Endoscopic retrograde cholecystogram (A) and cystic arteriogram (B) in a 65-year-old man with protruding-type gallbladder cancer with histopathologically proven subserosal invasion. An encasement in the superficial branch of the cystic artery suggests subserosal invasion (arrow).
A, Computed tomographic image in a 62-year-old man with pT4 tumor in the cystic duct. A histopathologically positive peripancreatic node is visible (arrow). This node is flat and less than 10 mm in diameter. The preoperative diagnosis of lymph node metastasis in such cases is difficult. B, Computed tomographic image in a 60-year-old man with pT3 tumor in the fundus and neck of the gallbladder. Marked swelling of a peripancreatic node (arrow) suggests pN2 disease.
A, Distribution of positive lymph nodes in pN1 cases (n = 21). Cystic nodes (N12c) were the most prevalent sites of metastasis, followed by pericholedochal nodes (N12b). There were no positive nodes in N12p. Sampling of N12c and N12b nodes would have detected 94.7% of the pN1 cases. N8a, N8p, N13a, and N17 nodes are N2 nodes, which are automatically assumed to be negative in pN1 cases. The metastatic rate (number of patients with positive nodes divided by number of patients sampled) of each lymph node group is expressed as a percentage. Regional lymph nodes were classified according to the Japanese Classification of Gastric Carcinoma,14 and General Rules for Surgical and Pathological Studies on Cancer of Biliary Tract15 (thick ovals, N1 nodes; thin ovals, N2 nodes). B, Distribution of positive lymph nodes in pN2 cases (n = 37). In pN2 cases, posterosuperior peripancreatic nodes (N13a) were the most prevalent sites of metastasis, followed by cystic (N12c), pericholedochal (N12b), and periportal (N12p) nodes. All N1 nodes except N12h had metastatic rates higher than 50%. Sampling of N13a nodes would have detected 79.3% of pN2 cases.
Relationship between positivity of the key nodes (N12b, N12c, and N13a) and pN in patients with image-T2 (A) and image-T3 or -T4 tumor where Image-T is the preoperative diagnosis of the T factor (B).
Effect of bile duct (BD) resection on overall survival in patients with nodal involvement and without bile duct invasion (P = .27, log-rank test).
Survival in patients with pN2 gallbladder carcinoma according to the extent of surgery. A, Pancreaticoduodenectomy (PD) (P = .10, log-rank test). B, Extended lymph node dissection (P = .01, log-rank test).
Algorithm for the extent of lymph node dissection based on preoperative diagnosis of the T factor (image-T) and intraoperative frozen section examination of key lymph nodes. Dashed rectangles indicate that a procedure has not yet been established because none of the patients in these subgroups survived for 5 years. A, Image-T1 disease; B, image-T2 disease; and C, image-T3 or -T4 disease.
Algorithm for the extent of local resection based on preoperative diagnosis of the T factor (image-T) and intraoperative frozen section examination of key lymph nodes. Dashed rectangles indicate that a procedure has not yet been established because none of the patients in these subgroups survived for 5 years. Extent of liver resection is determined by the extent of liver and bile duct invasion. PD indicates pancreaticoduodenectomy. A, Image-T1 disease; B, image-T2 disease; and C, image-T3 or image-T4 disease.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kokudo N, Makuuchi M, Natori T, et al. Strategies for Surgical Treatment of Gallbladder Carcinoma Based on Information Available Before Resection. Arch Surg. 2003;138(7):741–750. doi:10.1001/archsurg.138.7.741
Precise preoperative staging for gallbladder carcinoma is difficult, despite recent advances in hepatobiliary imaging. However, the most accurate preoperative staging may be possible by integrating preoperative key data.
To establish useful strategies for the surgical treatment of gallbladder cancer based on information available before resection.
University hospital and tertiary referral cancer center.
Patients and Methods
From January 1, 1978, through March 31, 2001, 152 patients with gallbladder cancer underwent surgical resection with curative intent. Preoperative diagnoses of the T factor (image-T) and N factor (image-N) in the TNM classification were determined by evaluating all findings of diagnostic imaging, including ultrasonography, enhanced computed tomography, endoscopic ultrasonography, and angiography. The distribution of lymph node metastasis and prognostic factors were also analyzed.
