Improvement in Perioperative and Long-term Outcome After Surgical Treatment of Hilar Cholangiocarcinoma:  Results of an Italian Multicenter Analysis of 440 Patients FREE

Hilar cholangiocarcinoma (HC), or Klatskin tumor,¹ represents more than 50% of all biliary tract cholangiocarcinomas.² Radical surgical resection is the only treatment offering a chance of long-term survival³; it combines resection of main biliary confluence with partial hepatectomy including the caudate lobe.^4- 9 The central location of the tumor and its close relationship with vascular structures at the hepatic hilum have resulted in a low resectability rate and high morbidity and mortality.³ During the past 2 decades, the rate of resectability has increased owing to wider application of aggressive surgery with concomitant vascular resection and reconstruction, and early and long-term outcomes have considerably improved.^10- 11 Most articles about the changing trends in the surgical approach to HC report results from single institutions, with data on a few patients collected.^12- 13 Moreover, the clinicopathologic characteristics of patients with HC differ from those of patients with intrahepatic cholangiocarcinoma involving the hepatic hilum and implicate a different prognosis after surgical resection,¹⁴ whereas in many studies^7,15- 17 the 2 conditions were considered to be the same entity.

The aim of this Italian multicenter study was to evaluate improvements in operative results and long-term outcome following surgical resection for HC during a 16-year period in a large cohort of patients with the same proven tumor entity.

Data were collected from 17 hepatobiliary Italian centers, members of the Italian Chapter of the International Hepato-Pancreato-Biliary-Association (IHPBA). The study included patients who underwent resection from January 1, 1992, through December 31, 2007, for histologically proven HC.¹ Patients who underwent resection for intrahepatic cholangiocarcinoma involving the hepatic hilum were excluded. Centers that collected 40 or more cases during the established period were considered high-volume centers. Time trend analysis was performed to evaluate improvements along time in perioperative and long-term results after liver resection.

For each patient, the following data were collected: demographic information, clinical presentation, use and type of preoperative biliary drainage, use of preoperative portal vein embolization, and staging laparoscopy. Operative details included type of resection, use of pedicle clamping, and use of blood transfusions. Early and late results included postoperative mortality and morbidity, 5-year overall survival (OS) and disease-free survival (DFS), and recurrence rate. Pathologic data included histologic type (sclerosing or papillary), size of tumor, grading, presence of perineural invasion, radicality of resection (biliary margin status), invasion of biliary ducts of caudate lobe, and lymph node involvement. Pathologic tumor staging was based on the TNM classification of the International Union Against Cancer staging system (6th edition).¹⁸ Extent of bile duct involvement was typed by the Bismuth-Corlette classification.¹⁹ Liver resections were defined according to the terminology of the IHPBA.²⁰

Overall survival and DFS were calculated using the Kaplan-Meier method. Patients who postoperatively died were included in the OS analysis (intention-to-treat criteria) and excluded from the DFS analysis. Differences between survival rates were analyzed with the log-rank test.

The χ² test was used to compare proportions. Significance was defined as P < .05. The Cuzick nonparametric test for trend analysis across ordered groups²¹ (an extension of the Wilcoxon rank-sum test) was performed to assess whether surgical outcomes and patient characteristics varied over time. Statistical analysis was performed using SPSS software, version 13.0 (SPSS, Inc).

From January 1, 1992, through December 31, 2007, a total of 440 patients were treated with curative intent surgery for HC (Table 1). Five of the 17 centers (29.4%) reported 40 or more patients who underwent resection and accounted for 317 patients (72.0%).

Liver resection combined with main biliary confluence excision was performed in 376 patients (85.5%) (Table 2). Figure 1 shows a steady yearly increase in the number of liver resections; during the most recent 2-year period listed (2006-2007), 121 hepatectomies were performed (32.2% of all hepatectomies).

