Volume 98, Issue 8 p. 1689-1700
Original Article
Free Access

Patterns of initial disease recurrence after resection of gallbladder carcinoma and hilar cholangiocarcinoma

Implications for adjuvant therapeutic strategies

William R. Jarnagin M.D.

Corresponding Author

William R. Jarnagin M.D.

Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York

Fax: (917) 432-2387

Hepatobiliary Service, Department of Surgery, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021===Search for more papers by this author
Leyo Ruo M.D.

Leyo Ruo M.D.

Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
Sarah A. Little M.B., Ch.B.

Sarah A. Little M.B., Ch.B.

Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
David Klimstra M.D.

David Klimstra M.D.

Department of Pathology, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
Michael D'Angelica M.D.

Michael D'Angelica M.D.

Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
Ronald P. DeMatteo M.D.

Ronald P. DeMatteo M.D.

Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
Raquel Wagman M.D.

Raquel Wagman M.D.

Department of Radiation Oncology, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
Leslie H. Blumgart M.D.

Leslie H. Blumgart M.D.

Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
Yuman Fong M.D.

Yuman Fong M.D.

Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York

Search for more papers by this author
First published: 21 August 2003
Citations: 377

Abstract

BACKGROUND

Current approaches to adjuvant treatment after resection of gallbladder carcinoma (GBCA) and hilar cholangiocarcinoma (HCCA) are based on an incomplete understanding of the recurrence patterns of these diseases. Through an in-depth analysis of the sites of initial recurrence after resection of GBCA and HCCA, the current study aimed to highlight differences in the biology of these tumors and to provide further insight for adjuvant therapeutic strategies.

METHODS

Patients with either GBCA or HCCA who underwent a potentially curative resection were identified prospectively from a maintained database. Specific sites of initial disease recurrence were identified retrospectively and categorized as locoregional (resection margin, porta hepatis, or retroperitoneal lymph nodes) or distant (peritoneal, extraabdominal, or discontiguous liver metastases). Differences in disease recurrence patterns, time to disease recurrence, and overall and site-specific survival were analyzed.

RESULTS

Between May 1990 and August 2001, 177 patients underwent potentially curative resection, 97 for GBCA and 80 for HCCA. Disease recurrence and follow-up data were available for 156 patients (80 with GBCA and 76 with HCCA). The median time to disease recurrence was shorter for patients with GBCA compared with patients with HCCA (11.5 vs. 20.3 months; P = 0.007). Overall, 52 (68%) patients with HCCA and 53 (66%) patients with GBCA had disease recurrene at a median follow-up of 24 months. Of those who developed disease recurrence, isolated locoregional disease as the first site of failure occurred in 15% of patients with GBCA compared with 59% of patients with HCCA (P < 0.001). By contrast, an initial GBCA recurrence involving a distant site, with or without concomitant locoregional recurrence, occurred in 85% of patients compared with 41% of patients with HCCA (P < 0.001). This pattern of disease recurrence was diagnosis specific and did not change significantly when patients were stratified by several clinicopathologic factors, including disease stage and its component variables. Using multivariate analysis, diagnosis was an independent predictor of the site of disease recurrence. Among patients who experienced disease recurrence, survival was greater among the patients with HCCA compared with patients with GBCA (29 months vs. 20.6 months, respectively; P = 0.037). For both tumors, the site of initial disease recurrence had no apparent impact on survival time.

CONCLUSIONS

After resection, recurrent GBCA is much more likely than recurrent HCCA to involve a distant site. GBCA is also associated with a much shorter time to recurrence and a shorter survival period after recurrence. The results demonstrated significant differences in the clinical behavior of these tumors and suggested that an adjuvant therapeutic strategy targeting locoregional disease, such as radiotherapy, is unlikely to have a significant impact in the overall management of GBCA. Conversely, there is at least some rationale for such an approach in patients with HCCA based on the pattern of initial recurrence. Cancer 2003. © 2003 American Cancer Society.

DOI 10.1002/cncr.11699

Adenocarcinoma of the gallbladder (GBCA) and the hepatic duct confluence (hilar cholangiocarcinoma [HCCA]) are the most common malignancies of the biliary tract. Even so, they are rare tumors, representing a fraction of all gastrointestinal carcinomas.1 As a result, to our knowledge only a few tertiary centers to date have had meaningful experience with these diseases and the impact of treatment has been difficult to assess.

Although it is clear that complete resection of these tumors is the most effective and only potentially curative treatment,2-6 definitive roles for adjuvant chemotherapy and radiotherapy have not been established, although both are used commonly. In particular, radiotherapy is recommended often after resection of GBCA and HCCA, based on assumptions that locoregional disease recurrence rates are high for both of these tumors and that improving locoregional disease control improves survival.7-12 However, these assumptions stem from an incomplete understanding of the behavior of these tumors and a presumption that they have a similar disease biology. Although a common molecular pathogenesis has been suggested,13 clinical data suggest that GBCA and HCCA are distinct entities with different clinical behavior.5, 6

Ultimately, detailed analyses of the molecular and genetic changes in human specimens will reveal the fundamental dissimilarities between these two tumors and may offer a rational approach to the use of adjuvant therapy. However, to our knowledge, currently, such data are sparse and of unclear relevance. The current study attempts to define the natural history of GBCA and HCCA by evaluating their recurrence patterns after complete resection and the impact of recurrence type (i.e., locoregional vs. distant) on survival. By analyzing sites of initial disease recurrence, we aim to provide information that will be useful for designing intervention strategies that might have an impact on outcome. We also highlight differences in disease biology, differences that may be correlated eventually with specific molecular and biologic derangements.

