Liver angiomyolipomas: A clinical, radiologic, and pathologic analysis of 22 patients from a single center

Jean M. Butte, MD,a Richard K. Do, MD, PhD,b Jinru Shia, MD,c Mithat Gonen,€ PhD,d Michael I. D’Angelica, MD,a George I. Getrajdman, MD,b Peter J. Allen, MD,a Yuman Fong, MD,a Ronald P. DeMatteo, MD,a David S. Klimstra, MD,c and William R. Jarnagin, MD,a New York, NY

Background. Liver angiomyolipomas (AML) are mesenchymal with an uncertain clinical behavior. The spectrum of presentation, imaging and histologic features, and outcomes were analyzed in all patients treated at Memorial Sloan-Kettering Cancer Center. Methods. Demographics, , pathologic, treatment, and outcome-related variables for consecutive patients were reviewed retrospectively. All imaging studies obtained at presentation were reexamined, categorized, and compared using Fisher and Wilcoxon tests. Results. Between 1989 and 2010, 238 patients with AML were seen and 22 (9.3%) had liver involvement (exclusive = 17; combined with = 5). The median age was 53 years; 18 were females, and 15 had symptoms. AML was not suspected at initial presentation in any patient. Fat-containing neoplasms on imaging were larger (P = .03), with more heterogeneous enhancement compared with fat-poor neoplasms (P = .001), but none of the imaging/histologic features correlated with outcome. Thirteen patients underwent resection (R0 = 9), 4 were observed, 2 received chemotherapy, 2 emboli- zation, and 1 plus intra-arterial chemotherapy. Two patients treated with R0-resection recurred and 2 treated with chemotherapy progressed, but no patient died of AML-related causes. At a median follow-up of 36 months, 7 patients were free of disease, 13 were alive with disease, 1 died of an unrelated cause, and another was lost to follow-up. Conclusion. AMLs are rare neoplasms with an indolent course in most patients. Subtypes based on fat content are recognized, but clinical behavior does not seem to be different. Recurrence after resection is not associated with disease-related mortality. Resection may be unnecessary in selected asymptomatic patients if the diagnosis can be established definitively. (Surgery 2011;150:557-67.)

From the Department of Surgery,a the Department of Radiology,b the Department of Pathology,c and the Department of Epidemiology and Biostatistics,d Memorial Sloan-Kettering Cancer Center, New York, NY

4 ANGIOMYOLIPOMAS (AML) are unusual mesenchymal misdiagnosed as adenoma or hepatocellular carci- neoplasms composed of blood vessels, smooth noma (HCC), which are much more common.3,5,6 muscle, and adipose cells.1 These neoplasms often When these neoplasms arise from the kidney, arise in the kidneys and may be associated with tu- and particularly when large, treatment is often berous sclerosis.2 The liver is a less common site of recommended to relieve pain or to treat or prevent origin, and liver involvement is often an incidental hemorrhage; however, there is no consensus re- finding on imaging studies or identified on evalu- garding the best treatment of hepatic AML, and in ation of nonspecific symptoms.3 Because hepatic recent years, this topic has become more contro- AML are rare and have imaging features that over- versial.7 Classically, this lesion has been defined as lap with other tumors, definitive preoperative diag- a benign , although it has been suggested nosis is difficult to achieve, with >50% of patients that rare cases may be associated with metastatic disease or recurrence after resection.7 Observation has been recommended as the best option for Accepted for publication March 22, 2011. asymptomatic liver lesions, provided the diagnosis 8,9 Reprint requests: William R. Jarnagin, MD, Department of Sur- can be made reliably. Because of the difficulty gery, Memorial Sloan-Kettering Cancer Center, 1275 York Ave- that can be encountered in establishing a defini- nue, New York, NY 10065. E-mail: [email protected]. tive diagnosis and the small but real potential 0039-6060/$ - see front matter risk of bleeding10 and/or malignant transforma- Ó 2011 Mosby, Inc. All rights reserved. tion in recent reports,2,11-13 some authors have doi:10.1016/j.surg.2011.03.006 advocated resection. Unfortunately, published