Preoperative ultrasonography was performed in all patients. Preoperative evaluation was performed by abdominal computed tomography and magnetic resonance cholangiopancreatography in 217 patients (49.3%), by computed tomography alone in 178 (40.5%), and by magnetic resonance cholangiopancreatography alone in 45 (10.2%). At imaging, a nodular lesion was documented in 218 patients (49.5%) and a focal stricture of main biliary confluence without evidence of neoplastic mass was documented in the remainder.

Staging laparoscopy was routinely performed in 4 centers (23.5%), in a total of 88 patients. This procedure detected peritoneal carcinomatosis in 1 of the 88 patients (1.1%).

Preoperative biliary drainage was performed in 294 patients (66.8%). Of these 294 patients, it was performed percutaneously in 161 (54.8%), endoscopically in 95 (32.3%), by both procedures in 18 (6.1%), and by unspecified procedure in 20 (6.8%).

The mean (range) number of stents placed by endoscopic retrograde cholangiopancreatography was 1.4 (1.0-4.0). The mean (range) number of percutaneously placed biliary drainages was 1.3 (1.0-3.0). In 80.4% of percutaneously treated patients, a transpapillary drainage was placed.

Preoperative cytology or histology of the neoplasm was sought in 33 of the 294 patients who underwent drainage (11.2%): by brushing for cytology in 19 and by biopsy for histology in 14. Malignancy was not confirmed in 18 patients: the overall sensitivity rate was 15 of 33 (45.5%), without significant differences for brushing (7 of 19 [36.8%]) vs biopsy (8 of 14 [57.1%]).

Right portal vein embolization was performed in 37 of 172 surgical procedures of right or right extended hepatectomies (21.5%). Liver resection followed after a mean (range) of 47 (30-90) days (median, 36 days).

Operations and results are shown in Table 2 and Table 3. Caudate lobectomy was performed in 293 of 376 liver resections (77.9%).

Of all 440 patients, lymphadenectomy of hepatoduodenal ligament was performed in 423 (96.1%). This procedure was extended to para-aortic lymph nodes in 125 patients (28.4%).

Of 376 patients who underwent liver resections, the postoperative mortality was 10.1% (38 patients). Postoperative morbidity after liver resection (47.6%) or after main biliary confluence resection alone (46.8%) was not significantly different. The rate of reinterventions was 6.4% for both.

Risk of mortality was assessed in a multivariate logistic regression model including age, sex, extent of biliary involvement, extent of hepatectomy, caudate lobectomy, vascular resection, pedicle clamping, and intraoperative blood transfusion. Intraoperative transfusion (P = .006) and extent of hepatectomy (right or right extended hepatectomy) (P = .03) emerged as independent predictors of mortality (Table 4). Table 5 shows differences in operative and pathologic factors between high- and low-volume centers.

The rate of liver resection was higher in high-volume than in low-volume centers: 90.9% (288 of 317) vs 71.5% (88 of 123) (P < .001). At multivariate logistic regression, blood transfusion and extent of hepatectomy were reconfirmed as predictors of postoperative mortality, whereas the factor high-volume center was not associated with significantly higher mortality (odds ratio, 1.66; P = .32), despite more aggressive surgical approaches.

Among the 376 patients who underwent liver resection, 299 presented with obstructive jaundice. Of these 299 patients, 252 (84.3%) underwent preoperative biliary drainage and 47 (15.7%) did not. Risk of mortality (assessed by direct standardization to adjust for blood transfusion and extent of resection) tended to be higher following hepatectomy without biliary drainage than following hepatectomy with biliary drainage (14.3% vs 10.7%; P = .41).

Final pathology showed nodular-sclerosing tumor in 94.3% of cases (Table 6). Data on neoplastic invasion of caudate lobe biliary ducts was available on 154 patients; invasion was found in 24.0% of them. R0 resections were performed in 340 patients (77.3%). There were 100 R1 resections; the cause was invasion of proximal bile duct margin in 89 patients (89.0%), of distal bile duct margin in 7 (7.0%), and of both margins in 4 (4.0%). The R0 resection rate was higher after associated liver resection than after biliary confluence resection alone (79.2% vs 65.6%; P = .01) (Table 6).

