Cancer of the bile duct (cholangiocarcinoma or

Intraductal Papillary Neoplasm of the Bile Duct: A Biliary Equivalent to Intraductal Papillary Mucinous Neoplasm of the Pancreas? Flavio G. Rocha, 1 Hwajeong Lee, 2 Nora Katabi, 2 Ronald P. DeMatteo, 1 Yuman Fong, 1 Michael I. D Angelica, 1 Peter J. Allen, 1 David S. Klimstra, 2 and William R. Jarnagin 1 Intraductal papillary neoplasm of the bile duct (IPNB) is a variant of bile duct carcinoma characterized by intraductal growth and better outcome compared with the more common nodular-sclerosing type. IPNB is a recognized precursor of invasive carcinoma, but its pathogenesis and natural history are ill-defined. This study examines the clinicopathologic features and outcomes of IPNB. A consecutive cohort of patients with bile duct cancer (hilar, intrahepatic, or distal) was reviewed, and those with papillary histologic features identified. Histopathologic findings and immunohistochemical staining for tumor markers and for cytokeratin and mucin proteins were used to classify IPNB into subtypes. Survival data were analyzed and correlated with clinical and pathologic parameters. Thirty-nine IPNBs were identified in hilar (23/144), intrahepatic (4/86), and distal (12/113) bile duct specimens between 1991 and 2010. Histopathologic examination revealed 27 pancreatobiliary, four gastric, two intestinal, and six oncocytic subtypes; results of cytokeratin and mucin staining were similar to those of intraductal papillary mucinous neoplasm (IPMN ) of the pancreas. Invasive carcinoma was seen in 29/39 (74%) IPNBs. Overall median survival was 62 months and was not different between IPNB locations or subtypes. Factors associated with a worse median survival included presence and depth of tumor invasion, margin-positive resection, and expression of MUC1 and CEA. Conclusion: IPNBs are an uncommon variant of bile duct cancer, representing approximately 10% of all resectable cases. They occur throughout the biliary tract, share some histologic and clinical features with IPMNs of the pancreas, and may represent a carcinogenesis pathway different from that of conventional bile duct carcinomas arising from flat dysplasia. Given their significant risk of harboring invasive carcinoma, they should be treated with complete resection. (HEPATOLOGY 2012;56:1352-1360) Cancer of the bile duct (cholangiocarcinoma or adenocarcinoma of the extrahepatic bile ducts) can occur anywhere along the biliary tree. In Abbreviations: BilIN, biliary intraepithelial neoplasia; IPMN, intraductal papillary mucinous neoplasm; IPNB, intraductal papillary neoplasm of the bile duct. From the 1 Hepatopancreatobiliary Service, Department of Surgery, and the 2 Department of Pathology, Memorial Sloan-Kettering Cancer Center, New York, NY. Received January 23, 2012; accepted April 6, 2012. Flavio G. Rocha is currently affiliated with the Department of Surgery, Virginia Mason Medical Center, Seattle, WA. Address reprint requests to: William R. Jarnagin, M.D., Chief, Hepatopancreatobiliary Service, Department of Surgery, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10065. E-mail: jarnagiw@mskcc.org;. Copyright VC 2012 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.25786 Potential conflict of interest: Nothing to report. the United States, approximately 2000 to 3000 cases are diagnosed each year, representing 3% of all gastrointestinal malignancies. 1 Historically, tumors arising from the biliary confluence or hilar cholangiocarcinoma have been the most common, but recent epidemiologic data suggest a rising incidence of intrahepatic cholangiocarcinoma. 2 Papillary cholangiocarcinoma is a rare variant of this disease compared with the more common mass-forming or periductal infiltrating phenotypes in the intrahepatic biliary tree or the nodularsclerosing phenotype of extrahepatic biliary tumors. Due to the exophytic nature of the lesions and intraductal growth pattern, papillary cholangiocarcinomas appear to have a more favorable prognosis. We have previously demonstrated, in a cohort of patients with hilar cholangiocarcinoma, that the papillary phenotype was associated with improved survival. 3 1352

