Introduction

Intrahepatic cholangiocarcinoma (ICC), a primary cancer of the bile ducts, is relatively uncommon, accounting for 4.4 % of all primary neoplasms of the liver in Japan [1]. Worldwide, cholangiocarcinoma accounts for 3 % of all gastrointestinal cancers and is the second most commonly occurring primary hepatic tumor [2]. Surgical resection is the only potentially curative treatment, however, the prognosis of advanced ICC, even if it is resectable, remains unsatisfactory [3–8]. Chemotherapy has improved the prognosis of patients with advanced or borderline-resectable ICC, as well as inoperably advanced ICC, however, curative resection is an essential prerequisite for long-term survival [8–13].

Numerous previous studies investigating the results of surgical resection for ICC have revealed several factors influencing the prognosis or risk of ICC recurrence, such as the macroscopic class, lymph node metastasis, multiple intrahepatic tumors, hilar invasion, positive/negative surgical margin, the histological differentiation grade, surgical curability according to the Japanese classification [1], serum carbohydrate antigen 19-9 (CA19-9) level, etc. [4, 5, 10, 14–40]. However, many of these risk factors require histological examination after surgical resection. Preoperative diagnosis of lymph node metastasis is quite difficult [41–43], and not particularly useful for deciding the operative indication without histological confirmation. On the other hand, with the development of minimally invasive surgery, laparoscopic liver resection has been applied for small peripheral ICCs [25], although evidence to justify not dissecting the regional lymph nodes in such patients is still lacking.

To clarify the appropriate indications for surgical resection in patients with ICC, it is necessary to consider whether the risk factors can be assessed preoperatively or need to be confirmed by histology. It is also important to know the risk of ICC recurrence after surgery, because high-risk patients for ICC recurrence may benefit from chemotherapy. There has been no study focusing on these issues so far, and appropriate treatment strategies for ICC have not yet been established to date.

In this study, we attempted to identify predictors of the overall survival rates and ICC recurrence that can be determined preoperatively and postoperatively in patients with advanced ICC, as well as the preoperative predictors of lymph node metastasis. With this objective, we retrospectively investigated the clinicopathological features of ICC and the outcomes of surgical resection in patients who underwent surgical treatment for ICC at our hospital between January 1980 and March 2012.

Patients and Methods

Aggressive surgical resection has remained the principle of treatment of ICC at our hospital, as long as there is no extrahepatic metastasis, no apparent paraaortic lymph node metastasis, or peritoneal dissemination. Vessel involvement is not a contraindication if it is reconstructable. If the tumor is found to invade the liver hilum, the common bile duct (CBD) is resected and hepatico-jejunostomy is performed. Lymphadenectomy of the hepatoduodenal ligament (lymph node Nos. 8, 9, 12, 13a according to the Japanese classification [1, 44, 45]) and of the paraaortic lymph node (lymph node No. 16 [1, 44, 45]) is routinely performed in cases where ICC is suspected preoperatively.

Postoperative adjuvant chemotherapy is not routinely administered for patients treated by curative resection, while intra-arterial infusion of 5FU or local radiation therapy is undertake for patients with non-curative or no resection.

Preoperative computed tomography (CT) is performed to evaluate the location, size and number of the tumors and any lymph node swelling. Metastatic lymph nodes are defined as round, bulky lymph nodes measuring more than 10 mm in diameter on CT, as diagnosed by radiologists. The level of involvement in cases of ICC is divided into 3 levels: peripheral, intermediate and hilar. Hilar level: tumor involvement of the left or right branch of the glissonean pedicle; intermediate level: tumor involvement of the second branch of the glissonean pedicle (anterior, posterior branch in the right lobe or left medial or lateral segment branch in the left lobe); peripheral level: more peripheral tumor involvement than the second branch of the glissonean pedicle. Preoperative evaluation of the anatomical level of invasion was performed on CT images by radiologists, while postoperative histological confirmation of the anatomical level of invasion is performed in resected specimen by pathologists.

