Journal of Hepato-Biliary-Pancreatic Sciences

, Volume 17, Issue 4, pp 490–496

Hilar cholangiocarcinoma: the Memorial Sloan-Kettering Cancer Center experience


  • Flavio G. Rocha
    • Hepatopancreatobiliary Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center
  • Kenichi Matsuo
    • Hepatopancreatobiliary Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center
  • Leslie H. Blumgart
    • Hepatopancreatobiliary Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center
    • Hepatopancreatobiliary Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center
Topics A new era of surgical treatment of hilar cholangiocarcinoma: comparison of leading Eastern and Western centers

DOI: 10.1007/s00534-009-0205-4

Cite this article as:
Rocha, F.G., Matsuo, K., Blumgart, L.H. et al. J Hepatobiliary Pancreat Sci (2010) 17: 490. doi:10.1007/s00534-009-0205-4



Hilar cholangiocarcinoma (HCCA) is a rare cancer with a low resectability rate, frequent recurrence after resection and an overall poor outcome. It is widely accepted that en bloc partial hepatectomy is a necessary part of the surgical therapy, but controversy surrounds other areas, including extent of lymphadenectomy and preoperative use of biliary drainage of the future liver remnant (FLR). This study analyzes the authors’ experience with HCCA, emphasizing outcome after resection in a more recent cohort.


All patients with HCCA evaluated at Memorial Sloan-Kettering Cancer Center (MSKCC) since 1991 were included in the initial analysis. Outcome after resection was specifically assessed in patients submitted to operation between January 2001 and September 2008. Patient demographics, preoperative evaluation, resection type, margin status, lymph node status, complications, morbidity and survival were examined. Preoperative disease staging was performed in all patients according to the Blumgart classification. Separate analyses were conducted to assess the impact of preoperative biliary drainage on the FLR and the optimal lymph node harvest. Outcomes for resected patients were analyzed by Fisher’s exact test and log rank tests.


Three hundred and fifty-two patients with HCCA were evaluated since 1991, of which 118 were seen between 2001 and 2008. During this latter period, 105 (89%) patients underwent exploration, and of the 60 patients that underwent resection with curative intent, 48 (80%) had R0 resections. There were 3 perioperative deaths (5%), and 22 (28%) patients had complications. Patients with an R0 resection had the highest disease-specific survival followed by those with R1 resection when compared to unresected patients. The median follow-up period was 18 months. Classification by the Blumgart preoperative staging system predicted resectability and the likelihood of R0 resection. The benefit of pre-operative biliary drainage of the FLR appeared to be limited to patients with a predicted FLR volume of <30%. In patients with node-negative tumors, survival was greater in those with more than 7 lymph nodes harvested.


R0 resection including hepatectomy with negative lymph nodes is feasible in the majority of patients with resectable HCCA. This strategy is associated with a prolonged disease-specific survival.


CholangiocarcinomaBile duct cancerKlatskin tumor


Hilar cholangiocarcinoma, originally described by Altemeier and Klatskin [1, 2], is defined as adenocarcinoma of the extrahepatic biliary tree, arising from the biliary confluence and/or the main left or right hepatic ducts. These tumors are distinct from intrahepatic cholangiocarcinoma, which arise from small peripheral biliary radicles but may secondarily involve the biliary confluence. Previously, HCCA comprised the majority (40–60%) of all cholangiocarcinomas and represented 3% of all gastrointestinal malignancies [35], although more recently, intrahepatic cholangiocarcinoma has become more prevalent [6].

Surgical therapy, either resection or transplantation, is the most effective treatment and the only intervention associated with prolonged survival [7]. However, the majority of patients, nearly two-thirds in some series, present with disease that is beyond surgical correction. Portal vein invasion and lobar atrophy are frequent findings in patients with HCCA, and although not necessarily indicative of irresectability, they are associated with both locally advanced tumors and metastatic disease. Over the past several years, data from a number of centers have shown that ipsilateral en bloc hepatectomy is generally required to achieve a complete resection [811].