SURGERY 557 558 Butte et al Surgery September 2011 clinical series are small, do not identify preopera- Radiologic study. A retrospective review was tive markers that might predict an aggressive performed for all patients, complemented by review clinical behavior, and do not define whether the of available pretreatment imaging studies, which association between kidney and liver AML corre- included both abdominal computed tomography sponds to multicentric or metastatic disease.11,14-18 (CT) and magnetic resonance imaging (MRI). Hepatic AML can range in size from 0.1 cm up Re-analysis of available studies was performed by to nearly 40 cm in some patients and can present the study radiologist (R.K.D.), who was unaware of with a wide array of morphologic and histologic the clinical information other than the diagnosis. appearances. Contributing further to the problems Specific radiologic characteristics evaluated were related to diagnosis, the 3 primary histologic com- number and size of lesions, fat content, presence of ponents may be present in varying proportions. intratumoral vessels, and homogeneity of enhance- Definitive diagnosis is based on anatomic patho- ment. Patients with incidental AML not identified logic analysis supported by immunohistochemical on preoperative imaging with noncontrast imaging staining. The cell component is the studies, such as 18F-fluorodeoxyglucose positron most specific to the diagnosis, and characteristi- emission tomography or unenhanced abdominal cally, these lesions stain positive for homatropine CT study (18FDG PET-CT), were not included in methylbromide-45 and Melan-A.1,19 Hepatic AML the radiologic re-analysis, but they were included are classified based on the line of differentiation in the outcome assessment. Imaging studies from and the predominant tissue component. The the remaining patients were re-analyzed (R.K.D.); most common type is the mixed lipomatous these studies consisted of abdominal CT and/or neoplasm, which contains $70% adipose tissue, MRI with intravenous contrast, which included whereas myomatous variants contain a much lesser dynamic contrast enhanced images with arterial proportion (10%); neoplasms composed mostly of phase imaging. smooth muscle elements (myomatous subtypes) Pathologic examination. The diagnosis of AML frequently cause substantial diagnostic confusion. was confirmed histologically in all patients. The The relationship among clinical behavior and pathologic material was re-analyzed at the time of radiologic and pathologic features are not well this review by 1 pathologist (J.S.). Features ana- characterized in these neoplasms. The objective of lyzed included tumor size, cellular composition this study was to characterize all patients with liver (epithelioid cell, spindle cell, and adipocyte com- AML evaluated at Memorial Sloan-Kettering Can- ponents), nuclear pleomorphism, mitotic activity, cer Center to determine whether clinical behavior and immunohistochemical staining pattern.1 correlates with radiologic and/or pathologic char- Treatment. Treatment varied according to the acteristics. Such a link, if it exists, would be helpful clinical and radiologic features. Specifically, for making rational treatment recommendations treatment recommendations took into account and predicting outcomes. tumor-related symptoms, whether the diagnosis was incidental, tumor size, extent of resection, and PATIENTS AND METHODS suspicion of malignancy. Resection, used in some Subjects and data collection. After approval by patients, was performed according to the authors’ the Institutional Review Board, a prospectively previously reported operative approach to partial maintained hospital database was queried for all hepatectomy.20 Hepatic artery embolization using patients with a diagnosis of AML between 1989 bland particles was used selectively, according to a and 2010. Patients with the diagnosis of liver AML previously published approach.21,22 Morbidity and were selected and analyzed. Recorded data in- mortality were defined as any complication or death cluded patient demographics, laboratory values, resulting from any procedure (operation or emboli- imaging studies, treatment administered, operative zation) at any time. procedures, perioperative outcomes, histopathol- Follow-up. Patient’s status was categorized at the ogy, follow-up, and long-term survival. This infor- time of last follow-up as follows: no evidence of dis- mation was supplemented by review of individual ease, alive with disease, dead of disease, operative medical records. Patient evaluation included mortality, or dead of other causes. The follow-up physical examination and imaging studies; for was the interval between the date of first consulta- some patients, serum tumor markers and/or tion and the date of last follow-up or death. Resec- percutaneous biopsy were obtained. Patients were tion with curative intent (R0 resection) was treated with different therapies, and final treat- defined as a complete resection without compro- ment recommendations were based on the results mise of any microscopic margins, whereas an R1 of these studies. resection represented the presence of microscopic Surgery Butte et al 559 Volume 150, Number 3 disease involvement in 1 resection margin and R2 resection as macroscopic compromise of any margin after final pathology report, or macro- scopic residual disease. Statistical analyses. Patient data were summa- rized using median values and ranges. Radiologic characteristics of patients with fat component versus without fat component were compared using the Fisher exact test for categorical variables and the Wilcoxon test for continuous variables. All tests were 2 sided. Statistical analysis was performed with S.A.S version 9.2 (SAS Institute, Inc., Cary, NC).