Other clinical factors, including histological diagnosis of lymph node metastasis, tumor differentiation grade, macroscopic class of the tumor, growth pattern (expansive growth or invasive growth), and preoperative laboratory data were also evaluated. The overall survival was investigated in patients who underwent laparotomy for curative resection (n = 111), including non-curative resection (RR group, n = 17), and simple laparotomy (NR group, n = 22), while the disease-free survival rate was investigated in patients in whom curative resection had been achieved (CR group, n = 72).

Continuous data were expressed as mean ± SD and group data sets were compared using the Mann–Whitney U test or Kruskal–Wallis test. Categorical data were presented as percentages, and differences between proportions were compared using the χ2 test. Univariate analysis of the rates of overall survival, ICC recurrence, lymph node metastasis was performed using Kaplan–Meier analysis (log rank test), logistic regression analysis, or COX proportional hazard model analysis. Variables that were identified to be correlated with the results in the univariate analysis (P < 0.1) and had no correlation with other factors were included in the multivariate logistic regression model or COX proportional hazard model analyses. For all tests, P values of <0.05 were considered to be significant. All analyses were performed using SPSS, version 14.0 (SPSS, Inc., Chicago, IL, USA).

Results

A total of 111 consecutive patients with ICC were operated at our hospital between January 1980 and March 2012. Curative resection (R0) was performed in 72 patients (CR group, 64.9 %), and resection with residual cancer (RR group, R1/R2) was performed in 17 patients (15.3 %). The reasons for non-curative resection in the 17 patients were positive bile duct margin with cancer invasion (n = 11), multiple intrahepatic tumors (n = 4), invasion of the IVC (n = 1), and others (n = 1). The remaining 22 patients (NR group, 19.8 %) underwent simple laparotomy without resection (R2) because of the advanced stage of the tumor with peritoneal dissemination (n = 12), para-aortic lymph node metastasis (n = 5), multiple intrahepatic tumors (n = 2), locally advanced tumor (n = 1), and others (n = 2).

Lymph node dissection was performed 87 patients (63 (87.5 %) in the curative resection group (CR group), 24 (61.5 %) in RR + NR group); 9 (12.5 %) of the 72 patients of CR group had not received lymph node dissection due to preoperative diagnosis of hepatocellular carcinoma (n = 5), old age (n = 1), or others (n = 3), and 15 (38.5 %) of the 39 patients of RR + NR group had not received lymph node dissection.

In 89 patients with ICC of the curative resection (R0) or non-curative resection (R1, R2) group, the CBD was resected and hepatico-jejunostomy was performed in 21 patients (23.6 %), while the CBD was not resected in 68 patients (76.4 %).

