Prediction of Microvascular Invasion of Hepatocellular Carcinoma: Preoperative CT and Histopathologic Correlation

Transcription

1 Gastrointestinal Imaging Original Research Gastrointestinal Imaging Original Research Chen-Te Chou 1,2 Ran-Chou Chen 2,3 Wei-Chan Lin 2,4 Chih-Jan Ko 5 Chia-Bang Chen 1 Yao-Li Chen 5,6 Chou CT, Chen RC, Lin WC, Ko CJ, Chen CB, Chen YL Keywords: CT, hepatocellular carcinoma, microvascular invasion, tumor capsule, tumor margin DOI: /AJR Received January 17, 2013; accepted after revision October 30, R. C. Chen and Y. L. Chen contributed equally to this study. The work was supported by a research grant from the National Science Council (grant B MY2). 1 Department of Radiology, Chang-Hua Christian Hospital, Changhua City, Taiwan. 2 Department of Biomedical Imaging and Radiological Science, National Yang-Ming Medical University, Taipei, Taiwan. 3 Department of Radiology, Taipei City Hospital, Taipei City, Taiwan. 4 Department of Radiology, Cathay General Hospital, Taipei, Taiwan. 5 Transplant Medicine & Surgery Research Centre, Chang-Hua Christian Hospital, No. 135 Nanhsiao St, 500 Chang-Hua, Taiwan. Address correspondence to Y. L. Chen (chouchente@yahoo.com.tw). 6 School of Medicine, Chung Shan Medical University, Taichung City, Taiwan. WEB This is a web exclusive article. AJR 2014; 203:W253 W X/14/2033 W253 American Roentgen Ray Society Prediction of Microvascular Invasion of Hepatocellular Carcinoma: Preoperative CT and Histopathologic Correlation OBJECTIVE. The objective of our study was to prospectively investigate whether nonsmooth margins detected on multiphasic CT images correlate with the presence and location of microvascular invasion (MVI) in hepatocellular carcinoma (HCC). SUBJECTS AND METHODS. A total of 102 patients with preoperative CT findings of solitary HCC were prospectively enrolled. size, tumor capsule, tumor margins, and peritumoral enhancement on preoperative CT images were assessed. Histopathologic results including the following were also recorded: tumor differentiation; liver fibrosis score; presence or absence of MVI; and, if present, the location of MVI. Correlation between tumor margin on preoperative CT images and histopathologic location of MVI was determined. RESULTS. Pathologic examination revealed MVI in 60 of the 102 HCC specimens. Although the results of the univariate analysis showed that tumor size, higher Edmondson- Steiner grade, and nonsmooth tumor margins were associated with MVI, multivariate analysis revealed that only nonsmooth margins correlated with the presence of MVI in HCC (p < 0.001). Of the 60 HCC specimens with histopathologic evidence of MVI, 40 exhibited focal nonsmooth margins. In addition, the locations of the nonsmooth margins and MVI were similar in 36 of the 40 specimens. CONCLUSION. Nonsmooth tumor margins correlated with the histopathologic presence and location of MVI. Therefore, nonsmooth margins detected on multiphasic CT may be predictive of MVI in HCC. H istopathologic evidence of vascular tumor invasion is a wellknown prognostic factor for patients who have undergone hepatic resection or liver transplant for the treatment of hepatocellular carcinoma (HCC) [1 4]. Macrovascular invasion, which is easily diagnosed using routine cross-sectional imaging, is a relative contraindication for surgery and transplant. Pathologic markers suggestive of an increased risk of microvascular invasion (MVI) include large tumor size, multiple tumors, intrahepatic micrometastasis, poor histologic grade, and gross anatomic subtype as defined by the Liver Cancer Study Group of Japan [5]. The three subtypes are defined as follows: a simple nodular type (a distinctly nodular tumor that frequently has a capsule); a simple nodular type with extranodular growth (a single nodular tumor with varying degrees of tumor growth beyond the tumor capsule); and a confluent multinodular type (a confluence of several minute to small nodules) [6 10]. Iwatsuki et al. [11] reported that the risk of recurrence of HCC after liver transplant was 4.4-fold higher in patients with MVI and 15- fold higher in patients with macrovascular invasion. Therefore, preoperative diagnosis of HCC with MVI is essential to help predict the outcome of patients who undergo liver transplant or curative liver resection. Recent studies have shown that PET [12] and diffusion-weighted MRI [13] can be used to predict the presence of MVI. Several studies [8, 14 16], including a retrospective study by our group [14], have shown that there is a certain degree of correlation between the contours of the tumors seen on CT or MRI and the Japanese Gross Classification System [5]. However, none of those studies was able to show a direct correlation between nonsmooth tumor margins and histopathologic evidence of MVI. In this prospective study, we investigated whether nonsmooth margins detected on multiphasic CT correlate with the presence and location of MVI in HCC. W253

