCC tumors revealed a homogeneous pattern, as shown in previous reports [32 34]. When comparing lipid-poor angiomyolipoma with all RCC, a homogeneous enhancement pattern still showed significant differences. This might be because of the relatively large case number of clear-cell RCC. Multiphasic contrast enhancement is another method that researchers have proposed using to differentiate lipid-poor angiomyolipoma and RCC. However, previous studies have produced variable results [5, 13], thus limiting the utility of this approach. These conflicting results might stem from a number of reasons: First, different subtypes of RCC showed different enhancement patterns [13, 32, 33, 35], and previous research did not analyze RCC by different subtypes. Second, the tumor size, which might affect vascularity, was not comparable between the study groups. Other factors including renal function, contrast medium concentration, bolus injection rate, scanning protocol, and type of CT scanner, also may influence the enhancement results. Thus, the results of enhancement measurements were not easy to reproduce in clinical practice. We found that both lipid-poor angiomyolipoma and clear-cell RCC showed strong enhancement in the corticomedullary phase and a slight washout pattern in the delayed phase, whereas chromophobe RCC and papillary RCC tumors showed completely different enhancement patterns that gradually increased enhancement over time. Although comparing lipid-poor angiomyolipoma with individual types of RCC revealed some significant enhancing parameters, most became insignificant when comparing all RCCs as a group. There are limitations of our study. First, other benign renal tumors (including oncocytoma or metanephric adenoma) may yield similar imaging findings to those of lipidpoor angiomyolipoma [34] but were not included in our study. Second, the slice thickness for the CT examinations included in our study was 5 mm. Thinner sections, such as 3 mm, may show the true nature of a renal mass more reliably, particularly for small renal masses. Third, we did not attempt to correlate the grading of renal tumor with imaging findings. Finally, this is a retrospective study and unintended selection bias inevitably exists. Further prospective study is necessary to support our results. In conclusion, an angular interface, hypodense rim, homogeneous enhancement pattern, and high unenhanced attenuation are specific CT features that may suggest lipid-poor renal angiomyolipoma and have high negative predictive value of renal cell carcinoma. By careful assessment of these features, unnecessary surgery may be avoided. Acknowledgments We thank Hui-Chen Lee, Division of Experimental Surgery and Biostatistics Task Force, for statistical analysis and consultation on the study. 1022 AJR:201, November 2013

