Introduction

Extrahepatic bile duct (EHBD) cancer is an aggressive disease with an overall 5-year survival rate of 30–42 % after surgical resection [1–4]. Surgical resection is the only possible curative treatment for patients with EHBD. On the basis of the location and extent of the tumor, appropriate surgical procedures including pancreaticoduodenectomy (PD), hepatectomy including caudate lobectomy with bile duct resection (HBR), or bile duct resection (BDR) are selected to achieve R0 resection. However, despite recent improvements in imaging techniques, it is difficult to evaluate the precise extent of longitudinal cancer spread along the bile duct. The reported incidence of a positive bile duct surgical resection margin in EHBD cancer patients was 18–35 % [2, 3, 5–7].

Several prognostic factors including the ductal margin status, depth of invasion, lymph node metastasis, perineural invasion, and differentiation grade have been reported in patients with EHBD cancer [1]. However, the survival benefit of a negative ductal margin of the bile duct (D-N) has been controversial [1–9]. Relationships between the ductal margin status and other prognostic factors are also unclear. Some patients may have postoperative survival dependent on tumor biology or stage of cancer progression rather than surgical margin of the bile duct. On the other hand, other patients may get benefit from achievement of D-N. It may be useful to clarify the patient population who benefits from D-N in clinical practice before deciding additional resection of the bile duct. The purpose of this retrospective study was to identify prognostic factors and to evaluate the prognostic impact of the bile duct surgical margin status in relationship with other prognostic factors in EHBD cancer patients.

Materials and methods

Patient enrollment

Patients with histologically proven EHBD cancer who underwent surgical resection at Kagoshima University Hospital between February 2002 and September 2014 were enrolled in this study, which was approved by the institutional ethics review board of Kagoshima University Hospital. Written informed consent was obtained for data analysis of the association between the ductal margin of the bile duct and postoperative prognosis. Patients with ampullary or gallbladder cancer were excluded. According to the location and extent of the tumor, PD, HBR, or BDR was performed. In all patients, regional lymph nodes including those in the hepatoduodenal ligament, posterior pancreaticoduodenal nodes, and nodes along the common hepatic artery were removed.

Diagnosis and definition of the bile duct surgical margin

Histological findings were described according to the TNM classification system for malignant tumors published by the International Union Against Cancer (UICC), 7th edition [10]. The proximal (hepatic side) and/or distal (duodenal side) bile duct margins were submitted for intra-operative frozen-section examination. If possible, additional resection was performed when the ductal margin was histologically positive. After surgery, the ductal margins were reevaluated histologically using formalin-fixed and paraffin-embedded resected specimens and tissues originally taken for intra-operative frozen diagnosis. The ductal margin status was classified into the following three categories: D-N, positive with carcinoma in situ (D-CIS), and positive with invasive carcinoma (D-INV). In the present study, high-grade dysplasia was included in the CIS category because it was difficult to distinguish between these epithelial lesions [2].

Follow-up after surgery

After discharge, all the patients were followed up by using physical examinations and laboratory tests that included carcinoembryonic antigen and carbohydrate antigen 19–9 (CA 19–9) levels every month for the first 2 years and every 3 months for the next 3 years. In addition, imaging examinations including computed tomography (CT) and magnetic resonance imaging (MRI) were performed every 3 months for the first 2 years and every 6 months for the next 3 years. The sites of recurrence were determined from the findings of CT or MRI. They were classified as hematogenous, local, lymph node, or peritoneal recurrence.

Statistical analysis

Disease-specific survival (DSS) and recurrence-free survival (RFS) curves were plotted using the Kaplan–Meier method and analyzed using the log-rank test. The DSS and RFS were analyzed from the date of surgical resection to the date of death from all causes and recurrence, respectively. The Cox proportional hazards model was used for the multivariate analysis of predictive variables found to be significant on the univariate analysis. Associations between different categorical variables were assessed using the Fisher exact test. P values <0.05 were considered statistically significant. Statistical evaluation was performed using SigmaPlot version 12.5 for Windows (HULINKS Inc., Tokyo, Japan).

Results

Patient characteristics

This retrospective study included 100 patients. PD, HBR, and BDR were performed in 69, 11, and 20 patients, respectively. In the present study, hepatectomy with concomitant pancreatoduodenectomy was performed in only 1 patient with perihilar cholangiocarcinoma, and the case was included in HBR group. Portal vein resection was performed in 3 patients (1 of BDR and 2 of PD), and there was no case with common hepatic artery resection. With regard to the bile duct surgical margin, 69 (69.0 %), 16 (16.0 %), and 15 patients (15.0 %) had D-N, D-CIS, and D-INV, respectively. Four of 15 patients with D-INV showed microscopically positive radial margins. Operative mortality was 1 % and that occurred in HBR group. And 16 patients (16.0 %) developed postoperative complications of Clavien–Dindo grade [11] 3 or more [1 of PD (15.9 %), 2 of HBR (12.5 %), and 3 of BDR (20.0 %)]. The most frequent complication was an intra-abdominal abscess (4 cases) followed by biliary leakage (3 cases) and pancreatic fistula (3 cases). Sixty of the 100 patients received either adjuvant chemotherapy or radiation after surgical resection. Radiation was mainly performed for patients with D-INV. The median follow-up period was 25.3 months (mean, 33.2 months).