Survival Analyses in the all 111 Patients Who Underwent Surgery

The characteristics of the 111 patients who underwent surgery are shown in Table 1. There were no significant differences in the age, gender, percentage with HBV/HCV infection, asparate aminotransferase (AST), alanine aminotransferase (ALT), percentage with lymph node metastasis, intrahepatic tumor number, tumor size, or histological tumor differentiation grade between the groups, while the gamma-glutamyltransferase (GGT), prothrombin time (PT), CA19-9, invasion level of the tumor, percentage with perineural and vascular invasion, and pTNM stage were different and more advanced in the RR group or NR group than in the CR group (Table 1). The 1-, 3- and 5-year overall survival rates were 65.8, 41.8 and 40.5 % in the 111 patients. According to the surgical curability, the 1-, 3- and 5-year overall survival rates were 85.0, 59.7 and 59.7 % in the CR group (n = 72), 43.8, 21.9 and 0 % in the RR group (n = 22), and 20.0, 0 and 0 % in the NR group (n = 17) (Fig. 1). Univariate risk factor analysis showed that tumor size >5 cm (P = 0.0004), multiple intrahepatic tumors (P < 0.0001), hilar invasion on CT (P = 0.004), final histological diagnosis (P = 0.0008), pathological lymph node metastasis (P < 0.0001), tumor differentiation grade (P = 0.0497) and curability of surgical resection (P < 0.0001) were significantly associated with the cumulative overall survival (Table 2). As for the preoperative laboratory data, AST (P = 0.018), ALT (P = 0.006), GGT (P = 0.010) and CA19-9 (P = 0.0001) were associated with the patients’ survival. On the other hand, lymph node metastasis diagnosed by CT, not conducting lymph node dissection, and expansive/invasive growth pattern were not associated with the overall survival (Table 2). To identify independent predictors of the overall survival, the risk factors identified by univariate analysis were entered into the multivariate analysis model. Hilar invasion on CT (P = 0.020, HR 3.67), multiple intrahepatic tumors (P = 0.0002, HR 7.09), and serum ALT (P = 0.024, HR 1.015) were identified as independent predictors, which could be evaluated preoperatively, of the overall survival, whereas multivariate analysis using the factors, including postoperatively confirmed factors, identified hilar invasion confirmed by histology (P = 0.020, HR 3.16), multiple intrahepatic tumors (P = 0.0009, HR 9.17) and lymph node metastasis by final histology (P = 0.003, HR 6.41) were independent risk factors associated with the patients’ survival (Table 2).

Table 1 Clinical backgrounds of patients in CR, RR, and NR groups
Full size table
Fig. 1
figure 1

Cumulative patient survival curve after surgery for intrahepatic cholangiocarcinoma (ICC). The 1-, 3- and 5-year overall year survival rates were 85.0, 59.7, 59.7 % in the patients with curative resection (CR group, n = 72, black line), 43.8, 21.9, and 0 % in the patients with non-curative resection (RR group, n = 22, dotted line), and 20.0, 0, and 0 % in the patients without resection (NR group, n = 17, gray line)

Full size image
Table 2 Uni- and multi-variate riskfactor analysis for overall survival (n = 111)
Full size table

Comparison Between Curatively and Non-curatively Resected Patients

Comparison of the preoperative and operatively confirmed clinical data between the CR group (n = 72) and RR + NR group (n = 39) by logistic regression analysis showed that the presence of hilar invasion on CT (P = 0.002) or on histology (P = 0.005) and presence of multiple intrahepatic tumors (P = 0.005) were significant risk factors for non-curative resection or no resection (RR + NR group), while tumor size >5 cm, lymph node metastasis diagnosed by CT, macroscopic class of the tumor, and histological differentiation grade were not. Multivariate logistic regression analysis using factors that could be evaluated by preoperative imaging studies revealed that the presence of hilar invasion on CT (P = 0.008) and presence of multiple intrahepatic tumors (P = 0.005) were independent risk factors for non-curative resection or no resection (RR + NR group) (Table 3).

Table 3 Uni- and multivariate risk factor analysis for the CR group (n = 72) versus RR + NR group (n = 39)
Full size table

Subclass analyses were next undertaken between all the patients with non-curative or no resection and the patients with certain risk factors (lymph node metastasis diagnosed histologically, tumor size >5 cm, and multiple intrahepatic tumors) among the patients with curative resection. These showed that the patients with positive lymph node metastasis (P = 0.592) or multiple intrahepatic tumors who had undergone curative resection (P = 0.054) had the same prognosis as the patients of the non-curative or no resection group, while the patients with ICC size >5 cm (P < 0.0001) or histological invasion of the liver hilum (P = 0.015) who had undergone curative resection showed better survival than the corresponding patients of the non-curative resection or no resection group (Fig. 2).