Imaging Differentiation of Benign and Malignant Renal Tumors References 1. Sherman JL, Hartman DS, Friedman AC, nature of solid renal masses in adults. J Urol 2003; 169:71 74 176:905 909 24. Millet I, Doyon FC, Hoa D, et al. Characterization Madewell JE, Davis CJ, Goldman SM. Angiomyo- 13. Zhang J, Lefkowitz RA, Ishill NM, et al. Solid re- of small solid renal lesions: can benign and malig- lipoma: computed tomographic-pathologic corre- nal cortical tumors: differentiation with CT. Radi- nant tumors be differentiated with CT? AJR 2011; lation of 17 cases. AJR 1981; 137:1221 1226 ology 2007; 244:494 504 197:887 896 2. Sant GR, Heaney JA, Ucci AA Jr, Sarno RC, 14. Verma SK, Mitchell DG, Yang R, et al. Exophytic 25. D Angelo PC, Gash JR, Horn AW, Klein FA. Fat Meares EM Jr. Computed tomographic findings in renal angiomyolipoma: an histologic correlation. Urology 1984; 24:293 296 3. Jinzaki M, Tanimoto A, Narimatsu Y, et al. Angiomyolipoma: imaging findings in lesions with minimal fat. Radiology 1997; 205:497 502 4. Simpfendorfer C, Herts BR, Motta-Ramirez GA, et al. Angiomyolipoma with minimal fat on MDCT: can counts of negative-attenuation pixels aid diagnosis? AJR 2009; 192:438 443 5. Kim JK, Park SY, Shon JH, Cho KS. Angiomyolipoma with minimal fat: differentiation from renal cell carcinoma at biphasic helical CT. Radiology 2004; 230:677 684 6. Kim JY, Kim JK, Kim N, Cho KS. CT histogram analysis: differentiation of angiomyolipoma without visible fat from renal cell carcinoma at CT imaging. Radiology 2008; 246:472 479 7. Catalano OA, Samir AE, Sahani DV, Hahn PF. Pixel distribution analysis: can it be used to distinguish clear cell carcinomas from angiomyolipomas with minimal fat? Radiology 2008; 247:738 746 8. Chaudhry HS, Davenport MS, Nieman CM, Ho LM, Neville AM. Histogram analysis of small solid renal masses: differentiating minimal fat angiomyolipoma from renal cell carcinoma. AJR 2012; 198:377 383 9. Kim JK, Kim SH, Jang YJ, et al. Renal angiomyolipoma with minimal fat: differentiation from other neoplasms at double-echo chemical shift FLASH MR imaging. Radiology 2006; 239:174 180 10. Tanaka H, Yoshida S, Fujii Y, et al. Diffusionweighted magnetic resonance imaging in the differentiation of angiomyolipoma with minimal fat from clear cell renal cell carcinoma. Int J Urol 2011; 18:727 730 11. Hindman N, Ngo L, Genega EM, et al. Angiomyolipoma with minimal fat: can it be differentiated from clear cell renal cell carcinoma by using standard MR techniques? Radiology 2012; 265:468 477 12. Dechet CB, Zincke H, Sebo TJ, et al. Prospective analysis of computerized tomography and needle biopsy with permanent sectioning to determine the renal masses: angular interface with renal parenchyma for distinguishing benign from malignant lesions at MR imaging. Radiology 2010; 255:501 507 15. Davenport MS, Neville AM, Ellis JH, Cohan RH, Chaudhry HS, Leder RA. Diagnosis of renal angiomyolipoma with Hounsfield unit thresholds: effect of size of region of interest and nephrographic phase imaging. Radiology 2011; 260:158 165 16. Jewett MA, Mattar K, Basiuk J, et al. Active surveillance of small renal masses: progression patterns of early stage kidney cancer. Eur Urol 2011; 60:39 44 17. Campbell SC, Novick AC, Belldegrun A, et al. Guideline for management of the clinical T1 renal mass. J Urol 2009; 182:1271 1279 18. Fujii Y, Komai Y, Saito K, et al. Incidence of benign pathologic lesions at partial nephrectomy for presumed RCC renal masses: Japanese dual-center experience with 176 consecutive patients. Urology 2008; 72:598 602 19. Schachter LR, Cookson MS, Chang SS, et al. Second prize: frequency of benign renal cortical tumors and histologic subtypes based on size in a contemporary series what to tell our patients. J Endourol 2007; 21:819 823 20. Snyder ME, Bach A, Kattan MW, Raj GV, Reuter VE, Russo P. Incidence of benign lesions for clinically localized renal masses smaller than 7 cm in radiological diameter: influence of sex. J Urol 2006; 176:2391 2395; discussion, 2395 2396 21. Jeon HG, Lee SR, Kim KH, et al. Benign lesions after partial nephrectomy for presumed renal cell carcinoma in masses 4 cm or less: prevalence and predictors in Korean patients. Urology 2010; 76:574 579 22. Bosniak MA, Megibow AJ, Hulnick DH, Horii S, Raghavendra BN. CT diagnosis of renal angiomyolipoma: the importance of detecting small amounts of fat. AJR 1988; 151:497 501 23. Milner J, McNeil B, Alioto J, et al. Fat poor renal angiomyolipoma: patient, computerized tomography and histological findings. J Urol 2006; in renal cell carcinoma that lacks associated calcifications. AJR 2002; 178:931 932 26. Hammadeh MY, Thomas K, Philp T, Singh M. Renal cell carcinoma containing fat mimicking angiomyolipoma: demonstration with CT scan and histopathology. Eur Radiol 1998; 8:228 229 27. Hélénon O, Merran S, Paraf F, et al. Unusual fatcontaining tumors of the kidney: a diagnostic dilemma. RadioGraphics 1997; 17:129 144 28. Lane BR, Aydin H, Danforth TL, et al. Clinical correlates of renal angiomyolipoma subtypes in 209 patients: classic, fat poor, tuberous sclerosis associated and epithelioid. J Urol 2008; 180:836 843 29. Rakowski SK, Winterkorn EB, Paul E, Steele DJ, Halpern EF, Thiele EA. Renal manifestations of tuberous sclerosis complex: incidence, prognosis, and predictive factors. Kidney Int 2006; 70:1777 1782 30. Verine J, Pluvinage A, Bousquet G, et al. Hereditary renal cancer syndromes: an update of a systematic review. Eur Urol 2010; 58:701 710 31. Silverman SG, Mortele KJ, Tuncali K, Jinzaki M, Cibas ES. Hyperattenuating renal masses: etiologies, pathogenesis, and imaging evaluation. Radio- Graphics 2007; 27:1131 1143 32. Kim JK, Kim TK, Ahn HJ, Kim CS, Kim KR, Cho KS. Differentiation of subtypes of renal cell carcinoma on helical CT scans. AJR 2002; 178:1499 1506 33. Sheir KZ, El-Azab M, Mosbah A, El-Baz M, Shaaban AA. Differentiation of renal cell carcinoma subtypes by multislice computerized tomography. J Urol 2005; 174:451 455; discussion, 455 34. Jinzaki M, Tanimoto A, Mukai M, et al. Doublephase helical CT of small renal parenchymal neoplasms: correlation with pathologic findings and tumor angiogenesis. J Comput Assist Tomogr 2000; 24:835 842 35. Herts BR, Coll DM, Novick AC, et al. Enhancement characteristics of papillary renal neoplasms revealed on triphasic helical CT of the kidneys. AJR 2002; 178:367 372 (Figures start on next page) AJR:201, November 2013 1023