MATERIALS AND METHODS

Surgical Approach

The authors' general approach to patients with resectable tumors of the hepatic duct confluence (HCCA) and gallbladder (GBCA) has been described previously.5, 6, 14 Cases and treatment plans are discussed at a twice-weekly multidisciplinary hepatobiliary disease management conference, attended by surgeons, gastroenterologists, medical and radiation oncologists, and diagnostic/interventional radiologists. The surgical approach for patients with HCCA includes resection of the entire extrahepatic biliary apparatus and a supraduodenal lymphadenectomy, usually combined with a partial hepatectomy. In most cases, a partial hepatectomy is comprised of a right or left hepatectomy (lobectomy) or extended hepatectomy (trisegmentectomy). A concomitant caudate resection is performed for patients with tumors extending into the left hepatic duct or for more centrally placed tumors with clear caudate duct involvement.

In patients with GBCA, en bloc cholecystectomy and partial hepatectomy are performed, along with supraduodenal lymphadenectomy, with or without resection and reconstruction of the bile duct. Reoperation and definitive resection are performed for patients who underwent a previous noncurative cholecystectomy (tumor invasion beyond the mucosa), and a similar resectional approach is used. The extent of the hepatic resection performed for patients with GBCA depends mostly on the site of the tumor and the presence or suspicion of major vascular involvement, in which case an extended resection is required. In the absence of such involvement, however, resection of segments IVB and V (which includes the gallbladder fossa) is generally performed. Likewise, biliary resection and reconstruction are performed if there is clear or suspected tumor involvement of the common hepatic or common bile duct. Patients who underwent a previous cholecystectomy and were found to have GBCA confined to the mucosa are typically not recommended for reoperation and resection. For the purposes of the current analysis, a major hepatectomy was defined as a right or left hepatectomy, extended hepatectomy, or central hepatic resection. After a complete resection of GBCA and HCCA is performed, adjuvant therapy typically is not used. After surgery, patients were followed with cross-sectional imaging studies (computed tomographic [CT] or magnetic resonance imaging [MRI] scans) every 4–6 months.

Data Analysis

Patients with HCCA or GBCA who underwent a potentially curative resection were identified from a prospectively maintained database and analyzed retrospectively. Patients who underwent palliative procedures, patients without complete follow-up regarding time to recurrence and/or site of initial disease recurrence, and patients who died perioperatively were excluded. The medical records of eligible patients were evaluated, including surgeon's notes, radiologic imaging, surgical and pathology reports, and discharge summaries. Data regarding the resection performed and pathologic variables including overall stage, T classification, N classification, tumor differentiation, and resection margins were collected. Tumor staging was performed according to the most recent guidelines of the American Joint Committee on Cancer (AJCC).15 For patients with GBCA diagnosed after initial cholecystectomy at another institution, the pathology slides were reviewed at Memorial Sloan-Kettering Cancer Center (MSKCC) to confirm the diagnosis and to determine the depth of invasion (T classification). The additional pathologic data obtained after definitive resection were incorporated into the final stage determination.

Tumor differentiation (well, moderate, or poor) was determined after histologic review of the resected specimen. In the current study, tumors classified as well to moderately differentiated or well differentiated with areas of moderate differentiation were not counted as well differentiated lesions. Long-term follow-up and patient status were determined by physician consult, a telephone questionnaire conducted with the patient or family member, or contact with the patient's primary care physician.

Specific sites of first disease recurrence, time to disease recurrence, and overall and disease-free survival were analyzed. The site(s) of initial disease recurrence were determined from cross-sectional imaging studies (CT or MRI). They were classified as local disease recurrence (hepatic resection margin, bilioenteric anastomosis, or porta hepatis), regional disease recurrence (retroperitoneal lymph nodes), and distant disease recurrence (intrahepatic, peritoneum, or extraabdominal sites). The majority of patients ultimately underwent biopsy confirmation of cancer recurrence, but the date of recurrence was recorded as that when the imaging studies initially identified the new finding(s). Radiologic evidence of tumor recurrence (suspicious new findings and progression of disease documented by serial imaging) was accepted in patients who did not undergo biopsy. When progression of disease was documented by serial imaging, the date of the first suspicious radiologic finding was used as the date of initial disease recurrence. Locoregional recurrences of tumors appearing either synchronous with or after disease recurrence at a distant site were recorded and the impact of such recurrences was recorded (e.g., surgical or radiologic intervention, change in therapy).