The overall diagnostic accuracy for image-T was 52.6% (95% confidence interval, 44.7%-60.6%) and was lower in patients with pT1 and pT2 disease (37.2% and 33.9%, respectively). However, image-T was a significant predictor of lymph node metastasis and patient outcome. Preoperative staging for N was more difficult, with only 24.5% (95% confidence interval, 12.4%-36.5%) of the node-positive patients being correctly diagnosed. An analysis of harvested lymph nodes showed that the cystic, pericholedochal, and posterosuperior peripancreatic nodes were the most prevalent sites of metastasis, and these were considered key nodes for the lymphatic spread of gallbladder cancer. By combining data on image-T and positivity of these key nodes, more accurate TNM staging was possible. Although an extended lymph node dissection provided significantly better survival in patients with pN2 disease, there was no survival advantage to more radical operations, including bile duct resection or pancreaticoduodenectomy.
Although precise preoperative TNM staging for gallbladder carcinoma was difficult, the most accurate staging before resection was possible by integrating image-T classification and data from the intraoperative histopathologic examination of key lymph nodes. Based on this staging, we propose algorithms for the surgical treatment of gallbladder carcinoma.
RECENT ADVANCES in methods for hepatobiliary imaging, including ultrasonography (US), computed tomography, and endoscopic ultrasonography (EUS), have contributed to an increase in the rate of detection of carcinoma of the gallbladder at a resectable stage.1,2 Because the operative procedures to be used and patient outcome are affected by the depth of cancer invasion and the extent of lymph node metastasis,3-7 the preoperative and intraoperative diagnosis of 2 important factors, pT and pN staging according to the TNM classification,8 is crucial. Although there have been several studies on the accuracy of diagnostic imaging for gallbladder cancer9,10 and on surgical strategies based on histopathologic staging,3-7 only a few have correlated diagnostic imaging and intraoperative diagnosis with surgical strategies, ie, extent of local resection and lymph node dissection.5,11,12
To establish useful strategies for the treatment of gallbladder cancer, we conducted a retrospective study to address 3 questions. (1) What is the accuracy of preoperative diagnostic imaging for the staging of primary tumor and nodal involve ment in gallbladder carcinoma? (2) How can we best assess lymph node metastasis before and during the operation? (3) How can we plan an operation based on preoperative and intraoperative findings? In this article, we propose algorithms based on preoperative findings of diagnostic imaging and intraoperative key lymph node sampling, data that are available before resection.
Between January 1, 1978, and March 31, 2001, 201 patients with gallbladder carcinoma were treated at the University of Tokyo Hospital and Cancer Institute Hospital. Of these, 162 patients underwent surgical removal of the tumor. Ten patients were excluded from the study because of apparently incomplete noncurative operations. The remaining patients consisted of 60 men and 92 women, aged 41 to 86 years (mean age, 65.0 years).
Preoperative diagnostic imaging consisted of routine US, enhanced computed tomography, EUS (after 1987), angiography, and endoscopic retrograde cholangiography. The latter 3 procedures were optional and were performed only when the physician in charge judged them to be necessary. Preoperative diagnosis of T factor (image-T) and N factor (image-N) according to the TNM classification8 was determined by evaluating all findings of diagnostic imaging. Diagnostic criteria for the depth of tumor invasion are listed in Table 1.9,13 Diagnostic criteria for lymph node metastasis were simple: a round lymph node, not a flat one, with swelling greater than 10 mm in diameter was considered metastatic.
Regional lymph nodes were classified according to the TNM classification of the American Joint Committee on Cancer and subdivided according to the Japanese Classification of Gastric Carcinoma14 and the General Rules for Surgical and Pathological Studies on Cancer of Biliary Tract15 (Table 2). Cystic nodes (N12c), pericholedochal nodes (N12b), nodes in the hepatic hilum (N12h), nodes along the proper hepatic artery (N12a), and periportal nodes (N12p) were considered N1, and others were considered N2. Involvement of the interaortocaval nodes (N16) was classified as M1 disease. A metastatic ratio was calculated for each node group, with the number of patients with positive nodes in a particular node group divided by the number of patients who underwent sampling of the same nodes.
Operative procedures were selected based on the preoperative diagnosis. A simple cholecystectomy was performed for image-T1 cases. Only N12c and N12b lymph nodes were dissected (partial dissection of N1 nodes). Extended cholecystectomy (cholecystectomy plus wedge resection of the liver bed) was performed for patients with T2 carcinoma. In general, a "standard lymph node dissection," including N1, N8a, N8p, and N13a nodes (Table 2), was performed in this subgroup of patients. If necessary, for patients with advanced carcinoma (T3 or greater), extended liver resection and combined resection of adjacent organs, including the duodenum, pancreas head, and transverse colon, were performed to achieve potentially curative surgery. A standard lymph node dissection or "extended dissection" (standard dissection plus interaortocaval node dissection around the left renal vein7,16) (Table 2) was also performed in this subgroup. If peripancreatic nodes (N13a) were positive for cancer, pancreaticoduodenectomy was performed. The bile duct was resected in 55 patients. The reason for bile duct resection was suspected bile duct invasion in 30 patients, presence of lymph node metastasis in 21, and concomitant pancreaticobiliary maljunction in 4. The surgical procedures performed are summarized in Table 3. There were 2 operative mortalities; these patients died of hepatic failure after hepatopancreaticoduodenectomy. There was no operative mortality after 1986.