There is little doubt that a concomitant partial hepatectomy is a necessary part of the surgical approach to HCCA; however, other management practices are much less clear cut. For example, en bloc hepatectomy enhances the R0 resection rate but also adds substantially to perioperative morbidity and mortality. This observation has led many to advocate routine preoperative, ipsilateral portal vein embolization and biliary decompression of the FLR, although clear evidence supporting this practice in all patients is lacking [12, 13]. Additionally, the optimal extent of regional lymphadenectomy during resection and the impact of nodal metastases on survival remain areas of debate.

The present study analyzes a consecutive cohort of patients with HCCA evaluated at MSKCC. The objectives of the study were to assess outcome after resection, with emphasis on results in more recently treated patients. We also sought to assess the impact of preoperative biliary drainage on perioperative outcome and to further define the role and extent of lymphadenectomy.


Consecutive patients with HCCA seen and evaluated at MSKCC since December 1991 were entered into a prospectively maintained database in the Department of Surgery. Permission for this study was obtained from the MSKCC Institutional Review Board; HIPAA compliance was assured. This study included only patients with adenocarcinoma arising from the proximal extrahepatic biliary tree (biliary confluence or right or left hepatic ducts); patients with other histologies were excluded. It should be noted that some patients included in the current analysis have been included in prior studies.

Patients were identified from the database, which contained demographic, laboratory, radiographic, histopathologic, operative, perioperative and follow-up data, and analyzed retrospectively. Additional data were obtained, as necessary, from review of the medical record. Final disease staging was based on the 6th Edition of the American Joint Commission on Cancer Staging Manual. Tumor grade was divided into well or not well-differentiated (moderate or poor). An R0 resection was defined as the absence of disease at all resection margins, while any margin involvement defined an R1 resection. All post-operative adverse events were recorded and graded according to a severity scale, as previously described. Post-operative mortality was defined as death related to a post-operative complication, whenever it occurred.

For all patients, the Blumgart preoperative clinical staging system was used to assess extent of the primary lesion. This staging scheme takes into account local tumor factors that influence resectability—biliary tumor extent, portal venous involvement, lobar atrophy—and has been previously described [10, 20]. The Blumgart stage was correlated with resectability and the presence of nodal or distant metastatic disease.

The authors’ general preoperative assessment of patients with confirmed or suspected HCCA has been described previously [10, 20]. Disease extent was evaluated with a combination of complementary imaging studies, including duplex ultrasonography, cross-sectional imaging (CT angiography or MRI), and either direct or indirect cholangiography (MRCP). Positron emission tomography (PET) was used selectively. All outside histology slides were reviewed by MSKCC pathologists. Most patients were referred after one or more attempts at biliary decompression. In some of these cases, additional procedures were required to correct malpositioned endoscopic or percutaneous stents. All cases were reviewed at a multidisciplinary hepatobiliary conference regardless of stage.

The authors’ operative approach to resection of HCCA is documented in prior studies [10, 20]. Hilar tumors were typically treated with partial hepatectomy, cholecystectomy, bile duct resection and porta hepatis lymphadenectomy en bloc. Staging laparoscopy was performed selectively just prior to laparotomy to exclude metastatic disease in patients with advanced disease or suspicious radiographic findings. Operative assessment for tumor resectability was performed according to previously described methods. Exposure of the biliary confluence and assessment of vascular involvement were accomplished by early transection of the common bile duct at the level of the duodenum with reflection superiorly. Resection of all lymph nodes within the porta hepatis, from the level of the common hepatic artery on the left and the retroduodenal area on the right and extending upwards to the base of the liver was routinely performed. Caudate lobectomy was performed for all tumors involving the left hepatic duct and in selected additional cases arising from the right hepatic duct in order to achieve complete tumor resection. Biliary continuity was achieved by Roux-en-y hepaticojejunostomy with an isoperistaltic 70 cm limb of jejunum. Patients deemed to be unresectable at presentation or intraoperatively were treated with systemic chemotherapy, with or without external beam radiation therapy.