Statistical Analysis

Differences in the site(s) of and time to initial disease recurrence between GBCA and HCCA, both overall and stratified by several clinicopathologic variables, were measured. Statistical calculations were performed using SPSS, version 11.0 (Statistical Package for the Social Sciences, Chicago, IL). Continuous variables were compared using the Student t test (two-tailed) and categorical variables were compared using a chi-square test. Survival probabilities were estimated using the Kaplan–Meier method16 and compared using the log-rank test. Survival (in months) was measured from the date of surgery to the date of death or date of last contact. In patients with GBCA who were subjected to an initial noncurative cholecystectomy, the date of definitive resection was used as the date of surgery. Cox regression17 was used to determine independent predictors of disease recurrence (overall, locoregional, and distant) using factors significant on univariate analysis as covariates.

RESULTS

Demographics

Between May 1990 and August 2001, 177 patients underwent a potentially curative resection. Thirteen patients who died perioperatively were excluded from the analysis. Tumor recurrence data were available for 156 patients, 80 with GBCA and 76 with HCCA. The demographics of the patient population are summarized in Table 1. There were 22 men (28%) and 58 women (72%) with GBCA and 41 men (54%) and 35 women (46%) with HCCA. There was a predominance of female patients with GBCA compared with HCCA (P = 0.001). The median age of the patients was 65.5 years for the entire cohort and was similar for both groups.

Table 1. Patient Demographics and Tumor Pathology Results
Characteristics GBCA (n = 80) HCAA (n = 76) All patients (n = 156) P valuea
No. (%) No. (%) No. (%)
Gender
 Male 22 (28) 41 (54) 63 (40) 0.001
 Female 58 (72) 35 (46) 93 (60)
Median age (yrs) (range) 65.0 (34–85) 66.5 (36–84) 65.5 (34–85) 0.9
Previous cholecystectomy
 Yes 54 (68)
 No 26 (32)
R0 resection 72 (90) 58 (76) 130 (83) 0.006
Stageb
 IA 2 (3) 5 (7) 7 (5)
 IB 24 (30) 23 (30) 47 (30)
 IIA 26 (33) 27 (36) 53 (34) 0.5
 IIB 26 (33) 18 (24) 44 (28)
 III 1 (1) 3 (4) 4 (3)
 IV 0 (0) 0 (0) 0 (0)
Stage ≥ IIA 53 (67) 48 (63) 101 (65) 0.61
T stage > 2 46 (58) 44 (58) 90 (58) 1.0
Lymph nodes (+) 26 (33) 19 (25) 45 (29) 0.3
Well differentiatedc 13 (18) 29 (39) 42 (29) 0.006
Adjuvant therapy 11 (14) 7d (9) 18 (11) 0.4
 Radiation 7 6 13
 Chemotherapy 4 2 6
  • GBCA: gallbladder carcinoma; HCCA: hilar cholangiocarcinoma.
  • a P values reflect comparisons between gallbladder carcinoma and hilar cholangiocarcinoma.
  • b Stage indeterminant in one patient with gallbladder carcinoma.
  • c Differentiation indeterminant in nine patients with gallbladder carcinoma and two patients with hilar cholangiocarcinoma.
  • d One patient received both radiotherapy and chemotherapy.

Surgical Procedures

Of the 80 patients with GBCA, 54 had undergone a previous cholecystectomy at another institution where an incidental cancer was discovered. These patients were referred subsequently to MSKCC. Reoperation was not performed in five patients with lesions that were considered to have been treated adequately with the initial resection. These patients are included in the current analysis. The median interval to definitive resection after a previous noncurative cholecystectomy was 1 month. An equal number of patients with GBCA underwent a concomitant major hepatectomy (i.e., 32 patients underwent a right or left hepatectomy or extended hepatectomy and 32 patients underwent segmentectomy IV/V), whereas the remaining patients underwent a wedge resection of the gallbladder fossa. Additional procedures performed included supraduodenal lymphadenectomy in 80 patients, biliary resection/reconstruction in 52 patients, partial colectomy in 7 patients, partial gastrectomyin 1 patient, and partial duodenectomy in 1 patients. All 76 patients with HCCA underwent bile duct resection/reconstruction and supraduodenal lymphadenectomy. This was the primary procedure performed in 18 patients, whereas in 58 patients this procedure was combined with a major hepatectomy (i.e., a right or left hepatectomy or extended hepatectomy in 55 patients or a central hepatectomy in 3 patients). Eighteen patients underwent en bloc caudate lobe resection, four patients had a portal vein resection/reconstruction, and two patients underwent a concomitant pancreaticoduodenectomy.

Adjuvant Therapy

Eighteen patients (11%) received adjuvant therapy after resection (11 patients with GBCA, 7 patients with HCCA). Thirteen patients received radiotherapy and six patients received systemic chemotherapy. One patient with HCCA received both radiotherapy and systemic chemotherapy (Table 1). The majority of patients who received adjuvant treatment were at high risk for disease recurrence—nine had a margin positive resection (R1) and six had metastatic disease to regional lymph nodes.

Pathology

Overall tumor stage, T classification, and the proportion of patients with lymph node involvement were similar between the two groups (Table 1). The majority of all patients had advanced-stage lesions (58% of the primary tumors were T3 or T4, 29% had regional lymph node metastases, and nearly two-thirds of patients had overall Stage IIA disease or worse). A higher proportion of patients with HCCA (39%) had well differentiated tumors compared with patients with GBCA (18%, P = 0.006), although the proportion of patients with poorly differentiated tumors was not significantly different (14% vs. 20%, respectively; P = 0.32). A complete resection with negative histologic margins (R0) was achieved more commonly in patients with GBCA than with HCCA (90% vs. 76%; P = 0.006).