HEPATOLOGY, Vol. 56, No. 4, 2012 ROCHA ET AL. 1353 Several reports from Asia have suggested that the progression of papillary biliary neoplasia from benign lesions to invasive cancers may represent an analogous pathway to that of intraductal papillary mucinous neoplasm of the pancreas. 4-7 Because both the bile ducts and the pancreas develop from the ventral endoderm of the foregut, some authors have suggested that these may develop malignancies along similar genetic and molecular oncologic pathways. 8 However, although intrahepatic stones and liver fluke infestation have been associated with papillary lesions of the bile duct in the Far East, the etiology and malignant potential of this tumor have not been examined in detail. 9 In addition, unlike their pancreatic counterparts, very little is known about the molecular pathways and prognostic variables in papillary bile duct tumors. It was only recently that the World Health Organization recognized intraductal papillary neoplasm of the bile duct (IPNB) as a distinct pathologic entity. 10 The purpose of this study was to define the natural history of this lesion in a Western patient cohort and to examine its putative tumor markers. Patients and Methods This study was approved by the Institutional Review Board at Memorial Sloan-Kettering Cancer Center. All cases of resected intrahepatic, hilar, and extrahepatic bile duct tumors with gross and microscopic features of intraductal growth were retrieved from the surgical pathology files (1981-2010) using SNOMED II (Systematized Nomenclature of Medicine). Search criteria included the terms biliary, intraductal, cholangiocarcinoma, adenocarcinoma, and papillary. In addition, a prospectively maintained database of bile duct tumors from the Hepatopancreatobiliary Service since 1991 was queried, and all papillary bile duct tumors were selected for histologic review. Pancreatic intraductal papillary mucinous neoplasms, ampullary and periampullary tumors of uncertain origin, and cases with intraductal epithelial dysplasia/neoplasia without compelling evidence of intraductal mass formation (biliary intraepithelial neoplasia) were excluded. Individual and departmental consult cases, biopsy cases, and cases with insufficient tissue for analysis were excluded from the study. Patients with bile duct tumors containing intraductal growth between 1993 and 2010 were identified. Available slides from the retrieved cases (5-61 hematoxylin and eosin slides per case) were reviewed, and slides containing tumors (1-33 hematoxylin and eosin slides per case) were examined. Tumors were evaluated Table 1. Antibodies for Immunohistochemistry Antigen Manufacturer Dilution Clone Pretreatment Method CK7 Dako 1:1000 OV-TL Ventana CK20 Dako 1:1000 Ks20.8 Ventana CEA-m Dako 1:500 11-7 Citrate buffer Manual CDX2 Cell Marque 1:90 EPR2764Y Ventana MUC1 Vector 1:100 Ma695 Ventana MUC2 Vector 1:200 Ccp58 Citrate buffer Manual MUC5AC NovaCastro 1:400 CLH2 Citrate buffer Manual MUC6 BD Pharmingen 1:100 CLH5 Ventana B72.3 BioGenics 1:400 B72.3 Ventana CA125 Ventana Prediluted OC125 Ventana Mesothelin Vector 1:50 5B2 Ventana Ki67 Dako 1:200 Mib1 Ventana p53 Dako 1:500 DO7 Ventana SMAD4 Santa Cruz 1:800 B-8 Citrate buffer Manual Biotechnology HepPar1 Dako 1:500 OCH1E5 Citrate buffer Manual b-catenin BD Transduction 1:2000 14 Ventana for overall architecture (papillary, tubular, solid), epithelial subtype (pancreatobiliary, intestinal, oncocytic, gastric), grade of dysplasia, presence of extraductal invasion, depth of extraductal invasive component, percentage of intraductal invasion in relation to intraductal tumor volume, growth pattern of invasive components, lymphovascular invasion, perineural invasion, lymph node metastasis, and resection margin status. Information on gross tumor size and any associated cyst formation was obtained from surgical pathology reports. Anatomic location of the tumors was confirmed by pathologic, radiographic, and operative findings. Choledocholithiasis, history of parasite exposure, underlying liver disease including viral infection, and autoimmune liver disease were noted. Formalin-fixed and paraffin-embedded tissue blocks were available in 32 cases, and one to two representative blocks from each case were subjected to immunohistochemical staining. In one case, six stained slides from a tissue block were available for review without the original block. In another case, immunohistochemical staining was performed on a tissue block from a locally recurrent tumor. Staining by hematoxylin and eosin as well as 16 immunostains was performed on 4-lm-thick tissue sections from the representative tissue blocks (Table 1). The staining patterns were separately documented for the invasive and noninvasive components, when both components were present, on selected tissue blocks. A binary system (positive or negative) was employed for the evaluation of staining. The cutoff for positive staining was 10% for CDX2, mesothelin, CEA, B72.3, HepPar1, MUC1, MUC2, MUC5AC, and MUC6 and was 25% for CK7, CK20, and p53. The cutoff of positivity for CA125 and Ki67 was 50%.