All values represent the mean ± SEM. Fisher exact test was used for 2 × 2 categorical data. Survival time was calculated from the date of surgery until death. Survival curves were generated using the Kaplan-Meier method and then compared using the log-rank test. A multivariate stepwise Cox regression analysis (backward elimination method)18 was performed to identify factors that were independently associated with survival after the operation. The SPSS program, version 6.0 for Macintosh (SPSS Inc, Chicago, Ill), was used for statistical analysis.
The diagnostic accuracy of preoperative imaging for pT is summarized in Table 4. Overall, correct staging of pT was possible in 80 patients (52.6%; 95% confidence interval [CI], 44.7%-60.6%). The diagnostic accuracy was lower for pT1 and pT2 disease (37.2% and 33.9%, respectively). Forty-four patients were misdiagnosed as having benign disease, and 4 of these turned out to have pT3 tumors. The diagnostic accuracy of each imaging modality is given in Table 5. Angiography was effective for detecting subserosal invasion (75.9% accurate; 95% CI, 60.3%-91.4%). Encasement or interruption of cystic artery branches was the most specific sign (Figure 1). Endoscopic ultrasonography was effective for staging pT1 lesions in 69.2% (95% CI, 44.1%-94.3%) of the cases. However, a precise diagnosis of pT2 tumor was possible in only 52.4% (95% CI, 31.0%-73.7%).
The diagnostic accuracy for pT before 1990 (when EUS was seldom used) and after 1990 (when EUS was readily available) was compared. The overall diagnostic accuracy before 1990 was 48.1% (n = 52; 95% CI, 34.5%-61.7%) and after 1990 it was 52.0% (n = 100; 95% CI, 42.2%-61.8%). The improvement in diagnostic accuracy was not statistically significant (P = .73).
The preoperative diagnosis of lymph node metastasis was more inaccurate (Table 6). Although the incidence of correct image-N staging was high (69.3%; 95% CI, 61.6%-76.9%), only 12 patients (24.5%; 95% CI, 12.4%-36.5%) were correctly diagnosed as having node-positive disease (Figure 2). Twelve patients were excluded from this analysis because neither computed tomography nor US was suitable for evaluation.
The distribution of positive lymph nodes in pN1 (n = 20) and pN2 (n = 35) cases is shown in Figure 3A and B, respectively. In pN1 cases, N12c (cystic node) was the most prevalent site of metastasis, followed by N12b (pericholedochal node). In pN2 cases, N13a (posterosuperior peripancreatic node) was the most prevalent site of metastasis, followed by N12c. Sampling of N12c and N12b nodes would have detected 94.7% of the pN1 cases and 100.0% of the pN1 cases under image-T2. Sampling of N13a nodes would have detected 79.3% of the pN2 cases.
The distribution of positive lymph nodes was then analyzed according to image-T. Among 25 patients who were preoperatively diagnosed as having T1 disease (image-T1), there was only 1 N-positive case (4.0%), in which metastasis was limited to the cystic node (N12c). Among 27 patients with image-T2 disease, 10 (37.0%) were positive for lymph node metastasis. Of these, 3 had pN1 and 7 had pN2 disease. The nodes that were most likely to show metastasis in this subgroup were N12c (metastatic rate of 33.3%), followed by N13a (31.8%). Patients with image-T3 or -T4 disease had a 64.3% (36/56) risk of lymph node metastasis, and the risk of pN2 or greater was 42.9% (24/56). The nodes that were most likely to show metastasis in this subgroup were N12b (metastatic rate of 53.7%), followed by N12c (50.0%) and N13a (38.7%).