Survival and perioperative outcome were analyzed for patients submitted to resection between January 2001 and December 2008. For an expanded cohort, the total lymph node count (TLNC) was analyzed as a survival variable; differences in survival according to TLNC were assessed in an effort to determine the optimal lymph node harvest during lymphadenectomy. In a smaller subset of patients, the impact of preoperative biliary drainage of the FLR on perioperative outcome was assessed [26]. For this analysis, patients were chosen based on availability of preoperative imaging studies suitable for semi-automated calculation of the FLR volume, which was determined retrospectively based on the actual resection performed. Studies pertaining to the biliary drainage procedure(s), if performed, were further analyzed to determine if the FLR had been decompressed; perioperative outcome (morbidity, liver insufficiency and mortality) was then analyzed, stratified by both FLR volume (<30 vs. ≥30%) and FLR biliary drainage.

Statistical analyses were performed using Stata 7.0 and Statistical Package for the Social Sciences (SPSS, version 16.0). Continuous variables were compared with Student’s t test while categorical variables were compared with chi-square and Fisher’s exact tests. Logistic regression was used to determine independent predictors of outcome. Disease-specific survival (DSS) was estimated by the method of Kaplan and Meier and comparison between groups was done using a log rank test.


Since December 1991, 352 consecutive patients with HCCA have been evaluated at MSKCC (Table 1). Two hundred and seventy-one (77%) were considered for an operation but only 146 (54%) had an attempt at complete resection. Stated otherwise, 206 patients, or 59% of the entire cohort, had unresectable disease either at presentation or exploration. The reasons for irresectability were metastatic disease in 109 (53%) patients, locally advanced disease in 71 (34%) patients, medical comorbidites precluding a major operation in 22 (11%). Four patients either refused surgery or pursued therapy elsewhere. Of those that had a complete gross resection, 111 (76%) had clear surgical margins (R0). Perioperative mortality was 8% (12 of 146) and overall median survival was 17.7 months.
Table 1

Hilar cholangiocarcinoma at MSKCC: summary of patients evaluated 1991–2008


352 patients

Age (mean)

65 years

Male gender (%)

203 (58%)

AJCC stage


68 (19%)


43 (12%)


39 (11%)


59 (17%)


21 (6%)


112 (33%)

 N/A or lost to follow-up

7 (2%)

Unresectable at presentation

81 (23%)

 Metastatic disease


 Locally advanced


 Medically unfit for operation


 Refused or lost to follow-up


Unresectable at exploration

125 (46%)



 Locally advanced


 Medically unfit



146 (54%)





Operative mortality (%)

12 (8%)

Survival (all patients, median)

17.7 months

From January 2001 to September 2008, 118 patients with HCCA were evaluated at MSKCC. There were 75 men with a mean age of 64 years and median follow-up time of 18 months. The management of these patients is outlined in Fig. 1. Thirteen (11%) patients did not undergo operation due to advanced disease or medical comorbidities. Of 105 (89%) patients that underwent exploration, 45 (43%) were found to have unresectable disease, while 60 (57%) had an attempt at a curative resection. Of those patients, 48 (80%) had R0 resections and the remaining 12 (20%) had involved microscopic margins or R1 resections.
Fig. 1

Flow diagram demonstrating the management of all patients with HCCA evaluated at MSKCC from 2001 to 2008

Clinicopathological parameters of patients submitted to resection are shown in Table 2. The mean age was 64 years in both R0 and R1 resection groups with an equal proportion of male patients. Mean tumor size was 2.9 ± 1.3 cm in patients with an R0 resection and 2.4 ± 1.2 cm in patients with an R1 resection and was not statistically different. The number of well-differentiated and papillary tumors were the same in both groups (12 vs. 17%) and (23 vs. 25%), respectively. There were also no significant differences in the rate of pre-operative biliary stenting (56 vs. 75%), hepatectomy (90 vs. 83%), or caudate resections (46 vs. 42%) between R0 and R1 groups. In addition, the estimated blood loss, transfusion rate and length of stay were equivalent. Metastasis to regional lymph nodes was noted in 23% of R0 and 33% of R1 resected patients, which was not statistically different.
Table 2