Fig. 2
figure 2

Cumulative patient survival curves after surgery for intrahepatic cholangiocarcinoma (ICC) according to the risk factors and resectability. a Lymph node metastasis and resectability. There was no significant difference of the patient survival rate between the CR group patients with lymph node metastasis (+) (n = 18) and the RR + NR group patients (n = 38) (P = 0.592, log rank test). b Tumor size and resectability. There was a significant difference between the CR group patients with size >5 cm (n = 28) and RR + NR group patients (n = 38) (P < 0.0001, log rank test). c Multiple intrahepatic tumors and resectability. There was no significant difference between the CR group patients with solitary tumor (n = 19) and RR + NR group (n = 38) patients (P = 0.054, log rank test). d Histological invasion of the liver hilum and resectability. There was a significant difference between the CR group patients with histological invasion to liver hilum (+) (n = 14) and RR + NR group (n = 38) patients (P = 0.015, log rank test)

Full size image

Preoperative Predictors of Lymph Node Metastasis

Lymph node metastasis was found as one of the strongest independent predictors of a poor prognosis. Preoperative diagnosis of lymph node metastasis on CT showed only 22.2 % sensitivity, 73.3 % specificity and 61.5 % accuracy. Clinical factors which were associated with lymph node metastasis were the tumor size (P = 0.003), hilar invasion level diagnosed by CT (P = 0.002), and multiple versus solitary intrahepatic tumors (P = 0.001), while lymph node metastasis as evaluated by preoperative CT was not associated with histological lymph node metastasis. Multivariate logistic regression analysis using factors which could be evaluated by preoperative imaging studies revealed hilar invasion on CT (P = 0.010) and tumor size (P = 0.043) as independent risk factors for histological lymph node metastasis (Table 4).

Table 4 Uni- and multi-variate risk factor analysis for histological lymph node metastasis (n = 111)
Full size table

The Incidence of Lymph Node Metastasis in Small and Peripheral ICC

To clarify the incidence of lymph node metastasis in association with solitary, small and peripheral ICCs, the patients with solitary ICC who underwent surgical resection with lymph node dissection (n = 46) were divided by 2 factors; tumor size (5 cm) and hilar invasion level diagnosed by CT scan (Table 5). Although the tumor size >5 cm and hilar invasion were risk factors for the presence of lymph node metastasis, hilar invasion showed the stronger correlation. Among the patients with solitary ICC, patients with hilar invasion had a 72.7 % incidence of lymph node metastasis, while none of the patients with tumor size <5 cm and peripherally located tumors showed lymph node metastasis (Table 5).

Table 5 Incidence of histological lymph node metastasis in the patients with solitary ICC (n = 46)
Full size table

ICC Recurrence Analyses in the Patients with Curative Resection (n = 72)

The 1-, 3- and 5-year disease-free survival rates in the 72 patients with ICC who underwent curative resection were 72.7, 53.9 and 51.7 %, respectively. ICC recurrence occurred in 32 patients (44.4 %) over a mean follow-up period of 3.35 years (0.03–18.2 years) (Fig. 3). The sites of recurrence of ICC in these 32 patients were the liver (n = 23, 71.9 %), regional lymph node (n = 7, 30.4 %), bone (n = 2, 8.7 %), peritoneum (n = 2, 8.7 %), lung (n = 1, 4.3 %), and others (n = 2, 8.7 %), including concurrent recurrence in the liver and lymph node (n = 3, 9.4 %), bone and lymph node (n = 1, 4.3 %), liver and bone (n = 1, 4.3 %).