2 Subjects and Methods Patients This prospective study was approved by the institutional review board of our institute, and informed consent was obtained from all of the study participants. The inclusion criteria for patients were as follows: chronic liver disease with dynamic CT evidence of a solitary hepatic tumor showing hypervascular enhancement in the arterial phase and contrast washout in the equilibrium phase, no history of malignancy, and agreement to undergo surgical resection or liver transplant for histopathologic diagnosis. During the period from January 2011 through March 2012, a total of 105 patients with radiographic evidence of HCC underwent elective curative hepatectomy or liver transplant in our institution. Of these 105 patients, partial hepatectomy was performed in 71 patients, 11 of whom underwent hepatectomy without resection of the adjacent major vessels because of impaired liver function; anatomic hemihepatectomy was performed in 27 patients; and liver transplant was performed in seven patients. Three patients were excluded from the study because of a diagnosis of hepatocholangiocarcinoma (two patients) or macrovascular portal venous thrombus (one patient). Therefore, the final study population comprised 102 patients (73 men and 29 women) with a mean age of 62.6 years (range, years). Underlying hepatic disease included hepatitis B in 42 patients, hepatitis C in 43 patients, combined hepatitis B and C in three patients, alcoholic liver disease in five patients, and cryptogenic liver disease in nine patients. According to the Child-Pugh liver disease classification, 91 patients had Child-Pugh class A disease, eight had Child-Pugh class B disease, and three had Child- Pugh class C disease. CT Image Acquisition CT images of the liver were obtained with a 16- MDCT scanner (LightSpeed Ultra 16, GE Healthcare) using the following parameters: z-axis modulation; a gantry rotation time of 0.6 second for the unenhanced studies and for the hepatic arterial and portovenous phases (beam pitch, 0.825:1) and a gantry rotation time of 0.8 second for the equilibrium phase (beam pitch, 1.1:1); 5-mm section thickness reconstructions; 120 kvp; ma; and a standard reconstruction algorithm. Patients were scanned in a craniocaudal direction. Nonionic contrast medium (iohexol [Omnipaque 350, GE Healthcare]) was administered at a total dose of ml with an injection rate of 3 ml/s through a 20-gauge venous cannula placed in the antecubital vein. For triphasic acquisitions, scanning was started with a 10-second scanning delay ( seconds after injection of the contrast Fig. 1 Illustration shows three patterns of tumor margins. (Drawing by Chou CT) Smooth margin agent) for the hepatic arterial phase after the attenuation value of the aorta reached 120 HU. Fifteen seconds after the endpoint of the hepatic arterial phase ( seconds after injection of the contrast agent), scans for the portovenous phase were acquired. Equilibrium phase images were acquired 120 seconds ( seconds after injection of the contrast agent) after the end of the acquisition of the portovenous phase. Whole-liver scanning was completed in 4 8 seconds while the patients held their breath. Coronary and sagittal images were reconstructed with a 5-mm slice thickness using raw imaging data for equilibrium phase CT and a standard reconstruction algorithm. Imaging Analysis The imaging analysis was performed on a dual-screen diagnostic workstation. Before starting the study, two experienced radiologists who were blinded to clinical information independently reviewed the preoperative CT images for each patient. Coronary and sagittal reformatted equilibrium phase images of the whole liver, which supplemented transverse plane images, were provided to the two reviewers. margins on equilibrium phase images were grouped into one of the three following categories depending on margin characteristics (Fig. 1): smooth margin, presenting as a smooth tumor normal liver interface seen on all axial, coronary, and sagittal images; nonsmooth margin, presenting as focal outgrowth of nodules protruding into the nontumor parenchyma (Fig. 2); or nonsmooth margin with multinodular type (Fig. 3), presenting as multifocal outgrowths protruding into the nontumoral parenchyma [14, 17]. For each tumor with nonsmooth margins, the reviewers recorded the location of nonsmooth margins on the liver map (location on transverse, coronary, and sagittal planes) for correlation with histopathologic findings. The presence or absence of tumor capsules was assessed in the equilibrium phase by identifying a thin, linear, enhanced structure encasing the tumor. s were then categorized into one of the Nonsmooth margins Single nodular type Focal extranodular type Multinodular type three following groups according to the presence or absence of tumor capsules: tumors presenting with capsules that completely surrounded the tumor circumference, tumors presenting with capsules that did not completely surround the tumor circumference, or tumors without radiologic evidence of a tumor capsule. The presence or absence of peritumoral enhancement was assessed in the dynamic CT study. Peritumoral enhancement was defined as the existence of a detectable arterial-enhancing area adjacent to the tumor border on arterial phase images that became isodense compared with the liver parenchyma on equilibrium phase images. Differences between the two reviewers were resolved by a third radiologist who joined a consensus conference and who was also blind to clinical information. Lesion Tracking In patients who underwent hepatectomy or liver transplant, consistency between CT and pathologic findings was reached by means of a lesion-tracking procedure. For each assessment, an investigator who did not participate in the initial review of the images selected the largest tumor diameter in the axial, coronal, and sagittal planes to prepare the liver maps (one map for each plane). Each liver map was completed by drawing each individual liver lesion and adjacent major vascular and anatomic landmarks around the tumor on a respective map according to the preoperative CT images and the Couinaud system of liver anatomy. The maps were drawn as accurately as possible for CT and pathologic findings. In each assessment, a continuous number was allocated to each lesion. In patients who underwent partial hepatectomy, the surgeon matched the location of the margin on the axial plane of a specimen with that of the markers so that a radiologic-pathologic correlation could be determined (Fig. 1C). The two satellite nodules in two patients with multinodular type HCC were excluded. W254

3 C Fig year-old man with solitary hepatocellular carcinoma (HCC) at segment VI of liver who underwent curative hepatectomy. A, Arterial phase CT image shows nodular tumor (arrow) with heterogeneous enhancement at segment VI of liver. B, shows extranodular extension (arrowhead) on late phase CT image. C, Surgeon indicated axial plane with sutures (three or four sutures) at border of specimen (arrow) to enable radiologic-pathologic correlation. D, Photomicrograph (H and E stain, 40) of histopathologic specimen reveals microvascular invasion of tumor (arrow). Histopathologic Examination All hepatic specimens were sliced in 5-mm increments by the pathologist. The plane of the section was chosen to be as close to the axial plane as possible on the basis of the markers provided by the surgeon and the major vascular and anatomic landmarks. The entire tumor margin in each specimen slice was divided into multiple sequential tissue blocks, coded with slice and sequential tissue block numbers, and submitted for microscopic examination. The pathologist recorded the slice and sequential tissue block numbers on the corresponding spatial positions of the tumor. All hepatic specimens were reviewed by a hepatic pathologist with more than 20 years of experience in hepatic pathology who was blinded to A the preoperative CT findings and clinical information. The histopathologic features of the tumors, fibrosis score, and location of MVI in tumor specimens were recorded. The degree of tumor differentiation was categorized according to the Edmondson-Steiner classification [18]. The degree of liver fibrosis was categorized according to the Metavir fibrosis scoring system [19]. In the HCC specimens with MVI, the pathologist marked the MVI site on the liver maps for each tumor. Correlation of Pathologic Specimens With Imaging Findings Radiologic-pathologic correlation was determined to evaluate the relationship between nonsmooth tumor margins and histopathologic evidence of MVI (Fig. 4). During the histopathologic examination, the octants of the tumors presenting with MVI were identified using the liver maps completed by the pathologist. During examination of the CT images, the octants of the tumors showing nonsmooth margins were identified using the liver maps completed by the radiologists. A radiologic-pathologic correlation was defined as specimens that showed the presence of MVI and a nonsmooth margin in the same octant. Statistical Analysis The interobserver difference between the initial two observers was evaluated with the kappa test. An independent Student t test was used to compare tumor size between the MVI-positive and D B W255