RESULTS Fig 1. Localization of 238 patients with AML. Clinical presentation. From 1989 to 2010, 238 patients with AML were evaluated at Memorial Imaging studies. All patients had $1 imaging Sloan-Kettering Cancer Center. The vast majority study of the liver. Seven patients had 1 study, 9 had were renal in origin, with only 22 (9.3%) present- 2, and 6 had 3 studies. Specifically, abdominal CT ing with liver-predominant lesions (Fig 1). Within was used in 22 patients, abdominal ultrasonogra- this group, there were 18 females and 4 males phy in 8, MRI in 8, and 18FDG PET-CT in 5 with a median age of 53 years (range, 35–75). patients. AML was not suspected in any of these pa- The disease was localized exclusively in the liver tients as a first diagnosis. The lesion was localized in 17 patients; 5 others had history of both kidney in the left lobe in 9 patients, in the right lobe in and liver AML. AML was the primary diagnosis in 6, and in both lobes in 7. The median size of the 15 patients and represented a secondary, inciden- largest liver lesion was 7 cm (range, 1–17), and tal diagnosis in 7. The most common presenting in all patients it was solid. Suspicious lymphade- symptom was abdominal pain (8 patients). There nopathy was not identified on imaging in any were no patients with a history of tuberous sclero- patient. Three patients had a synchronous kidney sis. Nine patients had $1 evaluated AML at the time of imaging evaluation. (alpha-fetoprotein in 7, carcinoembryonic antigen Sixteen of 22 patients had imaging studies avail- in 7, and carbohydrate antigenic 19.9 in 4), which able for review. Two patients underwent liver resec- were normal in all cases. Hepatitis serology was tion for unrelated reasons and had incidental AMLs tested in 6 patients, with only 1 positive for hepati- <1 cm that were not visible on preoperative imag- tis B virus exposure. ing; 2 other patients only had noncontrast imaging Pretreatment clinical diagnosis depended on available. Thus, reanalysis of imaging studies was presentation. In patients with incidentally discov- possible in 12 patients. Based on the available data, 2 ered neoplasms (n = 7), the initial diagnosis was pre- imaging patterns of hepatic AMLs were discernable sumed to represent metastasis from a preexisting and depended on the presence or absence of fat neoplasm, as listed in Table I. The other 15 patients (Table II). Hepatic AML with visible macroscopic fat had liver only (n = 10) or liver and kidney tumors component (n = 7) were larger (median, 12.5 cm; (n = 5). In patients with only liver disease, HCC range, 1.7–16.1) than those tumors without visible was suspected in 1, and primary liver neoplasm of fat component (ie, fat-poor AMLs; n = 5; median, uncertain histology in the other 9. In patients with 4.3 cm; range, 1.4–6.1; P = .03). Also, macroscopic kidney and liver tumors, the pretreatment diagnosis fat-containing AMLs demonstrated more heteroge- was with liver metastases in 2 and renal cell neous enhancement when they were compared with carcinoma with liver metastases in 3. Fourteen fat-poor AMLs, which were homogeneously enhanc- patients underwent a percutaneous liver biopsy, ing in 5 of 5 patients (P = .001). Macroscopic fat- which confirmed AML in 9 (64%), showed a non- containing AMLs were often solitary (5 of 7 patients specific spindle cell neoplasm in 2, suggested had a single lesion). Also, the presence of intratu- adrenocortical carcinoma or HCC (1 each) and moral vessels was not a distinguishing factor was nondiagnostic in 1. Three patients had diagno- between fat-containing and fat-poor AMLs (Fig 2). ses established with operative biopsies, one of which Hepatic fat-poor AML typically demonstrated ho- was obtained during laparoscopic staging. Table I mogeneous arterial enhancement with washout to summarizes clinical characteristics. isointensity on portal venous phase (Fig 3). 560 Butte et al Surgery September 2011

Table I. Clinical characteristics of 22 patients with ‘‘metastatic’’ AML (Fig 4), unlike all other AMLs, liver angiomyolipoma (AML) which were hypervascular on arterial phase. Liver AML n =22 Pathology. All neoplasms showed histologic and immunohistochemical evidence of AML. Median Gender number of tumors in patients treated with resec- Female 18 (81.8%) tion was 1, with 10 patients having a solitary lesion; Male 4 (18.2%) median size was 3 cm (range, 0.4–15), and 9 of 13 Age, y Median 53 patients who underwent operation had an R0 Range 35–75 resection (see below). A gross photograph of a Main diagnosis 22 resected neoplasm is illustrated in Fig 5, A. Liver AML 15 A re-review of the histopathologic material was Primary 10 carried out in 12 patients. Despite variable cellular Associated with kidney AML 5 composition, 11 of 12 patients had an epithelioid Other diagnosis + incidental primary 7 component (median, 70%; range, 30–100%), 5 of liver AML 12 had a fat component (median, 40%; range, Retroperitoneal sarcoma 1 20–50%), and 3 of 12 a spindle cell component Adrenal cancer with liver 1 (median, 70%; range, 40–100%). More than 1 com- metastases ponent was seen in 7 of 12 patients: Epithelioid fat Desmoid tumor with liver 1 metastases in 5 and epithelioid-spindle cells in 2 patients. Colon cancer (stage III) 1 There were no neoplasms with all 3 components. Colon cancer with liver metastases 1 Ten neoplasms had cellular and nuclear pleomor- Lung cancer with metastases 1 phism, which was mild in 6 patients, mild to Gastric gastrointestinal stromal 1 moderate in 1, moderate in 2, and severe in 1. tumor Two patients had $2 mitoses per 10 high power Symptoms* field (10 HPF); 1 patient with localized disease had Abdominal pain 10 2 mitoses per 10 HPF, and 1 patient thought to Incidental 5 have metastatic disease had >10 mitoses per 10 Back pain 1 HPF (Fig 5, B and C). Of note, the patient with >10 Cushing’s syndrome 1 mitoses per 10 HPF had multifocal tumors with se- Cough 1 Nausea 1 vere pleomorphism and absence of an obvious fat Rectal bleeding 1 component on imaging (ie, fat-poor tumors). Weight loss 3 Immunohistochemical staining pattern was evaluated in all patients. Seventeen of 17 cases *One patient had >1 symptom. demonstrated positive staining for homatropine methylbromide-45, 11 of 11 cases for actin, 2 of 2 For patients with an available abdominal MRI for desmin, and 2 of 2 for vimentin. In contrast, (n = 7), the non-fat component of the neoplasms S-100 protein was positive in 2 of 4 patients, were T1 hypointense and T2 hyperintense com- hepatocyte antigen was negative in 8 of 8 patients, pared with the liver, with increased heterogeneity and keratin was negative in 5 of 5 patients. in larger lesions. Bulk fat was demonstrated by Treatment. Different treatments were adminis- chemical shift artifact on in phase and opposed tered depending on the clinical and radiographic phase T1-weighted gradient echo imaging, unlike characteristics, the presence or absence of symp- intravoxel fat seen commonly with hepatic ade- toms, and the presumptive preoperative diagnosis. noma. Most lesions lacked a visible capsule on Some neoplasms were resected, particularly those imaging, and no capsule was found in the resected tumors felt to represent malignancy, whereas specimen. others with metastatic disease were treated with Two hepatic AMLs classified as ‘‘metastatic’’ had systemic chemotherapy, hepatic artery emboliza- prior imaging studies. One patient with an 18FDG tion, or embolization plus intra-arterial chemo- PET-CT study only demonstrated 2 large 18FDG- therapy. Four patients were observed. It should be avid hepatic lesions; unfortunately, this patient noted that the diagnosis of AML was not suspected did not have a contrast-enhanced CT or MRI. in any patient at the time of initial presentation The other patient had a liver MRI that demon- and was known in only 3 patients before operation; strated a single and homogeneous hypovascular however, the diagnosis was confirmed in all lesion on arterial and portal venous phase of patients submitted to nonoperative treatment or contrast enhancement and was later classified as observation (Table III). Surgery Butte et al 561 Volume 150, Number 3