Initial Disease Recurrence

At a median follow-up time of 24 months (range, 2–134 months), 105 patients developed tumor recurrence (53 of 80 [66%] patients with GBCA and 52 of 76 [68%] patients with HCCA; P = 0.68). Sixty-seven patients (64%) had biopsy confirmation, whereas disease progression was documented by serial imaging in 38 patients (36%). Table 2 shows the anatomic location(s) of all initial tumor recurrences. The initial site(s) could not be determined for 13 patients (i.e., 5 patients with GBCA and 8 patients with HCCA). Thirty-three patients (i.e., 18 patients with GBCA and 15 patients with HCCA) had multiple sites of initial disease recurrence, including 10 patients with GBCA and 5 patients with HCCA who had disease recurrence concomitantly at a locoregional and a distant site. Of all HCCA recurrences, 65% occurred at a locoregional site compared with 28% of all GBCA recurrences. By contrast, 72% of all GBCA recurrences occurred at a distant site compared with 36% of all HCCA recurrences. Fifty-one patients had no evidence of disease at the time of the current analysis (i.e., 24 [32%] patients with HCCA and 27 [34%] patients with GBCA).

Table 2. All Sites of Initial Recurrence for Gallbladder Carcinoma and Cholangiocarcinoma
Site No. of GBCA recurrences (%) No. of HCAA recurrences (%)
LOCAL 13 (17) 24 (41)
 Hilum 8 7
 Bilioenteric anastomosis 3 5
 Liver (resection margin) 2 12
REGIONAL 8 (11) 14 (24)
 Retroperitoneal lymph nodes
DISTANT 55 (72) 21 (36)
 Peritoneum 25 5
 Liver (intrahepatic) 8 6
 Lung or mediastinum 7 8
 Abdominal wall/incision 7 0
 Bone 3 0
 Skin 2 1
 Adrenal 2 0
 Axillary lymph nodes 1 0
 Brain 0 1
Total no. of recurrences 76 59
Total no. of patients with recurrence 53 (66) 52 (68)
  • GBCA: gallbladder carcinoma; HCAA: hilar cholangiocarcinoma.

Of the 92 patients whose site(s) of first disease recurrence were known with certainty (Table 3), a locoregional recurrence alone was observed in only 15% of patients with GBCA (7 of 48) compared with 59% of patients with HCCA (26 of 44; P < 0.001). By contrast, 85% of patients with GBCA (41 of 48) had an initial disease recurrence at a distant site, with or without a concomitant locoregional recurrence, compared with 34% of patients with HCCA (18 of 44; P < 0.001). The significantly greater propensity for GBCA to recur at distant sites and HCCA at locoregional sites was not affected by tumor stage, T classification, lymph node status, or the extent of the hepatic resection performed. In patients with microscopically involved resection margins (R1), locoregional recurrence was slightly higher for patients with GBCA (20%) and, paradoxically, somewhat lower for patients with HCCA (47%). However, the small number of patients (5 patients with GBCA, 15 patients with HCCA) likely accounts for the lack of a significant difference. Likewise, well differentiated GBCA appeared to have a somewhat higher likelihood of locoregional recurrence (25%), but again the number of patients was small (4 patients with GBCA, 14 patients with HCCA). Overall, the proportion of patients with GBCA who experience initial recurrence at a distant site (75–90%) and the proportion of patients with HCCA who experience recurrence at locoregional sites alone (47–75%) remained relatively constant in the current analysis. For patients with GBCA and patients with HCCA, the variations within each stratification subgroup were not statistically significant.

Table 3. Proportion of Patients with Initial Disease Recurrence at a Locoregional Site Only or at a Distant Site (with or without Concomitant Locoregional Disease Recurrence)
Characteristic Percentage of patients with locoregional disease recurrence only Percentage of patients with distant-disease recurrence
GBCA HCCA P value GBCA HCCA P value
All patients (%) 7/48 (15) 26/44 (59) < 0.001 41/48 (85) 18/44 (41) < 0.001
Tumor stage
 < IIA 10 56 0.018 90 44 0.018
 ≥ IIA 16 61 < 0.001 84 39 < 0.001
T stage classification
 ≤ 2 6 63 < 0.001 94 37 < 0.001
 > 2 19 56 0.004 81 44 0.004
Lymph node status
 Negative 12 56 < 0.001 88 44 < 0.001
 Positive 19 67 0.006 81 33 0.006
Resection margin status
 R0 15 66 < 0.001 85 34 < 0.001
 R1 20 47 0.29 80 53 0.29
Differentiation
 Well 25 57 0.26 75 43 0.26
 Moderate/poor 16 64 < 0.001 84 36 < 0.001
Preoperative interventiona
 Yes 19 53 0.004 81 47 0.004
 No 6 75 < 0.001 94 25 < 0.001
Major hepatic resection
 Yes 23 53 0.026 77 47 0.026
 No 8 71 < 0.001 92 29 < 0.001
Time to recurrence (mos)
 12 18 75 0.001 82 25 0.001
 18 14 65 < 0.001 86 35 < 0.001
 24 14 60 < 0.001 86 40 < 0.001
  • GBCA: gallbladder carcinoma; HCCA: hilar chloangiocarcinoma.
  • a Previous noncurative cholecystectomy in patients with GBCA, preoperative biliary drainage in patients with HCCA.