Resected patients 2001–2008


R0 Resection, N = 48

R1 resection, N = 12

Age (mean)

64 years

64 years

Gender (% male)

28 (58)

8 (67)

Tumor size (mean)

2.9 ± 1.3 cm

2.4 ± 1.2 cm

Pre-op biliary stent (%)

29 (56)

9 (75)

Well-differentiated (%)

6 (12)

2 (17)

Papillary (%)

12 (23)

3 (25)

Hepatectomy (%)

47 (90)

10 (83)

Caudate resection (%)

24 (46)

5 (42)

EBL (median)

800 cc

550 cc

Patients with transfusion (%) (mean)

20 (42)

5 (42)

LOS (median)



Positive lymph nodes (%)

12 (23)

4 (33)

Complications (%)a

15 (28)

6 (50)




Cardiovascular (PE, MI, AF)



Anastomotic leak/biloma



Hepatic failure



Mortality (%)

3 (6)


Survival (median)

74.3 months*

24.0 months*

* p < 0.008

aPatients may have more than 1 complication

There was a trend toward increased complications in the R1 group (50 vs. 28%) compared to the R0 group. There were three perioperative deaths in the R0 group, all of which were due to liver failure, on POD3, POD36 and POD50. There were no deaths in the R1 group. Disease-specific survival was significantly longer in the R0 resection group when compared to the unresected group (74.3 months vs. 13.1 months, p < 0.0001) as well as the R1 resection group (74.3 months vs. 24.0 months, p < 0.008). In addition, the R1 resection group had a significantly longer disease-specific survival than the unresected group (24.0 months vs. 13.1 months, p < 0.026) (see Fig. 2).
Fig. 2

Disease-specific survival (DSS) of all patients with HCCA stratified by resection status (R0, R1, unresectable). Median DSS was 74.3 months for R0 (n = 48), 24.0 months for R1 (n = 12), and 13.1 months for unresectable patients (n = 45). Log rank comparison of DSS R0 versus R1 was p < 0.008, and R1 versus unresectable was p < 0.026

All 118 patients evaluated from 2001-2008 were stratified according to the Blumgart preoperative staging system, previously described [10, 20] and detailed in Table 3. Forty-eight patients had primary tumor involvement of the biliary confluence but without unilateral extension into second-order biliary radicles, portal vein involvement or lobar atrophy and therefore classified as T1 tumors. Forty-one patients had T2 tumors due to ipsilateral lobar atrophy and/or portal vein involvement (n = 31 for both). There were 29 patients with T3 tumors, 10 due to main portal vein involvement, 7 due to extension to unilateral second-order biliary radicles and contralateral lobar atrophy, 10 due to extension to unilateral second-order biliary radicles and contralateral vascular involvement (portal vein in 9 and hepatic artery in 10), 2 due to extension to bilateral second-order biliary radicles. Using this system, resectability and the likelihood of an R0 resection decreased progressively with increasing stage (T1–T3) (odds ratio = 0.015 [0.002–0.119], p < 0.000) and (odds ratio = 0.028 [0.003–0.221], p < 0.001) respectively. Furthermore, the presence of metastatic disease precluding resection correlated with increasing T-stage (T1–T3) (odds ratio = 0.163 [0.057–0.443], p < 0.001).
Table 3

Resectability stratified by Blumgart preoperative staging system (2001–2008)



Operative exploration (%)

Resected (%)


Hepatic resection

Metastatic disease (%)

Median survival (months)



44 (92)

34 (71)



8 (17)




37 (90)

25 (61)



9 (22)




24 (83)

1 (3)