Table II. Imaging characteristics of hepatic angiomyolipomas in 12 patients who had contrast-enhanced CT and/or MRI Patient Fat component Number of lesions Size of largest lesion (cm) Intratumoral vessels Homogeneous enhancement 1 Yes 1 12.5 Yes No 2 Yes 1 9.2 Yes No 3 Yes 2 16.1 Yes No 4 Yes 1 15.0 No No 5 Yes 1 1.7 No No 6 Yes 1 13.4 Yes No 7 Yes 2 11.1 Yes No 8 No 4 3.1 Yes Yes 9 No 2 1.4 No Yes 10 No 1 4.7 No Yes 11 No 6 6.1 No Yes 12 No 2 4.3 No Yes

Fig 2. CT triple phase liver study with axial noncontrast (A), postcontrast arterial phase (B), and portal venous phase (C) images. A circumscribed mass containing fat (arrow) with heterogeneous arterial enhancement, typical of fat containing hepatic AML.

Thirteen patients underwent resection; 6 of 2–13). One patient experienced substantial intra- these procedures were performed as part of treat- operative bleeding from the neoplasms and re- ment for an unrelated diagnosis, with the AML quired packing; however, there was no mortality being an incidental finding. Nine patients had an related to operation or to other treatments R0 resection. Four patients were observed, 3 of (embolization and/or chemotherapy). whom had no changes on follow-up imaging, and Follow-up. After a median follow-up of 36 the other patient had an advanced lung cancer months (range, 1–192), patient status was as fol- with an incidental liver AML diagnosed during lows: Alive with disease (n = 13), no evidence of dis- staging. Two patients thought to have metastatic ease (n = 7), dead of other causes (n = 1), and lost AML were treated with chemotherapy (interferon to follow-up (n = 1). Two of 9 patients treated with in 1 and temsirolimus in the other), 2 were treated R0 resection developed recurrent disease at a me- with embolization (Fig 6), and a third under initial dian of 47 months after resection, 1 within the liver suspicion of unresectable HCC was treated with and the other at extrahepatic sites (lung and retro- intra-arterial chemotherapy. The therapy was dis- peritoneum). Of note, the second patient had 2 continued in favor of bland embolization when FDG-avid liver tumors documented on PET before the diagnosis of AML was made after a new review the hepatic resection, and predominance of spin- of the pathology report. Table III summarizes the dle cell component (70%), high pleomorphism, treatment, follow-up, and current status of all and >10 mitoses per 10 HPF was present on pathol- patients. ogy review. Both patients were alive with disease The median duration of hospital stay in those after a median follow-up of 131 months. Another patients treated with resection was 8.5 days (range, patient died as a consequence of metastatic 562 Butte et al Surgery September 2011