4 MVI-negative groups. Categoric variables, such as fibrosing score, Edmondson-Steiner grade of tumor, peritumoral enhancement, tumor margin, and radiologic evidence of tumor capsule, were analyzed by the chi-square test or Fisher exact test. The parameters found to have statistical significance by univariate analyses were entered into a stepwise, multiple logistic regression model to elucidate the independent predictors of MVI. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for the parameters that showed statistical significance by multivariate analysis. A p value < 0.05 was considered to indicate a statistically significant difference. TABLE 1: Clinical Characteristics of 102 Patients With Solitary Hepatocellular Carcinoma (HCC) Clinical Risk Factors All Patients (n = 102) Histopathologic MVI Negative (n = 42) Positive (n = 60) Age (y) Mean ± SD 62.6 ± ± ± 11.5 Sex (no. of patients) Male Female size (cm) < Mean ± SD 4.4 ± ± ± 4.3 Underlying liver disease (no. of patients) HBV HCV HBV and HCV Alcoholic liver disease Other Child-Pugh class (no. of patients) A B C AFP Mean ± SD (ng/ml) 1837 ± ± ± 11, < 20 ng/ml (no. of patients) ng/ml (no. of patients) Histologic grade a (no. of patients) I II III Liver fibrosis score b (no. of patients) Note MVI = microvascular invasion, HBV = hepatitis B virus, HCV = hepatitis C virus, AFP = α-fetoprotein. a According to Edmondson-Steiner classification [18]. b Metavir score [19]. Results Of the 102 specimens, 45 were determined to be focal extranodular type, nine were multinodular type, and the other 48 were single nodular type. Interobserver agreement was good between the two reviewers of the radiologic images. The kappa values were for tumor margin, for peritumoral enhancement, and for radiologic evidence of tumor capsule. Histopathologic analysis of the specimens revealed that 60 of the HCC lesions were positive for MVI. The results of the univariate analysis for the clinical characteristics and histopathologic findings in patients with and in those without MVI are shown in Table 1. size (MVI-positive group vs MVI-negative group: mean ± SD, 5.6 ± 4.3 cm vs 2.7 ± 1.7 cm, respectively; p < 0.001) and higher Edmondson-Steiner grade (grade III vs grades I and II; p = 0.020) were associated with the presence of MVI. There were no significant differences between the two patient groups in age, sex, liver fibrosis score, underlying liver disease, Child-Pugh class, or a-fetoprotein level. Univariate analyses of the radiologic findings of patients with and those without MVI are shown in Table 2. Histopathologic examination of the 45 focal extranodular HCC specimens revealed that 40 showed evidence of MVI. In addition, all nine of the multinodular HCC specimens showed MVI. Histopathologic examination revealed that five specimens with focal extranodular HCC showed no evidence of MVI and that 11 of 48 HCC specimens with smooth margins showed evidence of MVI. Of the 60 HCC specimens with histopathologic evidence of MVI, 40 showed evidence of focal extranodular extension, nine showed evidence of multinodular extension, and 11 had smooth margins. In 36 of 40 HCC specimens with focal extranodular extension, nonsmooth margins on CT images were located in the same octant as MVI in histopathologic specimens. The locations of MVI and nonsmooth margins were in different octants in the other four (4/40) HCC specimens. Of the nine HCC specimens showing multinodular extension, the location of MVI and that of nonsmooth margins was considered a mismatch. In the 49 HCC specimens with nonsmooth margins and MVI, the pathologic MVI octants were the same as the CT nonsmooth octants in 73% (36/49) of the cases. The presence of tumor capsules on CT images (p = 0.003) and the presence of nonsmooth tumor margins on preoperative CT images (p < 0.001) were associated with the presence of MVI. Of the 52 tumors with radiologic evidence of capsules, CT showed incomplete patterns for 30 tu- p W256