Yang et al. A D Fig. 1 Various patients with renal tumors. A F, CT images show specific findings including intratumoral calcified component (arrow, A), cystic component (asterisk, B), peritumoral engorged vessels (arrow, C), angular interface (dotted lines, D), coexisting lipid-rich renal angiomyolipoma (lra) with lipid-poor renal angiomyolipoma (lpa) (E), and fine hypodense rim between tumor and normal renal parenchyma on unenhanced scan (arrow, F). A Fig. 2 44-year-old woman with lipid-poor renal angiomyolipoma. A C, Multiphasic CT images show tumor with homogeneous high density of 44 HU (asterisk, A) on unenhanced phase image (A). Note thin hypodense rim at interface with normal renal parenchyma (arrow, A). Tumor shows substantial enhancement on corticomedullary phase image (B) and slight washout on nephrographic phase image (C). (Fig. 2 continues on next page) B E B C F C 1024 AJR:201, November 2013

Imaging Differentiation of Benign and Malignant Renal Tumors A D Fig. 2 (continued) 44-year-old woman with lipidpoor renal angiomyolipoma. D, Coronal reconstruction image clearly shows angular interface (dotted lines). D B E Fig. 3 CT and MR images of 45-year-old woman with lipid-poor renal angiomyolipoma. A C, Multiphasic CT images show 47-HU highdensity tumor (asterisk, A) with thin hypodense rim at interface with normal renal parenchyma (arrows, A) on unenhanced phase image (A). Tumor shows strong enhancement in corticomedullary phase (B) and washout in nephrographic phase (C). Note density of tumor is homogeneous in all three phases. D and E, In-phase (D) and out-of-phase (E) chemicalshift MR images. Hypodense rim shows high signal intensity on in-phase image (arrows, D) and signal drop in out-of-phase image (arrows, E), confirming nature as fat. C AJR:201, November 2013 1025

Yang et al. A C Fig. 4 69-year-old man with clear-cell renal cell carcinoma. A C, Multiphasic CT image on unenhanced phase (A) shows tumor (T) with similar density to normal renal parenchyma (34 HU). Tumor shows strong and heterogeneous enhancement on corticomedullary phase image (B) and washout on nephrographic phase image (C). D, Coronal reformation image of nephrographic phase reveals round shape and heterogeneous density. B D 1026 AJR:201, November 2013

Imaging Differentiation of Benign and Malignant Renal Tumors A C B D Fig. 5 34-year-old man with chromophobe renal cell carcinoma. A C, Multiphasic CT image in unenhanced phase (A) shows higher density of tumor (T) to normal renal parenchyma (34 HU). Tumor shows gradually increased enhancement on corticomedullary phase image (B) and subsequent nephrographic phase image (C). D, Coronal MIP reformation of nephrographic phase shows engorged collateral vessel (arrow). A B Fig. 6 33-year-old man with papillary renal cell carcinoma. A C, Multiphasic CT image in unenhanced phase (A) shows tumor (T) with slightly higher density (38 HU) to normal renal parenchyma as well as faint calcified component (arrow). Tumor enhancement is homogeneously modest on both corticomedullary phase image (B) and nephrographic phase image (C). C AJR:201, November 2013 1027

Yang et al. 140 120 Angiomyolipoma 100 80 60 40 20 0 Unenhanced Corticomedullary Phase Nephrographic Clear-cell RCC Chromophobe RCC Papillary RCC All RCC Fig. 7 Graph shows contrast enhancement over time of renal tumors. Lipid-poor renal angiomyolipoma and clear-cell renal cell carcinoma both show early strong enhancement and washout pattern, whereas chromophobe and papillary renal cell carcinoma show gradual enhancement over time. Papillary renal cell carcinoma shows lowest attenuation among all renal tumors in both corticomedullary and nephrographic phases. 1028 AJR:201, November 2013