Figure 2 shows the changing trends during the study period, assessed by locally weighted scatterplot smoothing. Postoperative mortality and morbidity peaked before 1996, then decreased until 2000, and thereafter moderately increased. During the last 4 years of study, compared with the initial 4 years, mortality and morbidity tended to decrease, the blood transfusion rate significantly decreased, and rates of associated vascular and caudate lobe resections and of hepatectomies for advanced stage and poorly differentiated tumors significantly increased.

The 5-year OS after surgical resection was 25.5% and the 10-year OS was 19.9% (median survival, 25 months) (Figure 3). The impact of associated liver resection on improved long-term results was analyzed in a subgroup of patients operated on in the first period of study with follow-up of 8 years or longer: patients undergoing associated liver resection had a significantly higher 5-year OS afterward than did those after biliary confluence resection alone (26.6% vs 0.0%; P = .02), with a 10-year OS of 20.0% (Figure 4).

Five-year OS after R0-liver resection was not significantly different in high- vs low-volume centers (35.2% vs 35.8%; P = .50).

Time trend analysis showed a significant increase in median survival during the period of study (Figure 5). Median survival during the initial 4 years was lower than during the last 4 years (16 vs 30 months; P = .05).

Recurrence was noted in 54.5% of patients (219 of 402 patients, excluding postoperative deaths). Mean (range) time to recurrence was 17 (1-120) months (median, 12 months).

At univariate analysis (Table 7), 8 factors were associated with significantly lower OS: perineural invasion, R1 resection, lymph node metastases, stage III or greater, T stage of 3 or greater, grading of 2 or greater, age 60 years or older, and caudate lobe invasion. These same factors, excluding perineural invasion and age of 60 years or older, were predictive of DFS.

At multivariate analysis (Table 8), R1 resection, lymph node metastases, T stage of 3 or greater, and perineural invasion were independent predictors of poor OS; except for perineural invasion, these same factors independently predicted poor DFS. Volume of the center was not an independent predictor of OS (P = .11) or DFS (P = .72).

Vascular resection was performed in 42 of 376 hepatectomies (11.2%). Of these 42 patients, portal resection was performed in 35 (83.3%), hepatic artery resection in 5 (11.9%), and resection of both in 2 (4.8%).

Postoperative mortality after hepatectomy with combined vascular resection (19.0% [8 of 42 patients]) exceeded that of liver resection alone (9.0% [30 of 334 patients]) (P = .04). Final pathology confirmed vascular invasion in 22 of 42 patients (52.4%). Five-year OS after hepatectomy with vascular resection was 22.8%.

By stratifying patients according to the combination of vascular resection and T stage, excluding postoperative deaths, 5-year OS for T1 or T2 stage tumors was significantly higher than for T3 or T4 stage tumors, irrespective of the use of vascular resection (Figure 6).

To our knowledge, this is the first multicenter Italian study on surgical management of HC. Data on 440 patients were collected, and prognostic factors of 376 patients who underwent associated major hepatectomy were analyzed. To address a unique and specific tumor entity, only patients with histologically proven HC¹ were included. Patients with intrahepatic cholangiocarcinoma involving the hepatic hilum (perihilar cholangiocarcinoma) were excluded. Previously, both types of patients were often analyzed together because they required the same kind of hepatobiliary resection^7,15- 17; however, their prognosis is different, with better OS for patients with HC.^14,22 Our objective was to evaluate improvements in operative results and long-term outcome following surgical resection for HC during a 16-year period in a large cohort of patients having the same proven tumor entity.