1354 ROCHA ET AL. HEPATOLOGY, October 2012 Clinical data were obtained from the electronic medical record or outside medical reports. Demographic, preoperative evaluation, operative therapy, postoperative course, long-term outcome, and recurrence patterns were included. Survival (in months) was measured from the date of operation, to date of death or date of last follow-up. Statistical analysis was performed using Stata 7.0. Continuous variables were compared using the Student t test, and categorical variables were compared using a chi-square test. Survival probability was estimated using the Kaplan- Meier method, and comparison between groups was performed using a log-rank test. A P value of < 0.05 was considered statistically significant. Results Table 2. Clinicopathologic Features of IPNB Feature Value Patients, n 39 Mean age, years 67 Sex, n Men 26 Women 13 Mean tumor size, cm 4.5 Presence of cyst/mucin/stones, n 5/3/1 Presenting symptoms, n (%) Abdominal pain 15 (39) Jaundice 14 (36) Elevated liver enzymes 6 (15) Cholangitis 2 (5) Asymptomatic 2 (5) Location, n (%) Hilar 3/144 (16) Intrahepatic 4/86 (5) Distal 12/113 (11) Total 39/343 (11) Operation, n Left hepatectomy 14 Right hepatectomy 2 Extended right 6 Extended left 2 Pancreatico-duodenectomy 12 Bile duct excision 3 Invasive component, n (%) 29/39 (74) Lymph node metastasis, n (%) 1/33 (3) Overall median survival, months 62 5-Year survival, % 50 Recurrence, n Locoregional 7 Distant 7 Combined 6 Demographics. Thirty-nine (11%) cases of intraductal papillary neoplasm were identified from a total of 343 bile duct tumors resected from 1991 to 2010. The mean age was 67 years, and men (67%) were more commonly affected. The most common presenting symptom was abdominal pain (39%), followed by jaundice (36%) and elevations in liver function tests (15%); less common was cholangitis (5%); 5% of patients were asymptomatic (Table 2). The symptoms did not differ significantly between invasive and noninvasive lesions. Treatment. Five tumors had cystic dilatations of the bile ducts but only three contained gross mucin, and one tumor occurred in conjunction with intrahepatic stones. The majority of tumors were located at the biliary confluence (59%), followed by the distal common bile duct (31%); tumors were least common within the liver (10%). However, the highest frequency of invasive lesions was found in the distal bile duct (93%) followed by the hilus (65%) and liver (25%). The most commonly performed procedure was a left hepatectomy, in 16 patients, compared with a right hepatectomy, in eight patients, and 12 patients required a pancreaticoduodenectomy for tumor clearance. Three patients underwent a bile duct excision alone. Two patients received adjuvant chemotherapy, and another two patients received adjuvant chemoradiation therapy. Pathology. Papillary architecture was noted in 31/ 39 IPNBs, whereas the remaining eight cases expressed tubular architecture, including one carcinosarcoma with rare tubules. All tumors were classified according to epithelial subtypes of the intraductal component (Fig. 1). The most common epithelial subtype was pancreatobiliary (n ¼ 27 [69%]), followed by oncocytic (n ¼ 6, [15%]), gastric (n ¼ 4, [10%]), and intestinal (n ¼ 2 [5%]). More than one epithelial type was recognized in the same tumor in 14 cases, including one case with squamous type epithelium, but the subtype was designated based on the dominant pattern. An invasive component was found in 29/39 cases, and whereas 6/10 noninvasive cases exhibited highgrade dysplasia, all invasive cases had high-grade dysplasia in the intraductal components. Twenty-seven out of 29 IPNBs with invasive components exhibited tubular carcinomas, three of which included anaplastic, squamous, or papillary features within the invasive tumor without mucin. Of the remaining two cases, one case showed invasive colloid carcinoma, and the other showed mucinous features with a minimal invasive component precluding the histologic typing of the invasive carcinoma. The latter two carcinomas with mucin were associated with an oncocytic type IPNB. In one case, invasive tumor was tubular with pancreatobiliary subtype, while the intraductal counterpart was pure intestinal subtype. Of the 10 cases without

HEPATOLOGY, Vol. 56, No. 4, 2012 ROCHA ET AL. 1355 microscopic evidence of invasion, five cases showed gastric type epithelium in the intraductal tumor as either predominant epithelial type, or minor type mixed with another predominant type. Fig. 1. Histologic appearance of epithelial subtypes of IPNB (magnification 200). (A) Gastric-type epithelium resembling pyloric glands. (B) Intestinal-type epithelium resembling colonic adenocarcinoma with elongated, stratified, and hyperchromatic nuclei. (C) Oncocytic-type epithelium consisting of relatively uniform cells with abundant eosinophilic cytoplasm and centrally located nuclei with prominent nucleoli was associated with occasional intraepithelial mucin-containing lumen formation. (D) Pancreatobiliary-type epithelium resembling monolayered malignant biliary epithelium with marked nuclear pleomorphism and high nuclear-to-cytoplasmic ratio. Fig. 2. Kaplan-Meier survival estimates of overall survival (A) and disease-specific survival (B) according to depth of extraductal invasion (none, <5 mm, and 5 mm) and resection type (R0 versus R1). P < 0.05 was considered significant.