Intra-operative frozen-section examination and additional resection of the bile duct

Intra-operative frozen-section examinations were performed in 93 patients (93.0 %). Results of the intra-operative frozen-section examinations were same with the final diagnoses with formalin-fixed specimens in 84 patients (90.3 %). Additional resections of the bile duct to achieve curative resection were performed in 29 of 46 patients who were diagnosed as positive ductal margin by the frozen-section examinations, and curative resection was achieved in 15 patients. On the other hand, additional resection could not be performed in 17 patients.

Association between the ductal margin status and clinicopathological factors

In univariate analysis, tumor location and surgical procedure were significantly associated with the ductal margin status of the bile duct (Table 1). In the perihilar group, 20 (57.1 %), 5 (14.3 %), and 10 (28.6 %) patients had D-N, D-CIS, and D-INV, respectively. By contrast, in the distal group, 49 (75.4 %), 11 (16.9 %), and 5 (7.7 %) patients had D-N, D-CIS, and D-INV, respectively. In the HBR group, 9 (81.8 %), 1 (9.1 %), and 1 (9.1 %) patients had D-N, D-CIS, and D-INV, respectively. In the PD group, 53 (76.8 %), 10 (14.5 %), and 6 (8.7 %) patients had D-N, D-CIS, and D-INV, respectively. By contrast, in the BDR group, 7 (35.0 %), 5 (25.0 %), and 8 (40.0 %) patients had D-N, D-CIS, and D-INV, respectively. Adjuvant therapy was performed for 11 (68.8 %) and 14 (93.3 %) patients with D-CIS and D-INV, respectively.

Table 1 Association between clinicopathological factors and margin status of the bile duct
Full size table

Prognostic significance of the ductal margin status

We analyzed postoperative survival according to the ductal margin status (Fig. 1). The median RFS differed significantly between patients with D-N, D-CIS, and D-INV, at 53.0, 32.0, and 15.5 months, respectively (Fig. 1a). There was a significant difference in RFS between each group. The median DSS of patients with D-N, D-CIS, and D-INV was 55.0, 53.0, and 24.0 months, respectively (Fig. 1b). A significant difference was noted only in the DSS between patients with D-N and D-INV. We analyzed prognostic significance of additional resection of the bile duct to achieve curative resection in 46 patients who were diagnosed as positive ductal margin by the frozen-section examinations. The median RFS in 29 patients who underwent the additional resection (32.0 months) was significantly longer than that in 17 patients who did not undergo the additional resection (13.0 months) (P = 0.002). On the other hand, the median DSS in the 29 patients (48.0 months) and the 17 patients (24.0 months) were not significantly different (P = 0.126).

Fig. 1
figure 1

Kaplan-Meier survival curves of extrahepatic bile duct cancer patients based on the surgical margin status of the bile duct. a There were significant differences in recurrence-free survival (RFS) among patients with a negative ductal margin (D-N), and carcinoma in situ at the ductal margin (D-CIS), and invasive carcinoma at the ductal margin (D-INV). b A significant difference in disease-specific survival (DSS) was observed only between patients with D-N and D-INV (P = 0.008)

Full size image

Association between the ductal margin status and recurrence site

There was no significant difference in recurrence sites according to the bile duct surgical margin status, with the exception of local recurrence (Supplementary Table 1). In patients with D-CIS and D-INV, 9 (56.3 %) and 10 (66.7 %) patients experienced local recurrence, respectively. By contrast, only 7 patients (10.1 %) with D-N experienced local recurrence. The distribution of the recurrence sites was similar between patients with D-CIS and D-INV.

Clinicopathological predictors of postoperative prognosis

As patients with D-INV showed significantly shorter DSS, we divided the patients into patients with D-N/CIS and patients with D-INV for further analysis of DSS. On univariate analysis, T3/4, N1, histological grade 2/3, and D-INV were significant predictors of shorter DSS (Table 2). In addition, we analyzed the relationship between the number of metastatic lymph nodes and prognosis. The median number of dissected lymph nodes per patient was 18.0. The median DSS was significantly shorter in patients with the number of metastatic lymph nodes ≥3 (28.5 months) compared to that in patients with the number of metastatic lymph nodes <2 (54.0 months) (P = 0.001). On multivariate analysis, N1 and D-INV were independent predictors of shorter DSS (Table 3).