Fig 3. Liver MRI with axial images from noncontrast T1-weighted gradient echo opposed-phase (A), in-phase (B), and T2-weighted fat suppressed fast spin echo (C) sequences. Axial images from 3-dimensional, fat-suppressed, T1-weighted gradient echo sequence before (D) and after contrast administration during the arterial (E) and portal venous (F) phases of enhancement. Three liver lesions are depicted, including a T1 and T2 isointense FNH (arrowhead), a mildly T1 hypo- intense and T2 hyperintense AML (arrow), and a markedly T1 hypointense and T2 hyperintense hemangioma (block arrow). Both AML and FNH demonstrate nearly uniform avid arterial enhancement (E) with washout to near iso intensity on portal venous phase (F). A second FNH lateral to the hemangioma is best seen on arterial phase imaging (E). adrenal cancer after a follow-up of 135 months. Of neoplasm of the falciform ligament/ligamentum the 4 patients submitted to an incomplete resec- teres, and unusual clear cell neoplasms of the pan- tion, none had evidence of disease progression creas, rectum, abdominal serosa, uterus, vulva, after a median follow-up of 19 months. thigh, and heart.24 Three of the 4 patients submitted to observation The present study confirms previous reports had stable disease after a median follow-up of 80 showing the kidney to be the most common months, and the 4th was lost to follow-up. One disease site, with the liver involved in <10% of patient treated with chemotherapy progressed af- patients.25 In addition, the large majority of pa- ter 13 months of treatment, whereas another had tients were symptomatic women, an observation stable disease. Those patients treated with emboli- that has been noted and may be related to hormo- zation (n = 3) had no progression of disease after nal stimulus.26 Specifically, studies have shown that an average follow-up of 12.7 months (range, 1–36). estrogen receptor subtype beta and androgen re- None of the patients in this cohort died of ceptor have been determined in a high proportion disease. The median overall survival from the time of AML localized in the kidney.27 Liver AML asso- of diagnosis was 36 months (range, 1–192). ciated with has been described, Disease-free survival in patients treated with an but this association was not observed in any of R0 or R1 resection (n = 10) was 47 months (range, the patients in this series and it has been suggested 3–135), whereas progression-free survival in pa- that <10% of patients with liver AML have this tients treated with an R2 resection or no resection association.28,29 This study confirmed the results (n = 12) was 13 months (range, 1–109). of previous reports showing that tumor markers are not specific or helpful to support the diagnosis DISCUSSION and may be useful only to help exclude other AML is the most common subtype of ‘‘PEComas,’’ diagnoses28; any association with hepatitis virus or neoplasms with perivascular epithelioid cell infection would seem to be a coincidence.28,30,31 differentiation.23,24 Other, less common subtypes AML has been described as a benign neoplasm are clear cell neoplasms of the lung, associated with tuberous sclerosis and usually aris- lymphangioleiomyoma, clear cell myomelanocytic ing in the kidney.19 The liver is an uncommon site Surgery Butte et al 563 Volume 150, Number 3

Fig 4. Liver MRI with axial images from noncontrast, T1-weighted, gradient echo opposed-phase (A), in-phase (B), and T2-weighted fat-suppressed fast spin echo (C) sequences. Axial images from 3-dimensinoal, fat-suppressed, T1-weighted gradient echo sequence before (D) and after contrast administration during the arterial (E) and portal venous (F) phases of enhancement. A single, T1 hypointense and T2 hyperintense liver lesion in segment 4a remains hypovascular with rim enhancement after contrast administration.

Fig 5. An AML showing a well-delineated border and slightly variegated and lobulated cut surface on gross examination (A). Microscopically, AML of the liver often contain abundant epithelioid tumor cells with mild degree of nuclear pleomorphism and no mitosis (B), or sometimes apparent nuclear pleomorphism and increased mitotic activity (C). of origin, and most cases are misdiagnosed as in different organs (ie, kidney and liver) repre- HCC, adenoma, , or sarcoma.32 Despite sents an expression of multicentric involvement the benign features, some cases suggesting malig- or metastatic disease.14 Certain histopathologic nant behavior have been described15-17,33-35; features, such as cytologic atypia of epithelioid however, given the extreme rarity of AML, it is dif- cells,36 extensive tumor necrosis,16 sarcomatous ficult to determine whether simultaneous disease transformation,33 and loss of expression of CD Table III. Description of treatment and follow-up of 22 patients with liver angiomyolipoma 564

Diagnosis Diagnosis al et Butte suspected confirmed Place of Time of recurrence Size at initial before recurrence or or progression Current Gender Age (cm) presentation treatment Treatment Additional resection R0/R1/R2 Tumor status progression (months) status Female 58 NI No No Resection No R2 Stable disease No No AWD Female 54 NI No No Resection No R0 Recurrence Liver 53 AWD Female 53 3 No No Resection Adrenalectomy R0 NR No No DOC Female 58 9 No Yes Resection Total hysterectomy R0 NR No No NED + bilateral salpingo- oophorectomy Male 38 1.2 No Yes Resection Partial colectomy R0 NR No No NED Female 58 10 No No Resection Partial colectomy R0 NR No No NED Male 35 16 No No Resection Cholecystectomy R0 NR No No NED Female 49 2.5 No No Resection Partial R2 Stable No No AWD Female 66 4.5 No No Resection No R2 Stable No No AWD Female 61 8 No No Resection Partial gastrectomy R0 NR No No NED and distal pancreatectomy Male 41 9 No No Resection No R0 Recurrence Lungs and 41 AWD retroperitoneum Female 51 1.7 No No Resection* No R0 NR No No NED Female 57 13.7 No Yes Resection No R1 NR No No NED Female 54 12 No Yes Embolization — — Stable No No AWD plus HAICh Female 48 17 No Yes Embolization — — Stable No No AWD Female 53 11 No Yes Embolization — — Stable No No AWD Female 69 8 No Yes Systemic — — Progression Liver 13 AWD chemotherapy Male 75 6 No Yes Systemic — — Stable No No AWD chemotherapy Female 43 3 No Yes Observation — — Lost to follow-up — — — Female 43 9 No Yes Observation — — Stable No No AWD Female 50 5 No Yes Observation — — Stable No No AWD Female 50 2.7 No Yes Observation — — Stable No No AWD