1356 ROCHA ET AL. HEPATOLOGY, October 2012 Table 3. Immunohistochemical Profiles of IPNB Subtype MUC1 MUC2 MUC5AC MUC6 CDX2 CK7 CK20 CEA Gastric/Pyloric 3 0 2 3 0 3 0 2 (n ¼ 3) Intestinal (n ¼ 2) 0 1 0 0 2 2 2 2 Oncocytic (n ¼ 5) 3 2 5 4 0 3 0 1 Pancreatobiliary (n ¼ 23) 19 3 10 13 4 22 3 11 Of the 29 invasive carcinomas, 16 had <5 mmof extraductal invasion, whereas 13 had 5 mm. Lymphovascular invasion was noted in 6 patients and perineural invasion was noted in nine patients. Of 33 patients submitted to a concomitant porta hepatis lymphadenectomy, only one had nodal metastases (1/22 lymph nodes involved with cancer). The majority of invasive carcinomas (20/29) were moderately differentiated, with only one well-differentiated and eight poorly differentiated. Immunohistochemical staining (Table 3) revealed that the majority of pancreatobiliary subtypes expressed MUC1 (19/23) and CK7 (22/23), whereas very few expressed MUC2 (3/23) or CDX2 (4/23). Similarly, the gastric subtypes all expressed MUC1 (3/3) and CK7 (3/3), but none expressed MUC2 or CDX 2. In contrast, intestinal-type tumors were MUC2þ (1/2), MUC1 (0/2), and both CK7 and CK20 were positive (2/2). Oncocytic tumors were MUC5ACþ (5/5) and MUC6þ (4/5). Invasive carcinoma components were noted in 100% of the intestinal tumors, 80% of pancreatobiliary tumors, 60% of oncocytic tumors, and 25% of gastric tumors. Outcome. There was one operative mortality (3%) due to overwhelming multiorgan failure after extended right hepatectomy. There were 21 complications in 15 patients with an overall morbidity rate of 38%. Intraabdominal abscess requiring drainage accounted for 29% (6/21) of the complications, whereas gastrointestinal complications of delayed gastric emptying, ascites, and anastomotic leak occurred in two patients each. One patient developed respiratory failure; two patients developed postoperative hemorrhage; one patient developed a deep venous thrombosis. Margin-negative resection was achieved in 30 (77%) patients, and nine (23%) patients had an R1 resection. Overall median survival was 62 months for the entire cohort and 49 months for the cases with an invasive component at a median follow-up of 86 months. Of the 20 documented recurrences, seven were locoregional, seven were distant, and six were both. In patients with R1 resections, 7/9 recurred compared with 11/30 in patients with R0 resections. Of the 10 patients with noninvasive tumors, all had R0 resections and two had a recurrence with an invasive component (one local, one distant), both at 5 years after resection. On univariate analysis, neither age nor sex was associated with survival. In addition, there was no difference in survival based on location of the primary tumor (hilar, distal, or intrahepatic) or epithelial subtypes. R0 resection was associated with an improved median survival of 82 months, compared with 36 months in the R1 resection group (P < 0.04). Both presence and depth of an invasive component correlated with survival. Depth of invasion, graded as 5 mm, <5 mm, and none, was associated with a survival of 39 months, 128 months, and 144 months, respectively (P < 0.007) (Fig. 2). In addition, percentage of invasive carcinoma components, graded as 10%, <10%, and none, was associated with a survival of 42 months, 128 months, and 144 months, respectively (P < 0.03). Patients with poorly differentiated carcinomas as well as those with lymphovascular invasion also suffered a worse outcome (Table 4). However, margin status did not affect survival in the cases with an invasive component. Of the immunohistochemical stains used in this study, only MUC1 and CEA were associated with a poorer prognosis. Patients whose tumors expressed MUC1 had a median survival of 58 months versus144 months in patients whose tumors lacked MUC1 expression (P < 0.009). The median survival of patients with CEAþ tumors was 42 months versus 128 months in patients with CEA tumors (P < 0.01). When considering only the cases with an invasive component, MUC1 expression remained a statistically significant predictor of survival, whereas CEA lost significance. None