Because N12b, N12c, and N13a nodes are the most common sites of metastasis in node-positive and image-T2 to -T3 or -T4 cases of gallbladder carcinoma, these nodes were considered key nodes for lymphatic spread in this disease. The relationship between the positivity of these key nodes and pN is summarized in Figure 4. In image-T2 disease, pN0 was confirmed if N12b, N12c, and N13a nodes were negative for cancer. Because none of the patients with positive N12b and N12c nodes and negative N13a nodes had metastasis in other N2 nodes, these patients could be confirmed as having pN1 disease by checking only the key nodes. In image -T3 or -T4 disease, pN0 was also confirmed if N12b, N12c, and N13a nodes were negative for cancer. In patients with image-T3 or -T4 disease with positive N12b and N12c nodes, the risk of pN2 was as high as 77.8% (14/18). Only half (4/8) of the patients with N13a-negative nodes in this subgroup had pN1 disease.
The overall 1-, 3-, and 5-year survival of the 152 patients was 82.9%, 71.4%, and 62.7%, respectively. In a univariate analysis, pT, pN, TNM stage, T, histopathologically proven bile duct invasion, and histopathologic type (well or moderately differentiated vs poorly differentiated or other types) were significant prognostic factors for patient survival. Only sex was not significant (Table 7). The data regarding these 6 significant prognostic factors were used in a multivariate analysis, which identified 3 significant prognostic factors: pT, pN, and histopathologic type.
To predict lymph node metastasis (pN) in each case, a logistic regression analysis was performed using preoperative data: image-T, image-N, and macroscopic bile duct invasion. Of these 3 factors, only image-T and bile duct invasion were independently significant (P<.05). Image-N did not have significant predictive value for detecting lymph node metastasis.
Of the 33 patients who had lymph node metastases but no bile duct invasion, 18 underwent bile duct resection and 15 did not. There was no significant difference in pT classification between the 2 groups (P = .491). The decision regarding whether to resect the bile duct was made by the attending surgeon. There was no significant difference in patient survival between the 2 groups (Figure 5). Sites of recurrence in these 33 patients are summarized in Table 8. There was no apparent difference in the incidence or pattern of recurrence with regard to bile duct resection.
Of the 37 patients with pN2 metastasis, 15 underwent pancreaticoduodenectomy to achieve complete clearance of N2 nodes. Although patients who underwent pancreaticoduodenectomy had slightly better survival, the difference was not statistically significant (Figure 6A).
Extended lymph node dissection was carried out in 14 cases of pN2 disease. These patients had significantly better survival than those who underwent standard dissection (n = 23). The mean ± SEM 50% survival time in each group was 28.0 ± 1.8 and 13.9 ± 4.7 months. However, the 3-year survival in the extended dissection group was as low as 17.1% (Figure 6B). None of the patients with pN2 disease have survived for 5 years.
In accord with previous studies,3-7 the survival of patients with gallbladder cancer was affected by the depth of primary tumor invasion (pT classification) and by the presence of lymph node metastasis (pN classification) (Table 7 and Table 8). However, precise preoperative staging was difficult in almost half of the cases. The preoperative diagnosis of tumor invasion (image-T) was correct in only 52.6% of the cases, and this rate was lower in patients with pT1 and pT2 disease (Table 4),12 who may have a good chance for cure after appropriate surgical resection.19 For pT1 disease, 37.2% of cases in the present study were correctly staged, which was comparable with the findings of previous studies.4,10,20 Most early cancers (pT1 disease) are macroscopically flat and difficult to detect by diagnostic imaging.21 Using EUS, Mitake et al9 reported excellent diagnostic rates of 87.5% and 66.7% for pT1 and pT2 tumors, respectively. Although the diagnostic accuracy of EUS in the present study was similar or slightly inferior to their results, EUS was done only in selected cases, primarily patients with image-T1 or -T2 disease, and not in patients having benign disease. Diagnostic accuracies for T classification before and after 1990 were essentially the same (48.1% and 52.0%, respectively).
Because of the limitation of the spatial and contrast resolution, conventional magnetic resonance imaging cannot determine whether a tumor extends into serosa (pT3 or greater) or not (pT2).22 We have not routinely used magnetic resonance imaging for patients with gallbladder carcinoma. New magnetic resonance imaging techniques, such as dynamic magnetic resonance imaging using a surface coil,22 may provide more useful information for preoperative staging. Although magnetic resonance cholangiopancreatography is useful in detecting bile duct involvement or liver invasion in advanced cases (pT3 or greater),23 it provides little additional information for the distinction among pT1, pT2, and pT3 disease.