*Laparoscopic resection. 2011 September AWD, Alive with disease; DOC, dead for other causes; HAICh, hepatic arterial infusion chemotherapy; NED, no evidence of disease; NI, no information; NR, no recurrence. Surgery Surgery Butte et al 565 Volume 150, Number 3

would use to describe the behavior of a malig- nancy, or a manifestation of multifocal involve- ment. Such a finding would be considered highly unusual for a purely benign hepatic neoplasm. On the other hand, neither patient has died of AML, despite the appearance of new disease after resection. The results of the current study support the conclusion that AML follow a generally indolent course. More than 85% of patients did not recur or progress, regardless of treatment, and none died of disease.28,37 Furthermore, none of the patients submitted to observation developed tumor- related complications, such as pain or hemor- rhage. Resection was the most commonly used treatment in this study; this approach was generally undertaken for resectable tumors when the diag- nosis was uncertain. In many of these cases, there was a suspicion of malignancy and/or the presence of symptoms, both of which are indications for resection. In patients with renal AML, arterial embolization is used commonly to mitigate the risk of hemorrhage, and this approach has been expanded to patients with liver involvement.38-41 In the present study, hepatic artery embolization with bland particles was used successfully in 3 pa- tients (Fig 6), all of whom had large neoplasms with biopsy confirmation of the diagnosis before treatment. The risk of spontaneous hemorrhage of hepatic AML seems to be less than those arising in the kidney, which are usually supplied by a sin- gle vessel and associated with . Large AML are soft neoplasms, with a consistency similar to that of hepatic adenomas or HCC, both of which carry some risk of bleeding. Given the low risk of embolization and its efficacy for preventing a potentially catastrophic complication, such an approach is not unreasonable, particularly for large, peripherally placed neoplasms. Emboliza- tion also may have some benefit for palliating tumor-related pain, as has been shown for other Fig 6. Pre-embolization images from (A) axial T2-weighted, fat-suppressed and (B) axial T1-weighted, postcontrast, arte- neoplasms. In contrast, there is a very limited rial phase sequences show a large heterogeneously hypervas- role for chemotherapy for treating AML. Greater cular mass in the right hepatic lobe. Postembolization, on a duration of follow-up in a larger cohort of patients postcontrast axial CT image (C), the mass now appears non- will be necessary, however, to confirm the findings enhancing. of the present study regarding therapeutic inter- ventions, which should be individualized based 11716 have been associated with more aggressive on clinical presentation, the presence or absence behavior and may identify patients with more ag- of symptoms, and the extent of disease. gressive disease. Two patients in the present study Establishing a diagnosis of AML remains chal- developed recurrent disease after an R0 resection, lenging. Indeed, given its very low incidence, AML one in the liver and one at multiple extrahepatic is often not considered in the differential diagnosis sites. In both patients, however, it is unclear of liver lesions, and its variable appearance on whether the emergence of new neoplasms repre- imaging adds further to the confusion. The radio- sented recurrence in the traditional sense, as one graphic characteristics depend on the proportion 566 Butte et al Surgery September 2011 of blood vessels, smooth muscle, and adipose tissue noted to have high mitotic indices. None of these in the tumor.1 The fat component varies from features, however, seemed to correlate with clinical <5% to >90% of tumor volume42; most AMLs behavior in this small cohort. have some amount of fat detectable on abdominal In conclusion, AMLs are rare neoplasms that CT and/or MRI.43 AMLs with abundant fat are involve the liver in <10% of patients with this hypoattenuating on noncontrast CT with absent diagnosis. Imaging studies may reveal certain char- or minimal enhancement after contrast injection, acteristic features, but given the variability of whereas they are hyperintense on T1- and appearance, definitive diagnosis requires histo- T2-weighted MRI sequences.42 These lesions, logic evaluation. Most patients had an indolent which are hypervascular and with a large amount course, few recurred or progressed, and none died of fat, have a characteristic appearance