of the other MUC proteins (CA 125, mesothelin, Ki67, p53, B72.3, or HepPar-1) was associated with survival (Table 5). Discussion Bile duct carcinoma usually develops through a multistep process involving one of two separate precursor pathways. The initial lesion is flat biliary dysplasia (biliary intraepithelial neoplasia [BilIN]) found in both the intra- and extrahepatic biliary tree and graded in severity as 1 (mild dysplasia), 2 (moderate dysplasia), or 3 (severe dysplasia/carcinoma in situ). This sequence resembles the progression of pancreatic intraepithelial neoplasia in pancreatic ductal adenocarcinoma. A less common variant is the intraductal papillary neoplasm characterized by exophytic proliferation of biliary epithelium on fibrovascular stalks within the bile duct

HEPATOLOGY, Vol. 56, No. 4, 2012 ROCHA ET AL. 1357 Table 4. Univariate Analysis of Clinicaopathologic Features Feature n Median Survival (Months) P Value Clinical features Age, years 0.42 70 21 59 <70 18 82 Sex 0.89 Men 26 62 Women 13 83 Subtype 0.69 Gastric/Pyloric 4 58 Intestinal 2 NR Oncocytic 6 82 Pancreatobiliary 27 53 Location 0.31 Hilar 23 82 Intrahepatic 4 33 Distal 12 49 Resection type 0.04 R0 30 82 R1 9 36 Pathologic features Depth of invasion 0.007 None 10 10 <5 mm 16 16 5 mm 13 13 Extraductal invasion 0.03 None 10 10 <10% 13 13 10% 16 16 Grade of invasion 0.04 Well 1 NR Moderate 20 62 Poor 8 32 Perineural invasion 0.10 Absent 20 53 Present 9 39 Lymphovascular invasion 0.04 Absent 23 53 Present 6 28 Abbreviation: NR, not reached. lumen, sometimes associated with mucin hypersecretion and cystic dilatation of the affected bile ducts. 11 These tumors can be entirely intraductal or they can have an associated invasive carcinoma component, and although many different terms have been used for this spectrum of neoplasms, recently they have been collectively termed IPNB. IPNBs can be separated from mucinous cystic neoplasms of the liver by their origin and connection to the biliary system. In addition, the latter lesion is characterized by an ovarian-type stroma resembling pancreatic mucinous cystic neoplasms. 12 In the present study, 12% of resected bile duct carcinomas met the pathologic definition of IPNB. 10 Previous reports found the incidence of IPNBs among all bile duct carcinomas to range from 7% to 38%. 13,14 However, unlike prior studies, mainly from Asia, there was a lack of association with hepatolithiasis or liver flukes in this cohort, as only one patient was found to have an intrahepatic stone and no infestations were noted. In addition, gross mucin production was present in only three cases. In two recent reports from Asia, tumors that secreted mucin were associated with decreased invasion and better survival, 15,16 but the protective role of mucin-producing tumors could not be assessed in the current study. The majority of lesions were found in the hilum and left-sided biliary ductal system. However, despite its variable location, the primary site did not affect the course of the disease or its prognosis, as there was no difference in survival according to anatomic site of origin. This finding was not surprising, because our previous analysis of extrahepatic bile duct carcinomas failed to demonstrate a difference in surgical outcomes or disease-specific survival between proximal and distal tumors after R0 resection. 17 Another study of patients with IPNB did not find any clinical or pathologic differences based on primary tumor location. 18 Table 5. Univariate Analysis of Tumor Markers Tumor Marker n Median Survival (in Months) P Value MUC1 0.009 Positive 22 58 Negative 12 144 MUC2 0.82 Positive 6 82 Negative 28 62 MUC5AC 0.85 Positive 17 82 Negative 17 62 MUC6 0.77 Positive 19 59 Negative 14 128 CDX2 0.14 Positive 6 42 Negative 27 82 CA125 0.33 Positive 5 33 Negative 28 128 CEA 0.01 Positive 16 42 Negative 17 128 Mesothelin 0.38 Positive 20 82 Negative 13 53 Ki67 0.66 Positive 14 59 Negative 19 82 p53 0.66 Positive 15 NR Negative 18 59 B72.3 0.17 Positive 21 128 Negative 13 53 HepPar1 0.56 Positive 14 82 Negative 20 62 Abbreviation: NR, not reached.