Obstructive jaundice was the most common symptom (69.1%). With incomplete biliary obstruction it may be absent, and HC may be diagnosed from anorexia, pruritus, abnormal liver test results, or otherwise.²³

All patients underwent preoperative computed tomography and/or magnetic resonance cholangiopancreatography; 49.3% of patients underwent both examinations. However, a nodular lesion at the hepatic hilum was evident in only 49.5% of cases. In the remainder, proximal biliary stenosis was shown. Diagnosis of HC is challenging, and no single imaging method exists for characterization of isolated hilar strictures.²⁴ Approximately 10% to 15% of patients who undergo operation for suspicion of HC are subsequently found to have benign disease.^25- 26 Therefore, the need for preoperative histologic diagnosis of cholangiocarcinoma is often discussed.²⁷ The most common tissue sampling methods are brush cytology and forceps biopsy by endoscopic retrograde cholangiopancreatography. However, the sensitivity of brushing in detecting cancer ranges from 18% to 60%, and the sensitivity of biopsy ranges from 43% to 81%.²⁷ Consistent results were reported in this study in which the sensitivity of brushing and that of biopsy were 36.8% and 57.1%, respectively, suggesting that preoperative histologic diagnosis is not mandatory to evaluate the indication to surgery. Hence, differentiating benign strictures from cancer remains problematic, and the treatment approach for patients with suspicious hilar lesions should remain resection for presumed malignant tumor.²⁶

Staging laparoscopy may identify unresectable extrahepatic biliary carcinoma, preventing unnecessary laparotomy in up to 20% of cases^28- 29; however, its role in patients with HC is yet undefined.³⁰ In our study, 23.5% of the centers routinely performed staging laparoscopy for HC, 88 patients underwent it before resection, and peritoneal carcinomatosis was detected in only 1 case (1.1%), suggesting that staging laparoscopy is not useful in patients with HC for preventing unnecessary laparotomies. Similar results come from a French multicenter study³¹ of 56 patients operated on for HC in 2008: laparoscopy contraindicated laparotomy because of peritoneal carcinomatosis in only 1 case.³¹ Indeed, the yield of staging laparoscopy is higher for gallbladder cancer, in which unresectability is mainly due to peritoneal and liver metastases, compared with HC, in which vascular and biliary infiltration are most frequently involved,³² and the assessment of local resectability is often resolved after laparotomy and extensive dissection of portal and biliary structures.³²

Bile duct resection combined with major hepatectomy is increasingly recognized as the standard treatment for HC, improving the obtainment of R0 resections and long-term survival.^3- 4,12 In our study, the rate of R0 resections was higher with associated liver resection than with biliary confluence resection alone (79.2% vs 65.6%; P = .01). Of course, this type of resection is challenged by the close relationship between tumor and vascular structures at the hepatic hilum. In high-volume compared with low-volume centers in our study, this has resulted in significantly higher rates of associated liver resections (90.8% vs 71.5%; P < .001), of caudate lobe resections (84.7% vs 56.9%; P < .001), and of operations in patients with complex biliary strictures (Bismuth type 3 or 4) and with advanced tumor stages (stage III or IV and/or presence of perineural invasion). However, this did not significantly increase postoperative morbidity and mortality (Table 5), as confirmed by the lack of significance of the factor high-volume center with regard to mortality at multivariate logistic regressions.

The positive effect of R0 liver resection on OS and DFS was confirmed at multivariate analysis (Table 8). The effect on OS was reassessed in a subgroup of patients who underwent resection between 1992 and 1999, having follow-up of 8 years or longer. These truly actual (not actuarial) results demonstrated no 5-year OS following biliary confluence excision alone, whereas 5-year OS following liver resection was 26.6% and 10-year OS was 20.0%. Such patients can be considered cured, and these results resemble the excellent results reported following hepatectomy for colorectal liver metastases.^33- 34

Five-year OS after R0 liver resection was similar for high-volume vs low-volume centers (35.2% vs 35.8%; P = .50), although a significantly higher rate of advanced tumors (stage III or IV) with complex biliary strictures (Bismuth type 3 or 4) were resected in high-volume centers.