and may of disease, regardless of treatment. Our results be somewhat easier to distinguish from other tu- suggest that an aggressive resectional approach mors (Fig 2). In contrast, fat-poor lesions are may be avoidable in selected asymptomatic pa- more difficult to differentiate from HCC. They tients with a proven diagnosis. Indeed, observation are hypointense on T1-weighted, precontrast MRI might be an appropriate course in such patients. sequences and tend to enhance after contrast Hepatic artery embolization seems to have efficacy injection.42 Their hypervascularity may be another and may be a reasonable alternative, if any means of discriminating these neoplasms, espe- treatment is indicated. cially from , myolipoma, metastatic ter- atoma, and focal fatty infiltration.3 Distinguishing AML from HCC, adenoma, and focal nodular hy- REFERENCES perplasia remains the greatest challenge, especially 1. Tsui WM, Colombari R, Portmann BC, Bonetti F, Thung SN, Ferrell LD, et al. Hepatic angiomyolipoma: a clinicopatho- in small tumors, because they tend to have early logic study of 30 cases and delineation of unusual morpho- vascular enhancement on contrast-enhanced, non- logic variants. Am J Surg Pathol 1999;23:34-48. specific, abdominal CT and MRI.43 The lack of a 2. Zeng JP, Dong JH, Zhang WZ, Wang J, Pang XP. Hepatic capsule, bulky fat, and prominent intratumoral angiomyolipoma: a clinical experience in diagnosis and vessels seem to be the most helpful findings for treatment. Dig Dis Sci 2010;55:3235-40. 3.PrasadSR,WangH,RosasH,MeniasCO,NarraVR,Middleton distinguishing AML from HCC, especially in the WD, et al. Fat-containing lesions of the liver: radiologic- 44 noncirrhotic liver. pathologic correlation. Radiographics 2005;25:321-31. Therefore, making the diagnosis of AML is 4. Yang CY, Ho MC, Jeng YM, Hu RH, Wu YM, Lee PH. Man- difficult for many reasons, and misdiagnosis is agement of hepatic angiomyolipoma. J Gastrointest Surg not uncommon, most often in patients with small 2007;11:452-7. 5. Nakamura N, Kudo A, Ito K, Tanaka S, Arii S. A hepatic li- neoplasms with low fat content. In contrast, larger poma mimicking angiomyolipoma of the liver: report of a tumors with high fat content have a more consis- case. Surg Today 2009;39:825-8. tent and characteristic appearance. Another situa- 6. Jiang TA, Zhao QY, Chen MY, Wang LJ, Ao JY. Diagnostic tion where diagnosis can be problematic is the analysis of hepatic angiomyolipoma. Hepatobiliary Pancreat incidental finding in patients undergoing evalua- Dis Int 2005;4:152-5. 7. Petrolla AA, Xin W. Hepatic angiomyolipoma. Arch Pathol tion for other neoplasms, in whom metastatic liver Lab Med 2008;132:1679-82. disease is a possibility. Laparoscopic exploration 8. Schwartz P, Travers P, Hunt DR. Angiomyolipoma of the and biopsy is a reasonable approach in patients liver. Aust N Z J Surg 1989;59:969-71. with diagnostic uncertainty. Alternatively, an ade- 9. Yeh CN, Chen MF, Hung CF, Chen TC, Chao TC. Angiomy- quate core biopsy, not a fine needle aspiration, olipoma of the liver. J Surg Oncol 2001;77:195-200. 10. Takanami K, Kaneta T, Hitachi S, Yamada T, Ishida K, Ri- may also be diagnostic. In addition to adequate kiyama T, et al. F-18 FDG PET/CT findings in two patients tissue for analysis, familiarity with this entity and an with hepatic angiomyolipoma with and without intratu- index of suspicion as a diagnostic possibility, as well moral hemorrhage. Clin Nucl Med 2010;35:18-21. as an experienced pathologist, are indispensible. 11. Takahara M, Miyake Y, Matsumoto K, Kawai D, Kaji E, Toyo- The results of the present study showed that, kawa T, et al. A case of hepatic angiomyolipoma difficult to distinguish from hepatocellular carcinoma. World J Gastro- based on the imaging appearance, AML can be enterol 2009;15:2930-2. divided into 2 general categories: Fat containing 12. Chen P, Yuan T, Liu H. Hepatic angiomyolipoma mimick- and fat poor. These 2 types were different in ing hepatic clear cell carcinoma. J Int Med Res 2009;37: several respects. Specifically, compared with fat- 257-63. poor neoplasms, high fat content was associated 13. Yip SK, Sim CS, Tan BS. Liver metastasis and local recur- rence after radical nephrectomy for an atypical angiomyoli- with larger, solitary lesions, the presence of intra- poma. J Urol 2001;165:898-9. tumoral vessels, and the absence of homogeneous 14. Saito M, Tsukamoto T, Takahashi T, Sai K, Fujii H, enhancement. Additionally, 2 fat-poor tumors were Nagashima K. Multifocal angiomyolipoma affecting Surgery Butte et al 567 Volume 150, Number 3