1358 ROCHA ET AL. HEPATOLOGY, October 2012 The immunohistochemical profiles of IPNBs in the present study were very similar to those published in cholangiocarcinoma as well as in pancreatic intraductal papillary mucinous neoplasm (IPMN). Zen et al. 9 initially reported a series of 110 cases of biliary neoplasm associated with hepatolithiasis, separating cases of BilIN and IPNB by cytokeratin and mucin staining. Cholangiocarcinomas arising in association with BilIN progressed to tubular adenocarcinomas, while IPNBassociated tumors progressed to either tubular or colloid carcinomas in that study of Asian patients. A follow-up report comparing benign papillomatosis, noninvasive and invasive papillary cholangiocarcinomas, nonpapillary bile duct carcinomas, and IPMNs of the pancreas demonstrated that the pancreatobiliary subtype was present in 50% of papillary bile duct lesions and 100% of nonpapillary bile duct carcinomas. In contrast, most pancreatic IPMNs were of the gastric or intestinal subtype. 4 MUC1 expression was found in all nonpapillary cholangiocarcinomas, compared with only 40% of IPNBs and 29% of IPMNs. Several studies have confirmed that the expression of MUC1 in cholangiocarcinoma is associated with poorer survival, suggesting that this phenotype is associated with more aggressive behavior. 19,20 The survival of patients with biliary papillary tumors also depended on the presence and histologic type of invasive carcinoma. Those with tubular adenocarcinomas had a prognosis similar to patients with nonpapillary tumors, whereas those with mucinous or colloid adenocarcinomas had a better prognosis, comparable to that of colloid carcinoma of the pancreas arising in association with an IPMN. In our cohort of IPNB, the majority of tumors (27/39) were pancreatobiliary subtype and MUC1þ, which was associated with worse survival. A similar pattern of expression was seen in a smaller study where 7/9 IPNBs with pancreatobiliary subtype were MUC1þ, with one being MUC2þ, whereas 6/8 IPNBs with intestinal subtype were MUC2þ and none were stained for MUC1. 21 We could only identify two cases of intestinal subtype, and although they were both MUC1þ and invasive, only one was associated with a mucinous carcinoma, whereas the other consisted of a tubular carcinoma. A larger study of 97 patients from Korea demonstrated a statistically significant increase of invasive tubular carcinoma in IPNBs with a pancreatobiliary subtype compared with other subtypes and was also associated with a poorer prognosis. 22 When we compared the immunohistochemical profile of IPNBs to that of a series of 43 resected IPMNs of the pancreas, 23 we noted several important similarities and differences. First, the majority of pancreatic tumors were gastric (56%) or intestinal (30%) subtypes, whereas only 9% were pancreatobiliary. Second, only 21% of IPMNs had an invasive component compared with 72% of IPNBs. So although both shared similar phenotypes and marker expression, the pattern shifted to the more aggressive MUC1-expressing pancreatobiliary subtype in IPNB, compared with the more indolent MUC2-expressing intestinal subtype in IPMN of the pancreas. A recent study from Europe directly comparing 20 IPNBs and 29 IPMNs supported the findings of the present study, whereby 85% of IPNBs contained a carcinoma in situ or invasive component, compared with 52% of IPMNs. 24 In addition, the most frequent phenotype in the IPNBs was pancreatobiliary (45%), followed by gastric (25%), intestinal (20%), and oncocytic (10%). In IPMNs the most frequent phenotype was intestinal (59%), followed by gastric (34%), and pancreatobiliary (7%). Interestingly, most tumors were colloid-type in both groups, 75% of IPNBs and 97% of IPMNs. Lymph node metastases were found in four and two patients, respectively. As in our previous study, the degree of invasion was highly correlated with outcome. When IPNBs were stratified by invasiveness, a significant survival advantage was seen in tumors with 0 mm to <5 mm of invasion. In fact, IPNBs that had a more than 10% invasive component in our previous series had an outcome that was identical to that of conventional nodular-sclerosing tumors. 3 The World Health Organization has recommended that when an invasive carcinoma is present in IPNBs, it should be separately staged. 10 Hong et al. 25 has proposed that the depth of invasive carcinoma should be incorporated into the American Joint Committee on Cancer staging guidelines. These authors noted that in 147 patients undergoing pancreaticoduodenectomy for distal bile duct carcinoma, those with >12 mm of invasion had the worst median survival at 12.9 months compared with those with 5-12 mm of invasion, who had a median survival of 28.9 months (P < 0.0001). Conversely, 78.6% of patients with <5 mm of invasion were still alive. On multivariate analysis, depth of invasion remained the strongest predictor of survival with a hazard ratio of 6.69 (confidence interval, 1.94-23.08; P < 0.003). Another study of 222 patients with extrahepatic bile duct tumors used recursive partitioning to discriminate optimal cutoff points for depth of invasion that would be predictive of survival. 26 As depth of invasion increased, survival decreased, with the slope of the regression line changing abruptly at 5 mm and 12 mm of invasion. By stratifying patients into three groups (<5 mm of invasion, 5-12 mm of invasion,