Eighteen patients had papillary tumors. All three patients with GBCA with this tumor morphology had disease recurrence at distant sites. Of the 15 patients with HCCA with papillary tumors, the overall disease recurrence rate was 73% compared with 67% for patients with nodular-sclerosing tumors (P = 0.65). Similarly, tumor morphology did not influence site-specific disease recurrence rates (locoregional only, 56% papillary vs. 60% nodular-sclerosing; P = 0.81; any distant site, 44% papillary vs. 40% nodular-sclerosing; P = 0.81).

The use of adjuvant therapy had little impact on the pattern of disease recurrence. Of the patients who received adjuvant radiotherapy, isolated locoregional disease recurrence occurred in 17% of patients with GBCA compared with 40% of patients with HCCA. Although a slight reduction in initial locoregional HCCA recurrence is suggested, it must be emphasized that a meaningful analysis of adjuvant therapy and recurrence is not possible because it was used selectively in a fraction of patients (n = 18; 11%), most of whom had tumor involved resection margins or regional lymph node metastases, and there was little influence on the overall results.

Fifty-four patients with GBCA underwent a previous, noncurative cholecystectomy. These patients had a similar disease recurrence rate (35 of 54 patients [65%]) compared with patients with no antecedent cholecystectomy (18 of 26 patients [69%]; P = 0.70). Similarly, there was no effect on the proportion in each group with initial disease recurrence at a distant site (81% with a previous cholecystectomy vs. 94% with no previous cholecystectomy; P = 0.25). Furthermore, of those with an initial disease recurrence at one or more distant sites, the proportion with peritoneal metastases or wound implants was nearly identical (65% with a pevious cholecystectomy, 68% with no previous cholecystectomy). Fifty-seven patients with HCCA (75.0%) underwent preoperative biliary intubation (25 intubations were performed percutaneously and 32 were performed endoscopically). Disease recurrence rates were similar in patients with any biliary stent compared with patients with no biliary stents (39 of 57 [68%] vs. 13 of 19 [68%], respectively; P = 1.0) and in patients with percutaneous (17 of 25 [68.0%]) versus endoscopic stents (22 of 32 [69%]; P = 0.95). An initial HCCA recurrence at a distant site was not significantly different in patients with any preoperative stent (47% compared with 25% no stent; P = 0.19) and after endoscopic versus percutaneous stents (50% vs. 43%, respectively; P = 0.69). Only one patient with GBCA underwent preoperative biliary decompression, precluding any meaningful analysis.

Of the 41 patients with GBCA whose initial recurrence involved a distant site, 14 (34%) developed locoregional disease at some point in the course of their illness (10 patients with synchronous disease, 4 patients with metachronous disease). By contrast, 11 of 18 patients (61%) with an initial distant HCCA recurrence also had locoregional involvement (5 patients with synchronous disease, 6 patients with metachronous disease). Thus, of the group of patients who had initial disease recurrence at distant sites alone, only 13% (4 of 31) of patients with GBCA developed late locoregional metastases compared with 46% (6 of 13) of patients with HCCA (P = 0.027). In addition, of the 25 patients (14 patients with GBCA and 11 patients with HCCA) with locoregional recurrence, either simultaneous with or after disease recurrence at a distant site, 12 (48%) had no signs or symptoms related to the locoregional disease. The remaining 13 patients developed problems that required intervention and/or impacted on subsequent therapy (e.g., obstructive jaundice in 11 patients, gastric outlet obstruction in 1 patient, and hemorrhage from tumor eroding into the duodenum in 1 patient).

The median time to disease recurrence was significantly shorter for patients with GBCA (11.5 months [range, 2–63 months]) compared with patients with HCCA (20.3 months [range, = 4–111 months]; P = 0.007). Sixty-two percent of GBCA recurrences occurred within 12 months of surgery and 88% occurred within 24 months, compared with 30% (P = 0.001) and 56% (P < 0.001), respectively, for patients with HCCA. The median time to a locoregional only GBCA recurrence was 10.6 months (range, 3.2–45 months) compared with 19.1 months (range, 5–111 months) for HCCA recurrence (P = 0.12). The median time to a distant GBCA recurrence was 11 months (range, 2–63 months) versus 20.1 months (range, 4–64 months) for HCCA recurrence (P = 0.009). In patients with HCCA, there was a trend toward more frequent distant site involvement (25–40%) as the time from the resection increased (Table 3). However, the median time to disease recurrence at a locoregional site only was not significantly different from the median time to disease recurrence at a distant site. No such trend was noted in patients with GBCA (82–86%), and like patients with HCCA, the median time to GBCA recurrence at a locoregional site only or a distant site was similar.