5 mors and complete patterns for 22 tumors. There was no significant difference in the presence of MVI between the tumors with CT findings showing an incomplete capsule and those with CT findings showing a complete capsule. Seven HCCs showed wedge peritumoral enhancement, one HCC showed circumferential enhancement, and the other HCCs showed no obvious peritumoral enhancement. Peritumoral enhancement revealed no statistically significant associations with MVI. The results of the multivariate logistic regression analysis of risk factors in patients with and those without MVI are shown in Table 3. Only nonsmooth tumor margin was associated with MVI (p < 0.001). The presence of nonsmooth tumor margins had a sensitivity of 81.7% (95% CI, %), a specificity of 88.1% (95% CI, %), an accuracy rate of 84.3% (95% CI, %), a PPV of 90.7% (95% CI, %), and an NPV of 77.1% (95% CI, %) in predicting MVI. TABLE 2: Univariate Analysis of Radiologic Findings for Patients With and Patients Without Microvascular Invasion (MVI) Radiologic Risk Factors Total No. of Patients Discussion In this study, we prospectively explored the correlation between nonsmooth tumor margins identified on dynamic four-phase CT with the presence of MVI in histopathologic specimens. Of the 54 HCCs with nonsmooth margins, 49 had MVI. The site of MVI in the pathologic specimens correlated with the site of focal extranodular extension on CT images in 73% (36/49) of the cases. In four specimens, MVI was seen along the smooth margin instead of the focal nonsmooth margin. Of the 48 HCC specimens with smooth margins, 11 showed evidence of MVI. Nonsmooth tumor margins as a predictor of MVI had a sensitivity of 81.7% and a specificity of 88.1%. Our results confirm that nonsmooth tumor margins identified on multiphasic CT can predict MVI and that MVI most frequently takes place at the site of extranodular extension. Kanai et al. [20] have proposed a gross classification system in which nodular HCC can be subclassified into a single nodular Histopathologic MVI Negative (n = 42 Patients) Positive (n = 60 Patients) Radiologic evidence of tumor capsule Negative Positive Incomplete capsule Complete capsule Peritumoral enhancement a Negative Positive margin < Smooth Nonsmooth Focal extranodular Multinodular a Fisher exact test. type, a single nodular type with extranodular growth, and a contiguous multinodular type. Studies have shown that the single nodular type with extranodular growth and the contiguous multinodular type are associated with increased risk of MVI, micrometastasis, and poorer outcome [7, 8, 18, 20, 21]. Preoperative prediction of growth subtype with imaging had been attempted in the past, but the imaging technique (10-mm slice thickness) was not sensitive enough [8]. In our study, we found that nonsmooth tumor p Fig year-old woman with solitary hepatocellular carcinoma (HCC) at segments V and VI of liver who underwent hepatectomy. A, Arterial phase CT image shows nodular tumor (arrows) with heterogeneous enhancement at segments V and VI of liver. B, shows multinodular extension (arrows) on equilibrium phase CT image. A B W257

6 4 Preoperative CT study Axial * Imaging investigator Extranodular extension Extranodular extension Axial Sagittal 1 * Coronal * Extranodular extension 2 Liver map TABLE 3: Multivariate Logistic Regression Analysis of Risk Factors in Patients With and Patients Without Microvascular Invasion Risk Factor Odds Ratio 95% Confidence Limits Lower Upper size Histologic grade Radiologic evidence of tumor capsule margin < Sagittal Axial Coronal Pathologist investigator MVI MVI Sagittal Coronal MVI Octant 1 Octant Match or not margins identified on dynamic CT (5-mm slice thickness) could predict the presence of MVI. A retrospective study by our group also showed that nonsmooth tumor margins on CT were predictive of MVI [14]. Highquality CT images to identify the tumor contour changes might be needed in the prediction of MVI. However, among the 60 HCC specimens with evidence of MVI, the location of MVI was found on the smooth border in 11 specimens with complete smooth margins and four specimens with focal extranodular extensions. Thus, the possibility of MVI occurring on a smooth tumor margin should be kept in mind. Chandarana et al. [22] reported that tumor multifocality was the only variable that was significantly correlated with the presence of MVI in explants. No other clinical, pathologic, or imaging features were useful in significantly predicting MVI [22]. The differences in findings between their study and our study might be because of different study designs, lower imaging resolution for their study (matrix, ), and smaller tumor size in their study cohort. Small tumor size might make the tumor contour harder to assess. Recent studies with gadoxetic acid showed that evidence of