HEPATOLOGY, Vol. 56, No. 4, 2012 ROCHA ET AL. 1359 and >12 mm of invasion), a significant shift toward more lymph node involvement and decreased survival was observed with deeper lesions. The major limitations of this study lie in the retrospective nature of the data collection. First, only a minority of patients could be identified preoperatively as having an IPNB by imaging, thereby limiting our ability to screen appropriately for this diagnosis. Second, all patients in the study underwent resection and were therefore subject to significant selection bias. It is certainly possible that those with either very early or advanced IPNB never came to operation. Despite these limitations, a few interesting observations can be extracted. First, the majority of IPNBs (74%) contained a component of invasive carcinoma. Previous single institution series of IPNBs have corroborated this observation, noting an incidence of invasive cancer in 70%-80% of specimens. 11,27 On the other hand, IPMNs of the pancreas have a much lower frequency of invasive carcinoma, with only 48% of resected cases containing an invasive component in a previous report from our institution, and in 21% of our recent comparison group. 23,28 Despite this observation, the survival of patients with IPNB has been consistently better than those with conventional bile duct carcinoma in several reports. 14,29,30 It is unclear whether this is due to the inherent biology of these tumors or to the primarily intraductal growth pattern of IPNB that may predispose to an early diagnosis from biliary obstruction before invasion into surrounding tissues occurs. Cholangiocarcinoma has significant heterogeneity with respect to expression of tumor markers and genes. 31,32 More detailed molecular analyses of methylation status and microsatellite instability have demonstrated that the genetic events involved in the pathogenesis of IPNB resemble those found in conventional cholangiocarcinoma, but retain some important differences highlighting the potential for two potentially overlapping pathways for the development of biliary tract cancer. 33,34 In conclusion, IPNBs represent a small proportion of biliary tumors that have some pathologic similarities to IPMN of the pancreas. These tumors are associated with better outcome compared with nonpapillary bile duct carcinoma and may represent an alternative carcinogenesis pathway in the biliary tract. Given their predisposition to invasion and the uncertainty in the preoperative diagnosis of these lesions, they should all be resected with negative margins in order to achieve long-term survival. References 1. Lazaridis KN, Gores GJ. Cholangiocarcinoma. Gastroenterology 2005; 128:1655-1667. 2. Endo I, Gonen M, Yopp AC, Dalal KM, Zhou Q, Klimstra D, et al. Intrahepatic cholangiocarcinoma: rising frequency, improved survival, and determinants of outcome after resection. Ann Surg 2008;248:84-96. 3. Jarnagin WR, Bowne W, Klimstra DS, Ben-Porat L, Roggin K, Cymes K, et al. Papillary phenotype confers improved survival after resection of hilar cholangiocarcinoma. Ann Surg 2005;241:703-712; discussion 712-714. 4. Zen Y, Fujii T, Itatsu K, Nakamura K, Minato H, Kasashima S, et al. Biliary papillary tumors share pathological features with intraductal papillary mucinous neoplasm of the pancreas. HEPATOLOGY 2006;44: 1333-1343. 5. Ji Y, Fan J, Zhou J, Wang BS, Liu HB, Wu ZW, et al. Intraductal papillary neoplasms of bile duct. A distinct entity like its counterpart in pancreas. Histol Histopathol 2008;23:41-50. 6. Suh KS, Roh HR, Koh YT, Lee KU, Park YH, Kim SW. Clinicopathologic features of the intraductal growth type of peripheral cholangiocarcinoma. HEPATOLOGY 2000;31:12-17. 7. Nanashima A, Kinoshita N, Nakanuma Y, Zen Y, Sumida Y, Abo T, et al. Clinicopathological features of intraductal papillary neoplasm of the bile duct and patient outcome after surgical resection. Hepatogastroenterology 2008;55:1167-1173. 8. Nakanuma Y. A novel approach to biliary tract pathology based on similarities to pancreatic counterparts: is the biliary tract an incomplete pancreas? Pathol Int 2010;60:419-429. 9. Zen Y, Sasaki M, Fujii T, Chen TC, Chen MF, Yeh TS, et al. Different expression patterns of mucin core proteins and cytokeratins during intrahepatic cholangiocarcinogenesis from biliary intraepithelial neoplasia and intraductal papillary neoplasm of the bile duct an immunohistochemical study of 110 cases of hepatolithiasis. J Hepatol 2006;44: 350-358. 10. Bosman FT, Carneiro F, Hruban RH, eds. WHO Classification of Tumours of the Digestive System. World Health Organization Classification of Tumours. Lyon, France: IARC; 2010. 11. Nakanuma Y, Zen Y, Harada K, Ikeda H, Sato Y, Uehara T, et al. Tumorigenesis and phenotypic characteristics of mucin-producing bile duct tumors: an immunohistochemical approach. J Hepatobiliary Pancreat Sci 2010;17:211-222. 12. Zen Y, Pedica F, Patcha V, Capelli P, Zamboni G, Casaril A, et al. Mucinous cystic neoplasms of the liver: a clinicopatohlogical study and comparison with intraductal papillary neoplasms of the bile duct. Mod Pathol 2011;24:1079-1089. 13. Barton JG, Barrett DA, Maricevich MA, Schnelldorfer T, Wood CM, Smyrk TC, et al. Intraductal papillary mucinous neoplasm of the biliary tract: a real disease? HPB (Oxford) 2009;11:684-691. 14. Yeh CN, Jan YY, Yeh TS, Hwang TL, Chen MF. Hepatic resection of the intraductal papillary type of peripheral cholangiocarcinoma. Ann Surg Oncol 2004;11:606-611. 15. Ohtsuka M, Kimura F, Shimizu H, Yoshidome H, Kato A, Yoshitomi H, et al. Similarities and differences between intraductal papillary tumors of the bile duct with and without macroscopically visible mucin secretion. Am J Surg Pathol 2011;35:512-521. 16. Yang J, Wang W, Yan L. The clinicopathological features of intraductal papillary neoplasms of the bile duct in a Chinese population. Dig Liver Dis 2012;44:251-256. 17. Allen PJ, Reiner AS, Gonen M, Klimstra DK, Blumgart LH, Brennan MF, et al. Extrahepatic cholangiocarcinoma: a comparison of patients with resected proximal and distal lesions. HPB (Oxford) 2008;10:341-346. 18. Choi SC, Lee JK, Jung JH, Lee JS, Lee KH, Lee KT, et al. The clinicopathological features of biliary intraductal papillary neoplasms according to the location of tumors. J Gastroenterol Hepatol 2010;25:725-730. 19. Matsumura N, Yamamoto M, Aruga A, Takasaki K, Nakano M. Correlation between expression of MUC1 core protein and outcome after surgery in mass-forming intrahepatic cholangiocarcinoma. Cancer 2002;94:1770-1776. 20. Park SY, Roh SJ, Kim YN, Kim SZ, Park HS, Jang KY, et al. Expression of MUC1, MUC2, MUC5AC and MUC6 in cholangiocarcinoma: prognostic impact. Oncol Rep 2009;22:649-657.