Using univariate analysis, several variables were associated with an increased risk of recurrent disease at any site (any recurrence). By contrast, only a diagnosis of HCCA and male gender were predictive of an initial, isolated locoregional disease recurrence, whereas a diagnosis of GBCA, tumor-involved regional lymph nodes, and moderate or poor tumor differentiation was associated with disease recurrence at a distant site (Table 4). Logistic regression analysis using the covariates listed in Table 5 identified positive regional lymph nodes, moderate/poor tumor differentiation, and a positive resection margin as independent predictors of any disease recurrence. Male gender was of borderline significance (P = 0.07). A diagnosis of HCCA was the only independent predictor of an isolated locoregional recurrence (P < 0.001; hazards ratio [HR] = 5.43, 95% confidence interval [CI], 2.13–13.81) and a diagnosis of GBCA was the only independent predictor of disease recurrence at a distant site (P = 0.006; HR = 2.9, 95% CI,; 1.35–6.21) when the multivariate model included only those variables significant for each on univariate analysis. When the variables univariately significant for any disease recurrence were added to the multivariate model, diagnosis remained the most powerful predictor of recurrence type. A diagnosis of HCCA was again the only factor associated with a locoregional recurrence. For disease recurrence at a distant site, a diagnosis of GBCA was again an independent predictor but a positive resection margin also emerged as a significant variable.

Table 4. Univariate Analysis of Clinicopathologic Factors Associated with any Recurrence, Locoregional Recurrence Only, and Distant Recurrence after Resection in all Patients
Variables Percentage of patients with any recurrence Percentage of patients with locoregional recurrence only Percentage of patients with distant recurrence (with/without locoregional recurrence)
Diagnosis
 GBCA 66 P = 0.77 9 P < 0.001 55 P = 0.001
 HCCA 68 38 26
Gender
 Male 76 P = 0.05 32 P = 0.04 41 P = 0.91
 Female 61 17 41
Age (yrs)
 < 60 y 69 P = 0.81 25 P = 0.7 42 P = 0.94
 ≥ 60 y 61 22 41
Resection margin status
 Positive 88 P = 0.02 35 P = 0.16 52 P = 0.21
 Negative 63 21 38
Tumor stage
 < IIA 57 P = 0.06 20 P = 0.56 33 P = 0.15
 ≥ IIA 72 25 45
T stage classification
 ≤ 2 62 P = 0.24 21 P = 0.67 38 P = 0.46
 > 2 71 24 44
Lymph node status
 Positive 87 P = 0.001 31 P = 0.19 54 P = 0.05
 Negative 59 29 36
Differentiation
 Well 50 P = 0.008 23 P = 0.74 23 P = 0.023
 Moderate/poor 73 26 44
Preoperative interventiona
 Yes 67 P = 0.79 23 P = 0.89 41 P = 0.80
 No 69 24 43
Major hepatic resection
 Yes 68 P = 0.88 26 P = 0.36 38 P = 0.41
 No 67 19 45
  • GBCA: gallbladder carcinoma; HCCA: hilar cholangiocarcinoma.
  • a Previous noncurative cholecystectomy in with GBCA patients, preoperative biliary drainage in patients with HCCA.
Table 5. Multivariate Analysis of Clinicopathologic Factors Associated with any Recurrence, Locoregional Recurrence only, and Distant Recurrence after Resection in all Patients
Variables Any recurrence Locoregional recurrence only Distant recurrence (CNCR CO430-O3)
P value HR (95% CI) P value HR (95% CI) P value HR (95% CI)
Diagnosis (HCCA/GBCA) < 0.001 6.29 (2.27–17.39) 0.002 0.25 (0.11–0.59)
Male gender 0.07 2.16 (0.95–4.92) 0.22 1.78 (0.71–4.43) 0.79 1.12 (0.48–2.62)
Positive resection margin 0.025 4.71 (1.21–18.36) 0.94 1.04 (0.34–3.17) 0.02 3.69 (1.26–10.8)
Tumor stage ≥ IIA 0.64 1.23 (0.52–2.94) 0.93 1.05 (0.35–3.13) 0.98 1.01 (0.4–2.57)
Positive lymph nodes 0.01 4.07 (1.4–11.82) 0.14 2.27 (0.77–6.72) 0.43 1.46 (0.57–3.79)
Moderate/poor differentiation 0.013 3.0 (1.26–7.12) 0.31 1.72 (0.61–4.82) 0.12 2.16 (0.83–5.62)
  • HR: hazards ratio; CI: confidence interval; HCCA: hilar cholangiocarcinoma; GBCA: gallbladder carcinoma.

Survival

The median overall survival was 31.3 months (range, 1.7–134.4 months) for all patients with GBCA and 38.3 months (range, 2.2–141.2 months) for all patients with HCCA (P = 0.89). In patients who developed disease recurrence, survival was longer in patients with HCCA (29 months[ range,4–141.2 months]) compared with patients with GBCA (20.6 months [range, 2.1–71.1 months;]; P = 0.037). Much of this difference is the result of the markedly better survival in patients with HCCA who experience disease recurrence at a distant site (37.1 months [range, 4–84.1 months]) compared with patients with GBCA (22.4 months [range, 2.1–66 months]; P = 0.017). Patients with HCCA who experienced initial recurrence at a locoregional site alone had a median survival period of 28.4 months (range, 6–141.2 months), which was longer than that for patients with GBCA (18.3 months [range, 4.9–71.2 months]) but was not significantly different (P = 0.35). In addition, the survival difference in patients with HCCA with an initial locoregional only versus a distant site recurrence was not significant (P = 0.66). This was also true of patients with GBCA (P = 0.95). The median time from disease recurrence to death was similar for both tumors (8.8 months for HCCA and 8.5 months for GBCA; P = 0.19), suggesting that the longer time to recurrence in patients with HCCA compared with patients with GBCA accounts for the overall difference in survival.