nonsmooth tumor margins in the hepatobiliary phase MRI was an indicator of MVI [15, 16]. An analysis of tumor margins in the hepatobiliary phase of gadoxetic acid enhanced MR images, therefore, seems to be useful in the preoperative prediction of MVI. In the current study, the univariate analyses revealed that larger tumor size, the presence of tumor capsules, and nonsmooth tumor margins on preoperative CT studies were associated with the presence of MVI; however, the multivariate analysis showed that only nonsmooth tumor margin was associated with MVI. On the basis of our re- Surgical specimen Slice 4, no. 2 Slice 1 Slice 2 Slice 3 Slice 4 Slice 5 Slice 6 Pathologist cut specimen along axial plane in 5-mm-thick increments according to markers made by surgeon and vascular anatomy Each slice was coded with a slice number Surgical specimen: slice 4 tissue block Slice 4, no. 1 Slice 4, no. 3 Slice 4, no. 4 Entire tumor margin on each slice was divided into multiple sequential blocks Each tissue block was coded with a slice number and a sequential tissue block number Histopathologic examination p 8 6 Radiologic 8 6 Pathologic MVI site of tumor was located according to slice number and sequential tissue block number Fig. 4 Flowchart shows steps in radiologic-pathologic correlation. MVI = microvascular invasion. (Drawing by Chou CT) W258

7 sults, we recommend that radiologists review tients who underwent anatomic hepatectomy operative prediction of the microvascular invasion and communicate any significant findings relating only are needed. of hepatocellular carcinoma with diffusion-weighted to tumor margins on preoperative CT reports and inform the surgeon of the possibility of MVI. Patients with HCC that has nonsmooth tumor margins may require more aggressive treatment, such as more extensive surgery with or without adjuvant therapy. Peritumoral enhancement has been shown to be suggestive of an increased risk of microvascular tumor invasion when seen on MR images [16] and on combined CT hepatic angiographic and CT arterioportographic images [23]. In our study, peritumoral enhancement on CT was not a significant risk factor for MVI. This discrepancy in findings might be due to differences among the different imaging modalities. Further studies of the relationship between peritumoral enhancement and MVI on dynamic CT images are needed. Adachi et al. [24] reported that blood vessels of the fibrous capsule were frequently invaded by cancer cells and stated that the presence of a fibrous capsule was a predictor of portal venous invasion. Witjes et al. [25] also reported that MVI was significantly associated with the presence of a capsule on dynamic contrast-enhanced MR images. Our results were similar to those in the studies by Adachi et al. and Witjes et al., in which radiologic evidence of a capsule correlated with MVI. However, Ariizumi et al. [15] reported that an incomplete capsule was likely to be associated with MVI. In other studies, the presence of fibrous capsules in HCC has been shown to be a favorable prognostic factor because the capsule may prevent invasion of HCC into the adjacent liver parenchyma [25, 26]. The relationship between the presence or absence of tumor capsules and MVI remains controversial [27]. This study has two limitations. First, the underlying liver disease in the majority (88/102) of patients in the study was viral hepatitis. When HCC develops in patients with chronic liver disease with underlying alcoholic complications, HCC may have a different clinical course. Thus, further studies are required to investigate MVI characteristics when HCC develops in patients with alcoholic liver disease. Second, 11 patients underwent hepatectomy without resection of the adjacent major vessels because of impaired liver function. With no major vascular landmarks as reference points, specimens from these patients might have caused some bias in the radiologic-pathologic correlation. However, the surgeon placed the referent markers to indicate the CT axial plane as accurately as possible. We believe this strategy might have minimized the bias. Future studies on the radiologic-pathologic correlation of MVI in pa- In conclusion, evidence of nonsmooth tumor margins on preoperative CT images is a predictor of MVI. The characteristics of tumor margins on preoperative CT scans should be considered when developing a treatment strategy for patients with HCC. References 1. Shah SA, Cleary SP, Wei AC, et al. Recurrence after liver resection for hepatocellular carcinoma: risk factors, treatment, and outcomes. Surgery 2007; 141: Ikai I, Arii S, Kojiro