Survival after resection for HC improved in recent years together with improvements in surgical technique and perioperative care.^35- 36 Ercolani et al¹³ demonstrated how operative results and long-term outcome improved together with a more aggressive surgical policy. Our multicenter study (Figure 1) shows a progressive increase in the yearly rate of hepatectomies for HC, with 121 hepatectomies (32.2% of all hepatectomies in the 16-year period) performed in the final 2 years.

Time trend analysis (Figure 2) clearly shows changes over time in operative and pathologic results. Indeed, mortality and morbidity peaked before 1996 then trended downward until 2000. This was obviously correlated with advances in surgical technique and perioperative care, as confirmed by a steady, significant decrease in blood transfusion. After 2000, together with improved surgical expertise, there was a more aggressive surgical approach, with increasing rates of caudate lobectomies, vascular resections, and operations on poorly differentiated and advanced stage tumors. This aggressive policy resulted in a moderately increasing trend of operative risk. However, median OS after liver resection significantly increased over the years (Figure 5). Because adjuvant therapies have not significantly improved the long-term outcome after liver resection for HC, these advances are strictly correlated with increased expertise in hepatobiliary surgery.

Biliary drainage in patients with jaundice before liver resection for HC has been an object of debate.^37- 38 Major liver resection in patients with jaundice carries a high risk of postoperative liver failure.³⁷ Conversely, preoperative biliary drainage carries a 20% to 30% risk of cholangitis and sepsis, with increased postoperative infectious morbidity.³⁹ There are no prospective randomized studies analyzing the utility of biliary drainage before major hepatectomy for HC.⁴⁰ Although studies performed to demonstrate the benefits of preoperative biliary drainage were inconclusive, presently major hepatectomy combined with preoperative biliary drainage is established as a safer strategy for HC in Eastern and Western countries.^3,7,9,13,41 The Memorial Sloan Kettering Cancer Center experience⁴² showed that adequate future liver remnant, even in patients with jaundice, can support an extended resection, whereas marginal future liver remnant requires biliary drainage to enhance postoperative function; future liver remnant volume of less than 30% appeared to be a strong indication for preoperative biliary drainage.⁴² In our study, similar results were reported. Indeed, at multivariate logistic regressions, extent of hepatectomy and intraoperative blood transfusion were independent predictors of postoperative mortality. Most centers usually performed biliary drainage before major hepatectomy, and 84.3% of patients with jaundice underwent preoperative biliary drainage. Patients without biliary decompression before right or right extended hepatectomy had higher postoperative mortality than did those operated on following biliary drainage (14.3% vs 10.7%), although this did not reach statistical significance. These results suggest that preoperative biliary drainage should always be performed before right or right extended hepatectomy for HC in patients with jaundice.

Finally, the role of portal vein resection and reconstruction during operations for HC is still debated.^43- 45 Neuhaus et al⁴³ proposed systematic portal vein resection as part of “no-touch” resection of the tumor, showing a 5-year OS of 65%; however, 60-day mortality was 17% and noncurative resections were excluded. Indeed, portal vein reconstruction increases the risk of major hepatectomy, with postoperative mortality of 10% to 20%.³ In our study, postoperative mortality after combined vascular resection exceeded that of liver resection alone (19.0% vs 9.0%; P = .04), and OS for advanced HC (stage T3 or T4) was significantly lower compared with that for less advanced HC, irrespective of the use of vascular resection (Figure 6). These results were confirmed at multivariate analysis in which advanced T stage, and not vascular resection, was an independent predictor of OS and DFS, suggesting that systematic portal vein resection should not be recommended (Table 8).

In conclusion, our multicenter study showed that surgical management of HC has improved over the years with decreased operative risk despite more aggressive surgical policy. Long-term survival after curative liver resection significantly improved, despite the operation on patients with more advanced HC. Preoperative biliary drainage represented a safe strategy before major hepatectomy for HC in patients with jaundice. Finally, pathologic factors (such as advanced T stage, N1 stage, and R1 resection) were the only independent predictors of OS and DFS.