The impact of the site of disease recurrence on survival was analyzed further by comparing patients who experienced recurrence at distant sites alone, with no locoregional involvement at any time, with patients who developed locoregional disease at some point, either alone or combined with a distant site recurrence (synchronous or metachronous). For patients with GBCA (Fig. 1) and patients with HCCA (Fig. 2), the median survival in patients with locoregional metastases as a part of or the sole manifestation of the recurrence pattern was not significantly different from that in patients with distant-only disease.

Details are in the caption following the image

Median survival (23.2 months) of 24 patients with gallbladder carcinoma who experienced disease recurrence at distant sites alone, with no locoregional involvement at any time (dashed line), compared with the median survival (18.3 months) of 21 patients who developed locoregional disease at some point, either alone or combined with disease recurrence at a distant site (synchronous or metachronous; (solid line).

Details are in the caption following the image

Median survival (23.8 months) of eight patients with hilar cholangiocarcinoma who experienced disease recurrence at distant sites alone, with no locoregional involvement at any time (dashed line), compared with the median survival (35.1 months) of 37 patients who developed locoregional disease at some point, either alone or combined with disease recurrence at a distant site (synchronous or metachronous; solid line).

DISCUSSION

GBCA and HCCA, although the most common malignant tumors of the biliary tree, are rare, accounting for a fraction of all gastrointestinal carcinomas. As a result, there remains considerable uncertainty regarding the natural history and optimal management of these diseases.

This uncertainty is particularly apparent in the area of adjuvant therapy, for which there is no consensus and no standard approach. Although resection is now well established as the most effective and only potentially curative treatment, it remains unknown whether adjuvant chemotherapy and/or radiotherapy reduces the high incidence of disease recurrence. This is due largely to the rarity of these diseases and the consequent difficulty in accruing patients into randomized trials. Despite the lack of proven efficacy,18, 19 adjuvant treatment, particularly radiotherapy, is recommended and used frequently. This practice has been fostered by a number of small, single-institution, nonrandomized studies that have suggested a survival benefit of radiotherapy as an adjuvant to resection of both GBCA and HCCA.7-12 However, a review of the pertinent literature fails to provide a clear rationale for its use. In a recent review of postoperative radiotherapy in biliary tract carcinoma, Macdonald and Crane9 summarized the problem with the following observation: “Although few studies provide data on the pattern of failure, improvement in survival presumably is caused by a reduction in the recurrence of locoregional disease” (p. 942).

The approach to adjuvant therapy for GBCA and HCCA is based more on a presumption that they behave similarly and that the same adjuvant therapeutic strategy is equally likely to offer a potential benefit. To date, however, there have been little data published regarding disease recurrence patterns and their impact on survival after resection. Therefore, there is no firm basis to recommend an adjuvant approach that targets only locoregional disease recurrence, such as radiotherapy. The current study attempts to fill this gap through an in-depth analysis of the anatomic sites of initial disease recurrence, and in so doing, perhaps also to highlight differences in disease biology. Disease recurrence sites were categorized as locoregional or distant based not only on proximity to the resection field and likely mechanism of dissemination but also, in part, on the likelihood that such anatomic sites would be included in a regimen of postoperative radiotherapy. Locoregional disease recurrence along the resection margin, in the porta hepatis, or in the retroperitoneum likely arises from microscopic residual disease or from disease in lymphatics and potentially could be included in a radiation treatment field. By contrast, intrahepatic and other disease recurrences at a distant site result from hematogenous spread and would not be treatable with this modality.

The data demonstrate several dissimilarities between GBCA and HCCA, the most striking of which is the marked propensity for initial GBCA recurrences to involve a distant site, either alone or concomitant with locoregional disease recurrence. This is in contrast to HCCA, which recurs initially at locoregional sites alone in the majority of cases. This pattern of first disease recurrence was diagnosis specific and varied little when patients were stratified by a number of clinicopathologic factors, including tumor stage and its component variables. Furthermore, on multivariate analysis, diagnosis was the most potent predictor of initial disease recurrence site. This was also true when the multivariate model included variables that were univariately significant for any disease recurrence. Although not significant in the analyses of site-specific disease recurrence, these additional variables were added to ensure that no potentially important associations were missed. Indeed, for disease recurrence at a distant site, a positive resection margin emerged as a significant variable, although it was not on univariate analysis. This finding suggests that both a diagnosis of GBCA and a positive resection margin are predictive of disease recurrence at a distant site. However, this finding also may, in part, reflect the relation between diagnosis and an R0 resection that was shown to exist.