1360 ROCHA ET AL. HEPATOLOGY, October 2012 21. Sclabas G, Barton J, Smyrk T, Barrett DA, Khan S, Kendrick ML, et al. Frequency of subtypes of biliary intraductal papillary mucinous neoplasm and their MUC1, MUC2, and DPC4 expression patters differ from pancreatic intraductal papillary mucinous neoplasm. J Am Coll Surg 2012;214:27-32. 22. Kim K, Lee J, Shin J, Lee KH, Lee KT, Sung JY, et al. Clinicopathologic features of intraductal papillary neoplasm of the bile duct according to histologic subtype. Am J Gastroenterol 2012;107:118-125. 23. Lee H, Rocha FG, Klimstra DS, Young GQ, Jarnagin WR, Katabi N. Intraductal papillary neoplasm of the bile duct (IPN): clinicopathologic study of 39 cases in comparison to pancreatic intraductal papillary mucinous neoplasm (IPMN) in a single North American institution [Abstract]. Lab Invest 2011;91:367A. 24. Kloek JJ, van der Gaag NA, Erdogan D, Rauws EA, Busch OR, Gouma DJ, et al. A comparative study of intraductal papillary neoplasia of the biliary tract and pancreas. Hum Pathol 2011;42: 824-832. 25. Hong SM, Pawlik TM, Cho H, Aggarwal B, Goggins M, Hruban RH, et al. Depth of tumor invasion better predicts prognosis than the current American Joint Committee on Cancer T classification for distal bile duct carcinoma. Surgery 2009;146:250-257. 26. Hong SM, Cho H, Moskaluk CA, Yu E. Measurement of the invasion depth of extrahepatic bile duct carcinoma: an alternative method overcoming the current T classification problems of the AJCC staging system. Am J Surg Pathol 2007;31:199-206. 27. Paik KY, Heo JS, Choi SH, Choi DW. Intraductal papillary neoplasm of the bile ducts: the clinical features and surgical outcome of 25 cases. J Surg Oncol 2008;97:508-512. 28. D Angelica M, Brennan MF, Suriawinata AA, Klimstra D, Conlon KC. Intraductal papillary mucinous neoplasms of the pancreas: an analysis of clinicopathologic features and outcome. Ann Surg 2004;239:400-408. 29. Tajima Y, Kuroki T, Fukuda K, Tsuneoka N, Furui J, Kanematsu T. An intraductal papillary component is associated with prolonged survival after hepatic resection for intrahepatic cholangiocarcinoma. Br J Surg 2004;91:99-104. 30. Hoang MP, Murakata LA, Katabi N, Henson DE, Albores-Saavedra J. Invasive papillary carcinomas of the extrahepatic bile ducts: a clinicopathologic and immunohistochemical study of 13 cases. Mod Pathol 2002;15:1251-1258. 31. Hezel AF, Deshpande V, Zhu AX. Genetics of biliary tract cancers and emerging targeted therapies. J Clin Oncol 2010;28:3531-3540. 32. Albores-Saavedra J, Murakata L, Krueger JE, Henson DE. Noninvasive and minimally invasive papillary carcinomas of the extrahepatic bile ducts. Cancer 2000;89:508-515. 33. Abraham SC, Lee JH, Boitnott JK, Argani P, Furth EE, Wu TT. Microsatellite instability in intraductal papillary neoplasms of the biliary tract. Mod Pathol 2002;15:1309-1317. 34. Abraham SC, Lee JH, Hruban RH, Argani P, Furth EE, Wu TT. Molecular and immunohistochemical analysis of intraductal papillary neoplasms of the biliary tract. Hum Pathol 2003;34:902-910.