the liver and lung without tuberous sclerosis. J Clin 29. Short MP, Richardson EP Jr, Haines JL, Kwiatkowski DJ. Pathol 2004;57:221-4. Clinical, neuropathological and genetic aspects of the tu- 15. Sato K, Ueda Y, Tachibana H, Miyazawa K, Chikazawa I, Kaji berous sclerosis complex. Brain Pathol 1995;5:173-9. S, et al. Malignant epithelioid angiomyolipoma of the kid- 30. Sajima S, Kinoshita H, Okuda K, Saito N, Hashino K, Sugi- ney in a patient with tuberous sclerosis: an autopsy case re- moto R, et al. Angiomyolipoma of the liver---a case report port with p53 gene mutation analysis. Pathol Res Pract and review of 48 cases reported in Japan. Kurume Med J 2008;204:771-7. 1999;46:127-31. 16. Nguyen TT, Gorman B, Shields D, Goodman Z. Malignant 31. Bilbao I, Castro E, Dopazo C, Castro S, Allende E, Genesca J, hepatic angiomyolipoma: report of a case and review of lit- et al. [Hepatic angiomyolipoma in two patients infected with erature. Am J Surg Pathol 2008;32:793-8. hepatitis C virus]. Gastroenterol Hepatol 2007;30:222-8. 17. Deng YF, Lin Q, Zhang SH, Ling YM, He JK, Chen XF. Malig- 32. Basaran C, Karcaaltincaba M, Akata D, Karabulut N, Akinci nant angiomyolipoma in the liver: a case report with patho- D, Ozmen M, et al. Fat-containing lesions of the liver: cross- logical and molecular analysis. Pathol Res Pract 2008;204: sectional imaging findings with emphasis on MRI. AJR Am J 911-8. Roentgenol 2005;184:1103-10. 18. Takamura K, Miyake H, Fujii M, Nishi M, Tashiro S, Shi- 33. Ferry JA, Malt RA, Young RH. Renal angiomyolipoma with mada M. Multiple hepatic angiomyolipomas with a soli- sarcomatous transformation and pulmonary metastases. tary omental angiomyolipoma. J Med Invest 2005;52: Am J Surg Pathol 1991;15:1083-8. 218-22. 34. Lin WC, Wang JH, Wei CJ, Pan CC, Chang CY. Malignant re- 19. Huang PC, Chen JT, Ho WL. Clinicopathologic analysis of nal epithelioid angiomyolipoma with aggressive behavior renal and extrarenal angiomyolipomas: report of 44 cases. and distant metastasis. J Chin Med Assoc 2003;66:303-6. Zhonghua Yi Xue Za Zhi (Taipei) 2000;63:37-44. 35. Mizuguchi T, Katsuramaki T, Nobuoka T, Nishikage A, 20. Jarnagin WR, Gonen M, Fong Y, DeMatteo RP, Ben-Porat L, Oshima H, Kawasaki H, et al. Growth of hepatic angiomy- Little S, et al. Improvement in perioperative outcome after olipoma indicating malignant potential. J Gastroenterol hepatic resection: analysis of 1,803 consecutive cases over Hepatol 2004;19:1328-30. the past decade. Ann Surg 2002;236:397-406. 36. Nonomura A, Mizukami Y, Shimizu K, Kadoya M, Matsui O. 21. Brown KT, Nevins AB, Getrajdman GI, Brody LA, Kurtz RC, Angiomyolipoma mimicking true lipoma of the liver: report Fong Y, et al. Particle embolization for hepatocellular carci- of two cases. Pathol Int 1996;46:221-7. noma. J Vasc Interv Radiol 1998;9:822-8. 37. Ren N, Qin LX, Tang ZY, Wu ZQ, Fan J. Diagnosis and treat- 22. Covey AM, Maluccio MA, Schubert J, BenPorat L, Brody LA, ment of hepatic angiomyolipoma in 26 cases. World J Gas- Sofocleous CT, et al. Particle embolization of recurrent he- troenterol 2003;9:1856-8. patocellular carcinoma after hepatectomy. Cancer 2006; 38. Wolff N, Kabbani W, Bradley T, Raj G, Watumull L, Brugar- 106:2181-9. olas J. and temsirolimus for epithelioid angiomy- 23. Bonetti F, Pea M, Martignoni G, Zamboni G. PEC and sugar. olipoma. J Clin Oncol 2010;28:e65-8. Am J Surg Pathol 1992;16:307-8. 39. Ramon J, Rimon U, Garniek A, Golan G, Bensaid P, Kitrey 24. Strzelczyk JM, Durczynski A, Szymanski D, Jablkowski M, ND, et al. Renal angiomyolipoma: long-term results following Dworniak D, Sporny S. Primary perivascular epithelioid selective arterial embolization. Eur Urol 2009;55:1155-61. cell tumor (PEComa) of the liver: report of a case. Surg To- 40. Bissler JJ, McCormack FX, Young LR, Elwing JM, Chuck G, day 2009;39:916-21. Leonard JM, et al. Sirolimus for angiomyolipoma in tuber- 25. Barnard M, Lajoie G. Angiomyolipoma: immunohistochem- ous sclerosis complex or . ical and ultrastructural study of 14 cases. Ultrastruct Pathol N Engl J Med 2008;358:140-51. 2001;25:21-9. 41. Lenton J, Kessel D, Watkinson AF. Embolization of renal an- 26. Cho NH, Shim HS, Choi YD, Kim DS. Estrogen receptor giomyolipoma: immediate complications and long-term is significantly associated with the epithelioid variants of outcomes. Clin Radiol 2008;63:864-70. renal angiomyolipoma: a clinicopathological and immu- 42. Valls C, Iannacconne R, Alba E, Murakami T, Hori M, nohistochemical study of 67 cases. Pathol Int 2004;54: Passariello R, et al. Fat in the liver: diagnosis and charac- 510-5. terization. Eur Radiol 2006;16:2292-308. 27. Boorjian SA, Sheinin Y, Crispen PL, Lohse CM, Kwon ED, 43. Yen YH, Wang JH, Lu SN, Changchien CS. Contrast- Leibovich BC. Hormone receptor expression in renal angi- enhanced ultrasonography in hepatic angiomyolipoma. omyolipoma: clinicopathologic correlation. Urology 2008; J Ultrasound Med 2005;24:855-9. 72:927-32. 44. Jeon TY, Kim SH, Lim HK, Lee WJ. Assessment of triple- 28. Zhou YM, Li B, Xu F, Wang B, Li DQ, Zhang XF, et al. Clin- phase CT findings for the differentiation of fat-deficient he- ical features of hepatic angiomyolipoma. Hepatobiliary Pan- patic angiomyolipoma from hepatocellular carcinoma in creat Dis Int 2008;7:284-7. non-cirrhotic liver. Eur J Radiol 2010;73:601-6.