Fig. 3
figure 3

Disease-free survival rates in the CR group (n = 72). a The disease-free survival rates at 1, 3 and 5 years after surgery were 72.7, 53.9, 51.7 %, respectively. b Recurrence-free survival rates according to the presence/absence of lymph node metastasis. There was a significant difference in the recurrence-free survival rate after curative surgery for ICC between patients who were lymph node metastasis (+) and (−) (P < 0.0001). c Recurrence-free survival according to the number of tumors. There was a significant difference in the recurrence-free survival rate after curative surgery for ICC between patients with solitary and multiple intrahepatic tumors (P < 0.0001)

Full size image

Univariate cox proportional hazard model analysis showed that multiple intrahepatic tumors (P < 0.0001), lymph node metastasis confirmed by histology (P < 0.0001), tumor size >5 cm (P = 0.034), hilar invasion confirmed by histology (P = 0.024), curability (P = 0.0019) and preoperative serum CA19-9 level (P = 0002) were significantly associated with the risk of ICC recurrence, while hilar invasion preoperatively diagnosed on CT, lymph node metastasis preoperatively diagnosed by CT scan, not conducting lymph node dissection, and preoperative serum total bilirubin, GGT, AST, ALT and carcinoembryonic antigen (CEA) were not associated with the risk of ICC recurrence.

To identify the independent predictors of recurrence-free survival, factors identified by univariate analysis were entered into the multivariate analysis model. Multiple intrahepatic tumors (HR 4.59, P = 0.003) and elevated preoperative CA19-9 level (HR 1.00, P = 0.038), which could be evaluated preoperatively, were identified as independent predictors of ICC recurrence, whereas multivariate analysis using the factors including postoperatively confirmed factors identified the presence of multiple intrahepatic tumors (HR 8.20, P = 0.002) and lymph node metastasis confirmed by final histology (HR 15.9, P < 0.0001) as the independent risk factors for ICC recurrence (Table 6). Recurrence-free survival analyses showed that these two risk factors allowed very good stratification of the prognosis (Fig. 3).

Table 6 Uni- and multi-variate riskfactor analysis for recurrence free survival in CR group (n = 72)
Full size table

Discussion

ICC is a rare primary malignant tumor of the liver, and its clinicopathological characteristics are still not clearly understood. We retrospectively analyzed the data of 111 patients who had undergone surgery for ICC at a single cancer center over a period of more than 30 years. During this long time period, while our principle of an aggressive treatment strategy has not changed, new chemotherapeutic agents such as gemcitabine and S-1 have become available, and many studies have reported improvement of the prognosis following addition of chemotherapy in ICC patients [8–13]. However, curative resection is still one of the essential prerequisites for obtaining long-term survival in patients with ICC.

Previous studies have identified numerous risk factors, such as the presence of lymph node metastasis, multiple tumors and/or lymphatic invasion as being associated with a high risk of tumor recurrence and/or poor survival in patients with ICC [4, 5, 10, 14–39]. Other studies have pointed out the location of the tumor as having a significant influence on the risk of ICC recurrence. Tumors invading the liver hilum are associated with a poorer prognosis and a high incidence of lymph node metastasis [14, 23, 25, 29], while some studies have denied any influence of the presence/absence of hilar invasion on the prognosis [46].

Lymph node metastasis as confirmed by histology has been reported as one of the strongest predictors of the prognosis in patients with ICC [7, 15, 18, 19, 22–24, 28, 29, 34, 36, 37, 39]. However, preoperative evaluation of the status of lymph node metastasis by CT is quite difficult because of its low accuracy, consistent with the finding of this study as well as other studies [41–43]. Recently, positron emission tomography–computed tomography (PET–CT) was reported to be better method for detecting lymph node metastasis, although its accuracy has not yet been confirmed to be satisfactory [42, 43, 47]. Therefore, as of the present, lymph node metastasis can be evaluated accurately only by histology, which is possible by fine needle biopsy preoperatively, but is more often diagnosed intra- or post-operatively. Therefore, other preoperative predictors of postoperative survival than lymph node metastasis are very important in considering the indication for surgical treatment.

In this background, the significance of this study was as follows. First, we considered that the extent of invasion of ICC may be a key factor in predicting the prognosis in patients with ICC, and used the new classification of the extent of invasion in cases of ICC based on both preoperative imaging and histology. Secondly, our study is one of the very few [40] to investigate the risk factors separately as preoperative and postoperatively confirmed factors, which is considered to be quite useful to clarify the appropriate operative indications for surgery both in patients with advanced as well as early ICC.