16 (55)




105 (89)

60 (51)



33 (28)


In a cohort of patients submitted to operation from 1991 to 2007, the impact of nodal metastases and the total lymph node count (TLNC) on outcome was evaluated. In the 144 patients included in this analysis, the median TLNC was 3 with a range of 0–16. On multivariate analysis, lymph node metastasis was an independent prognostic factor for disease-specific survival (median DSS: 46.6 ± 7.5 months vs. 22.8 ± 3.5 months, p < 0.008, HR 2.0 [CI 1.5–5.8]). Among patients who had an R0 resection with negative lymph nodes, DSS was higher in those with a greater TLNC. Maximal chi-square analysis suggested that an optimal lymph node harvest or TLNC for HCCA was 7. Among the 97 patients who underwent an R0 resection with en bloc hepatectomy, those classified as N0 based on a TLNC ≥7 (n = 13) experienced prolonged survival (85% 5-year, median not reached); however, when TLNC fell below this number, node-negative DSS was significantly worse (48% 5-year, median = 52 months, p = 0.05), although it was still greater compared to patients with node-positive tumors (18% 5-year, median = 25 months).

The role of preoperative biliary drainage of the future liver remnant (FLR) prior to extended hepatectomy was examined in 60 patients with adequate imaging available for retrospective volumetric analysis. All of these patients underwent ipsilateral en bloc hepatectomy for a potentially curative resection; only 1 patient underwent preoperative portal vein embolization. Preoperative biliary drainage was performed in 49/60 (83%) patients: 33 percutaneous, 14 endoscopic and 2 operative. Of these 49 patients, only 31 (63%) had adequate decompression of the FLR, while 18 had stents positioned that drained the ipsilateral liver (i.e, the portion of liver that was resected). Twenty-one patients had a predicted FLR volume of less than 30%, 9 underwent drainage of the FLR while 12 patients did not. In the latter group, absence of preoperative FLR biliary drainage was strongly associated with postoperative hepatic insufficiency (defined by a rise in bilirubin of greater than 5 mg/dl that persisted for longer than 5 postoperative days) and death (p = 0.009), 4 of these patients died in the perioperative period and 5 others developed postoperative hepatic dysfunction from which they eventually recovered. Of the 9 patients that underwent adequate drainage of the FLR, no patient died or developed evidence of hepatic insufficiency. By contrast, in the 39 patients with a FLR ≥30%, there were only 2 perioperative deaths and no cases of hepatic insufficiency. The two deaths occurred in patients who underwent FLR decompression, while in group of the 17 patients with a FLR volume ≥30% and no biliary drainage, none developed hepatic insufficiency and none died peri-operatively. It should be noted that patients with an FLR ≥30% were more likely to have radiographic evidence of ipsilateral lobar atrophy and hypertrophy of the FLR (46.2 vs. 9.5% in patients with FLR <30%, p = 0.004).


Hilar cholangiocarcinoma is a disease characterized by frequent locoregional invasion into porta hepatis structures that often precludes resection; however, contrary to popular belief, metastatic disease is frequently encountered. In the present study, approximately half of all patients had unresectable disease, either at presentation or exploration, and this was true also in the more recent period. While locally advanced disease was a major factor in this regard, metastatic disease was a more common finding. Accurate disease staging is clearly critical for identifying patients who would benefit from an operation and for avoiding a non-therapeutic laparotomy. Diagnostic laparoscopy and positron emission tomography (PET) may be helpful in this regard, although their impact on improving resectability rates are unclear [1417]. We have previously proposed a preoperative staging scheme [10, 20], which takes into account local factors that determine resectability, unlike the American Joint Commission on Cancer (AJCC) and Bismuth-Corlette systems [10, 18, 19]. Using this Blumgart classification, resectability, the probability of an R0 resection, and the likelihood of metastatic disease correlated with higher disease stage. The findings suggest that the incidence of non-therapeutic laparotomy could potentially be improved by targeting a more aggressive preoperative investigational approach to patients with more locally advanced tumors (i.e., higher Blumgart stage).