Table 2 Association between clinicopathological factors and disease-specific survival
Full size table
Table 3 Cox regression analysis of disease-specific survival in patient with extrahepatic bile duct cancer
Full size table

Prognostic significance of the ductal margin status according to lymph node metastasis

To determine the impact of lymph node metastasis on DSS, we analyzed the prognostic significance of the ductal margin status based on the lymph node status (Fig. 2). In patients with lymph node metastasis, the survival curves of patients with D-N, D-CIS, and D-INV were similar, without any significant difference (Fig. 2a). The median DSS in patients with D-N, CIS, and D-INV was 37.5, 23.0, and 24.0 months, respectively. By contrast, in patients without lymph node metastasis, D-INV was a significant predictor of a shorter DSS (Fig. 2b). The median DSS in patients with D-N, CIS, and D-INV was 120.0, 55.0, and 22.0 months, respectively.

Fig. 2
figure 2

Kaplan-Meier survival curves of extrahepatic bile duct cancer patients based on lymph node status. a There was no significant difference in disease-specific survival (DSS) according to the ductal margin status in patients with lymph node metastasis. b In patients without lymph node metastasis, the DSS of patients with D-INV was significantly shorter than that of patients with D-N (P = 0.006)

Full size image

Discussion

Surgical resection has been the mainstay of curative treatment for EHBD cancer. However, most patients who undergo surgical resection experience postoperative recurrence [1–4]. We investigated the significance of the surgical margin status of the bile duct on recurrence and prognosis. In the present study, the rates of D-CIS and D-INV were 16.0 and 15.0 %, respectively. Patients who underwent BDR showed significantly higher incidence of D-CIS and D-INV compared to patients who underwent PD or HBR. Patients with D-INV/CIS showed significantly shorter RFS compared to those with D-N. On the other hand, only patients with D-INV showed significantly shorter DSS compared to those with D-N, whereas no significant difference was noted in DSS between patients with D-CIS and those with D-N. Several previous studies had reported that there was no significant difference in the prognosis of patients with D-N and those with D-CIS [2, 3, 6, 7]. In the present study, D-INV and lymph node metastasis were independent predictors of shorter DSS on multivariate analysis.

The rates of local recurrence in patients with D-CIS and D-INV were 56.3 and 66.7 %, respectively. These rates were significantly higher compared to those in patients with D-N (10.1 %). Higuchi et al. previously reported that the rates of local recurrence in patients with D-N, D-CIS, and D-INV were 18, 31, and 53 %, respectively [12]. While Nakanishi et al. reported that initial recurrence most frequently occurred at a local site in patients with D-CIS (60 %) and D-INV (44 %), it most commonly occurred in the liver of patients with D-N (52 %) [6]. These results may indicate that D-CIS has the potential to become a massive tumor, although this may take several years. The tumor type displaying extensive superficial spread that tends to be D-CIS was reported to be less malignant, with shallower invasion, better histological differentiation, and a more localized gross appearance compared to conventional bile duct cancer [13]. There was no significant difference in DSS between patients with D-N and those with D-CIS. However, patients with D-CIS showed significantly shorter RFS and their DSS decreased after 50 months compared to patients with D-N. In the patient subpopulation without lymph node metastasis, decrease of DSS after 50 months in patients with D-CIS was more remarkable. Therefore, it may be preferable to avoid not only D-INV but also D-CIS in order to achieve actual cure. Jang et al. emphasized the necessity of long-term follow-up to ensure that the patients were cured because late recurrence after 5 years is detected frequently [1].

Because the exact diagnosis of carcinoma spread is difficult, diagnosis using frozen sections is necessary to ensure a disease-free surgical margin of the bile duct [14]. If the intra-operative frozen section examination revealed a cancer-positive ductal margin, additional resection is performed to achieve D-N. However, additional resection of the bile duct can be challenging for patients with a poor general condition or with a high operative risk. Oguro et al. reported that additional resection of the bile duct to achieve curative resection did not always improve the prognosis of patients with perihilar cholangiocarcinoma, although it did prolong the postoperative survival of patients with a low serum CA 19–9 level and without distant metastasis [15]. Our study suggests that additional resection of the bile duct to achieve curative resection may improve the prognosis of EHBD cancer patients and should be considered, particularly for patients without lymph node metastasis. Lymph node metastasis is a potent predictor of shorter postoperative survival in patients with bile duct cancer [4, 16, 17].

This study had several limitations. It had a retrospective design, and there was no standard strategy for postoperative therapy. Adjuvant therapy was mainly performed for patients with positive ductal margin. Therefore, whether adjuvant therapy improve the prognosis of all patients with EHBD is unclear. Next, the analyses of prognosis according to lymph node status have small number of patients with D-CIS and D-INV. To confirm the results of the present study, we are planning further study including more number of patients. Furthermore, the preoperative diagnosis of lymph node metastasis in patients with bile duct cancer is difficult [18, 19]. However, additional resection of the bile duct in patients with a poor general condition or with a high operative risk should be avoided if they have obvious lymph node metastasis.

Conclusion

Our findings indicate that the surgical margin status of the bile duct is a significant prognostic factor in EHBD cancer patients. D-CIS and D-INV were significant predictors of shorter RFS. D-INV was an independent predictor of shorter DSS, and its relevance to DSS was observed, particularly in patients without lymph node metastasis.