Analysis of the overall survival rate in the patients with ICC who had undergone elective surgery at our institution confirmed three preoperative factors (presence of hilar invasion on CT, presence of multiple intrahepatic tumors and serum ALT) and three postoperatively confirmed factors (presence of hilar invasion by final histology, multiple intrahepatic tumors, and lymph node metastasis as confirmed by final histology) as independent predictors of a poor overall survival rate, while two preoperative factors (hilar invasion on CT and multiple intrahepatic tumors) and two postoperatively confirmed factors (multiple intrahepatic tumors and lymph node metastasis confirmed by final histology) were identified as factors predictive of ICC recurrence after curative resection.

Comparison of the preoperative and operatively confirmed clinical data between the patients with curative resection (CR group, n = 72) and the patients with non-curative resection or no resection (RR + NR group, n = 39) by logistic regression analysis showed that hilar invasion and multiple intrahepatic tumors were independent risk factors of non-curative resection of ICC; this result could be interpreted as reflecting the fact that patients with hilar invasion or multiple intrahepatic ICC tumors have a high risk of peritoneal dissemination, lymph node metastasis or tumor growth along with surgical margin.

Analysis of the data of ICC patients with multiple intrahepatic tumors or lymph node metastasis confirmed by histology showed that the prognosis was similar between the group that underwent curative resection and the group that underwent non-curative resection or no resection, whereas that of the patients with hilar invasion showed that the survival was better in the group that had undergone curative resection than that in the group that had undergone non-curative resection or no resection. Therefore, surgical resection may not be indicated in ICC patients with multiple intrahepatic tumors and/or histologically confirmed lymph node metastasis because it appears to offer no survival benefit, whereas in ICC patients with hilar invasion, curative resection appears to be associated with an improved prognosis (Fig. 2). Therefore, it would be advisable to perform lymph node dissection first and intra-operatively evaluate the specimens histologically for metastasis before proceeding to hepatectomy, and in case the histological metastasis to the lymph node was confirmed, hepatectomy should not be performed.

Not the diagnosis by CT (P = 0.897), but the presence of hilar invasion (P = 0.010) and the tumor size (P = 0.043) on CT was identified as a preoperative predictor of lymph node metastasis. Thus, these two factors were strongly associated with the histological presence of lymph node metastases, and were quite useful as preoperative predictors.

On the other hand, the incidence of lymph node metastasis in the 19 patients with solitary, small (<5 cm) and peripheral ICC tumors was 0 %, suggesting that lymphadenectomy is not necessary for such cases, and provides the rationale for considering laparoscopic liver resection without lymphadenectomy for such cases in patients with solitary, small (<5 cm) and peripheral ICCs.

Most patients in our series did not receive perioperative chemotherapy, unless the surgery was non-curative or no resection, residual cancer was confirmed, or recurrent tumor was confirmed postoperatively; in such cases, we administered 5FU-based chemotherapy either by peripheral or intra-arterial injection. It is possible that the patient survival rates can be improved by using the newly introduced therapeutic agents, but the degree of such improvement would be similar among patients with non-curative resection, no resection, histological lymph node metastasis and/or multiple intrahepatic tumors if the newer agents were used similarly for these patients.

In conclusion, we identified that preoperatively diagnosed hilar invasion, multiple intrahepatic tumors and histologically confirmed lymph node metastasis were the main determinants of an adverse postoperative prognosis in patients with ICC. Preoperative evaluation of the lymph node metastatic status was quite difficult by image studies, while we also identified with hilar invasion on CT and tumor size as independent risk factors for histological lymph node metastasis. On the other hand, patients with solitary, small (<5 cm) and peripheral ICCs show a very low probability of lymph node metastasis, so that lymph node dissection could be omitted for these patients.