M, et al. Reevaluation of prognostic factors for survival after liver resection in patients with hepatocellular carcinoma in a Japanese nationwide survey. Cancer 2004; 101: Lauwers GY, Terris B, Balis UJ, et al. Prognostic histologic indicators of curatively resected hepatocellular carcinomas: a multi-institutional analysis of 425 patients with definition of a histologic prognostic index. Am J Surg Pathol 2002; 26: Hemming AW, Cattral MS, Reed AI, Van Der Werf WJ, Greig PD, Howard RJ. Liver transplantation for hepatocellular carcinoma. Ann Surg 2001; 233: Liver Cancer Study Group of Japan. The general rules for the clinical and pathological study of primary liver cancer, 2nd English ed. Tokyo, Japan: Kanehara, Shirabe K, Aishima S, Taketomi A, et al. Prognostic importance of the gross classification of hepatocellular carcinoma in living donor-related liver transplantation. Br J Surg 2011; 98: Eguchi S, Takatsuki M, Hidaka M, et al. Predictor for histological microvascular invasion of hepatocellular carcinoma: a lesson from 229 consecutive cases of curative liver resection. World J Surg 2010; 34: Hui AM, Takayama T, Sano K, et al. Predictive value of gross classification of hepatocellular carcinoma on recurrence and survival after hepatectomy. J Hepatol 2000; 33: Sumie S, Kuromatsu R, Okuda K, et al. Microvascular invasion in patients with hepatocellular carcinoma and its predictable clinicopathological factors. Ann Surg Oncol 2008; 15: Gouw AS, Balabaud C, Kusano H, Todo S, Ichida T, Kojiro M. Markers for microvascular invasion in hepatocellular carcinoma: where do we stand? Liver Transpl 2011; 17(suppl 2):S72 S Iwatsuki S, Dvorchik I, Marsh JW, et al. Liver transplantation for hepatocellular carcinoma: a proposal of a prognostic scoring system. J Am Coll Surg 2000; 191: Kornberg A, Freesmeyer M, Bärthel E, et al. 18 F- FDG-uptake of hepatocellular carcinoma on PET predicts microvascular tumor invasion in liver transplant patients. Am J Transplant 2009; 9: [Erratum in Am J Transplant 2009; 9:1255] 13. Suh YJ, Kim MJ, Choi JY, Park MS, Kim KW. Pre- imaging. Liver Transpl 2012; 18: Chou CT, Chen RC, Lee CW, Ko CJ, Wu HK, Chen YL. Prediction of microvascular invasion of hepatocellular carcinoma by pre-operative CT imaging. Br J Radiol 2012; 85: Ariizumi S, Kitagawa K, Kotera Y, et al. A nonsmooth tumor margin in the hepatobiliary phase of gadoxetic acid disodium (Gd-EOB-DTPA)-enhanced magnetic resonance imaging predicts microscopic portal vein invasion, intrahepatic metastasis, and early recurrence after hepatectomy in patients with hepatocellular carcinoma. J Hepatobiliary Pancreat Sci 2011; 18: Kim H, Park MS, Choi JY, et al. Can microvessel invasion of hepatocellular carcinoma be predicted by pre-operative MRI? Eur Radiol 2009; 19: Nakashima Y, Nakashima O, Tanaka M, Okuda K, Nakashima M, Kojiro M. Portal vein invasion and intrahepatic micrometastasis in small hepatocellular carcinoma by gross type. Hepatol Res 2003; 26: Edmondson HA, Steiner PE. Primary carcinoma of the liver: a study of 100 cases among 48,900 necropsies. Cancer 1954; 7: Bedossa P, Poynard T. An algorithm for the grading of activity in chronic hepatitis C: the META- VIR Cooperative Study Group. Hepatology 1996; 24: Kanai T, Hirohashi S, Upton MP, et al. Pathology of small hepatocellular carcinoma: a proposal for a new gross classification. Cancer 1987; 60: Nagano Y, Shimada H, Takeda K, et al. Predictive factors of microvascular invasion in patients with hepatocellular carcinoma larger than 5 cm. World J Surg 2008; 32: Chandarana H, Robinson E, Hajdu CH, Drozhinin L, Babb JS, Taouli B. Microvascular invasion in hepatocellular carcinoma: is it predictable with pretransplant MRI? AJR 2011; 196: Nishie A, Yoshimitsu K, Asayama Y, et al. Radiologic detectability of minute portal venous invasion in hepatocellular carcinoma. AJR 2008; 190: Adachi E, Maeda T, Kajiyama K, et al. Factors correlated with portal venous invasion by hepatocellular carcinoma: univariate and multivariate analyses of 232 resected cases without preoperative treatments. Cancer 1996; 77: Witjes CD, Willemssen FE, Verheij J, et al. Histological differentiation grade and microvascular invasion of hepatocellular carcinoma predicted by dynamic contrast-enhanced MRI. J Magn Reson Imaging 2012; 36: Ng IO, Lai EC, Ng MM, Fan ST. encapsulation in hepatocellular carcinoma: a pathologic study of 189 cases. Cancer 1992; 70: Torimura T, Ueno T, Inuzuka S, Tanaka M, Abe H, Tanikawa K. Mechanism of fibrous capsule formation surrounding hepatocellular carcinoma: immunohistochemical study. Arch Pathol Lab Med 1991; 115: W259