Once deemed potentially resectable, complete removal of the tumor generally requires extended hepatic resection, in addition to resection of the bile duct and porta hepatis lymphadenectomy en bloc, in order to achieve clear margins. In our prior reports, concomitant partial hepatic resection was shown to be an independent predictor of survival even among patients who had clear margins [10, 20]. In the more recent time period, a somewhat higher proportion of patients were subjected to partial hepatectomy (n = 67, 89%) and this appeared to be associated with some improvement in survival. In addition, these resections were associated with an operative mortality of 5%, which is lower than we have previously reported. This observation suggests an improvement in perioperative outcome and supports an aggressive resectional approach when technically feasible.

Historically, hepatic resection for HCCA has been associated with significant morbidity and mortality, the rates for which are notably higher than for other diseases [21, 22]. An important consideration in this regard is that patients with biliary obstruction are predisposed to post-operative liver failure, as well as other complications, following major hepatic resection. In an effort to improve outcome, preoperative biliary drainage, often combined with portal vein embolization, has been advocated as a means of improving the functional status of the FLR and reducing the rate of postoperative hepatic insufficiency. The potential advantages of this strategy do come at a cost, however, specifically the risk of increased postoperative infections related to bacterbilia, as well as possible procedural complications such as hemobilia, cholangitis, or neoplastic seeding of the tube tract. While some recent retrospective studies have suggested that preoperative biliary drainage with or without portal vein embolization may improve outcome, the contribution has been difficult to assess in the absence of prospective, randomized data [23, 24]. We have previously argued against the routine use of these procedures, given the potential complications, favoring instead selective use [25]. Indeed, our recent analysis would suggest that patients with a FLR volume ≥30% appear to derive little benefit from these procedures. This likely reflects the fact that, in many cases, hypertrophy of the FLR has already taken place, related to prolonged biliary obstruction and/or ipsilateral portal vein involvement, and suggests that the functional status of the FLR is sufficient to support a major hepatic resection. On the other hand, a FLR volume less than 30% had a remarkably high rate of hepatic insufficiency and death if the FLR was undrained [26]. These data appear to support a selective approach to the use of these preoperative procedures.

Metastasis to portal lymph nodes is a frequent occurrence in HCCA and serves as an important prognostic factor; our recent analysis showed a reduction in survival of over 50% with node-positive tumors [27]. Extended lymphadenectomy, to include nodal groups beyond the hepatoduodenal ligament, almost certainly does not improve survival, but the optimal lymph node dissection had not been defined. The AJCC classifies nodal stage for cholangiocarcinoma as either involved (N1) or uninvolved (N0). Unlike other gastrointestinal tumors such as gastric, pancreas and colon, there is no standard minimum number of lymph nodes that must be evaluated for accurate staging [28]. Based on our experience, patients who had an R0 resection for HCCA and found to be N0 by examination of at least 7 lymph nodes had a better disease-specific survival than those who fewer nodes staged. This observation is almost certainly due to understaging and presence of more advanced disease as opposed to a therapeutic effect of lymphadenectomy.

In summary, surgical management of HCCA remains a challenge. At MSKCC, the past several years have seen progressively greater use of en bloc partial hepatectomy with a slight decrease in operative mortality. Unresectable disease identified at laparotomy remains a problem. The Blumgart preoperative staging system may help identify patients at high risk for irresectable disease and therefore more likely to benefit from additional diagnostic studies, such as PET and staging laparoscopy. Although lymphadenectomy likely adds little to the therapeutic benefits of resection, an adequate lymph node sampling is necessary to avoid disease understaging. Selective use of preoperative biliary drainage, targeting patients with FLR <30%, appears to be a reasonable alternative to routine use of biliary drainage in all patients.

Copyright information

© Japanese Society of Hepato-Biliary-Pancreatic Surgery and Springer 2009