Biliary tract interventions before definitive resection were common in patients with GBCA (cholecystectomy) and in patients with HCCA (biliary drainage), interventions that could potentially cause tumor cell dissemination and affect the patterns of disease recurrence. For GBCA in particular, it has been speculated that a previous noncurative cholecystectomy might predispose patients to peritoneal implantation.20-23 Although peritoneal metastases represented a major proportion of distant GBCA recurrences in the current study, a previous cholecystectomy was not a contributing factor. Patients who underwent definitive resection for GBCA at the initial surgical procedure had an equally high rate of disease recurrence at a distant site and the proportion of patients with peritoneal recurrence was nearly identical. In fact, patients with no antecedent cholecystectomy had a higher incidence of initial disease recurrence at a distant site, although this difference was not significant. Similarly, preoperative biliary intubation did not impact the overall disease recurrence rate nor the pattern of disease recurrence in patients with HCCA.

The extent of resection was neither a predictor of overall nor a predictor of of site-specific disease recurrence. The surgical management of these tumors has been the subject of some controversy, but it is now clear that complete resection typically requires a partial hepatectomy and often an extended resection.5, 6 This is particularly true of HCCA, which often extends to second-order biliary radicals and/or involves major portal vascular structures, features that mandate a major hepatectomy for complete tumor clearance. GBCA is somewhat less likely to be associated with such findings, allowing a more limited hepatic resection in some patients. Indeed, recent studies have shown superior survival when resection of HCCA also includes a major hepatectomy.6, 24 In the current study, patients with HCCA who underwent a limited resection had a higher incidence of disease recurrence at any site but the pattern of recurrence was no different from those who underwent a concomitant major hepatectomy (data not shown).

Based on the patterns of disease recurrence observed in the current study, it would be difficult to advocate the routine use of adjuvant radiotherapy for GBCA, given the fraction of patients who experience disease recurrence at locoregional sites alone. There would appear to be no potential benefit, with respect to survival, to the majority of patients who develop disease recurrence at distant sites. Although certain characteristics of GBCA were associated with a slightly higher incidence of locoregional disease recurrence (e.g., well differentiated tumors and a positive resection margin), the number of patients in these subgroups was small and the benefit of adding adjuvant radiotherapy would seem to be limited. Conversely, there would at least be some rationale for its use after resection of HCCA because it is much more likely to be associated with isolated locoregional disease. However, distant site disease recurrence was not an insignificant problem in patients with HCCA and was higher proportionally in the current study than has been reported previously.25

Because locoregional disease recurrence often results in significant hepatobiliary-related morbidity, it is perceived to result in a more precipitous clinical decline than disease recurrence at other sites. This has been used as an argument for the liberal use of adjuvant radiotherapy. Even if adjuvant radiotherapy is not curative, it might provide a prolonged period of symptom-free survival by delaying the appearance of clinically relevant locoregional disease and may extend overall survival, notwithstanding the presence of distant metastases.9 However, the results of the current study do not support this view for either tumor. For both GBCA and HCCA, survival was not influenced by the site of disease recurrence. Initial disease recurrence at a locoregional site alone was not associated with a shorter survival compared with disease recurrence at a distant site. Similarly, patients who experienced disease recurrence only at distant sites, without any locoregional failure, had a survival period that was no different from that of patients with some component of locoregional disease recurrence. Furthermore, of the patients who had a distant site disease recurrence and either a synchronous or late locoregional recurrence, nearly one-half did not develop any apparent symptoms related to the local problem. These observations suggest that recurrent locoregional disease does not necessarily equate to a shorter survival and, in some patients, may have little influence on the clinical course.

The marked disparity in disease recurrence patterns between patients with GBCA and patients with HCCA not only has implications for adjuvant therapeutic strategies but also underscores differences in underlying disease biology. Additional differences in patient demographics, time to disease recurrence, and the survival period after disease recurrence further support this view. That the distinct clinical features of GBCA and HCCA reflect dissimilarities in the underlying molecular abnormalities is not an unreasonable assumption. To a large extent, these abnormalities dictate tumor behavior. However, only a full elucidation of the molecular genetic changes associated with these tumors will show conclusively that they arise and progress through different mechanisms and will allow a firm conclusion in this regard. Studies concerning the molecular biology of biliary tract carcinomas have demonstrated a number of abnormalities, including mutations of protooncogenes, cell cycle regulatory/tumor suppressor genes, and mismatch repair genes.26-29 The available data suggest that the cellular and molecular abnormalities in biliary malignancy vary by location, although there is some overlap, and the studies to date have included small numbers of patients. It is expected that further research in this area will identify the specific changes underlying the observed differences in clinical behavior, provide the basis for improvements in current nonsurgical therapy, and perhaps identify patients who are most likely to benefit from an adjuvant strategy that targets locoregional disease recurrence.

Despite the emergence over the past several years of safe hepatic resection as a potentially curative treatment for biliary malignancy, high disease recurrence rates remain a major barrier to long-term survival for many patients. Effective adjuvant therapy is necessary if the mortality related to these tumors is to be reduced further. However, future studies in this area must take into account the marked differences in disease recurrence patterns between GBCA and HCCA. Adjuvant treatment targeting locoregional disease, such as radiotherapy, is not likely to have an impact in GBCA. Conversely, such an approach at least fits the pattern of failure of HCCA and may have a role, although efficacy has yet to be shown.