The preoperative detection of lymph node metastasis (image-N) was more difficult. Although the localization of visualized or enlarged lymph nodes is usually easy, the presence or absence of metastasis in each node can be difficult to determine. Positive nodes were correctly diagnosed in only 24.5% (12/49) of node-positive cases in the present series (Table 6). Few studies have addressed the sensitivity of diagnostic imaging for detecting lymph node involvement. Haribhakti et al24 reported a sensitivity of 50% using US. The multivariate analysis in the present study showed that image-T and presence of bile duct invasion had a stronger predictive value for the pN classification than image-N. A strong correlation has previously been reported between the T and N classifications.7
Because the precise preoperative staging of pT and pN can be difficult, it is important to establish surgical strategies based on information that is available before resection. Based on the results of the present study, we selected image-T (preoperative data) and histopathologic findings of key lymph nodes as the 2 determinants for our proposed algorithms (Figure 7 and Figure 8). Histopathologic data can be obtained from the intraoperative examination of frozen sections of key nodes. The preoperative diagnosis of lymph node metastasis (image-N) was excluded because of its low predictability.
Decision trees for the extent of lymph node dissection are shown in Figure 7. For patients with image-T1 tumor, the risk of pN1 is 4.0% and that of pN2 is 0%. N12b and N12c nodes should be checked first because they are known to be the initial sites of spread from gallbladder cancer.11,25,26 If the sampled N12c and N12b nodes are both negative for cancer in image-T1 disease, the risk of pN1 is 0% and no further dissection is necessary. If these nodes are positive, complete dissection of N1 nodes is adequate because the risk of pN2 is 0% in image-T1 cases.
Patients with image-T2 disease had a 37.0% (10/27) risk of lymph node metastasis, and the risk of pN2 was 25.9% (7/27). Dissection and intraoperative frozen section diagnosis should first be conducted for N12b and N12c nodes, followed by N13a. N13a is considered to be a point of lymphatic communication between the right and left channels in the hepatoduodenal ligament.26 If N12b, N12c, and N13a are negative for cancer, no further dissection is necessary because the risk of pN1 or greater is 0% in such instances. If N12b and N12c are positive and N13a is negative, complete dissection of N1 nodes is adequate because the risk of pN2 or greater is 0% in this subgroup. If N13a is positive, ie, pN2, standard dissection (clearance of N1, N8a, N8p, and N13a nodes) with or without pancreaticoduodenectomy should be tried if metastasis to other N2 nodes is not extensive. The risk of metastasis to other N2 nodes is 33.3% in such instances.
The risk of pN2 in patients with image-T3 or -T4 disease can be estimated in individual cases by sampling the key nodes. Overall, the probability of lymph node metastasis in patients with image-T3 or -T4 disease is as high as 64.3%. However, the risk of pN1 or greater is 0% if N12b, N12c, and N13a are negative for cancer. When N12b or N12c is positive and N13a is negative, the risk of only pN1 disease is 50.0%. Standard lymph node dissection will provide a chance for cure in patients with pN1 disease. Extended lymph node dissection is a controversial surgical challenge4,5,7,27 for pN2 disease. We cannot recommend extended dissection because there were no long-term survivors among this group in the present series (Figure 6B). However, because the 50% survival time in patients who underwent extended resection was more than twice that in the standard dissection group, extended dissection should receive further attention.
Decision trees for the extent of local resection were structured based on the algorithms for lymph node dissection (Figure 8). For patients with image-T1 tumor, the risk of understaging is 36.0% (9/25), but in no case is the understaging greater than pT2 (Table 4). Because the only positive nodes in patients with image-T1 disease are N12b and N12c, sampling and intraoperative examination of frozen sections of these key nodes are the next steps. If these nodes are negative for cancer, simple cholecystectomy is adequate because the risk of pN1 is 0%. Several studies4,5,28 have shown excellent results with cholecystectomy alone for pT1 disease. Eight patients with pT2 disease (image-T1) in the present series underwent simple cholecystectomy. There has been no recurrence among these patients. If the N12b or N12c nodes are positive, the patient has pT2 disease, for which liver bed resection is the most standard operation. Wedge resection of the liver bed with complete clearance of the subserosal layer of the gallbladder has been considered as providing a chance for cure.29
Patients with image-T2 tumor have a 14.8% risk of understaging (pT3 or pT4 tumor) (Table 4). Because the chance for pT1 is only 16.7%, even if the N12b, N12c, and N13a nodes are negative, liver bed resection is recommended as a standard procedure for pT2 disease.29 If the N12b and N12c nodes are positive and N13a is negative, there is no chance for early cancer. Liver bed resection is recommended in these cases. Because there is no risk of pT3, pT4, or pN2, there is a good chance for cure in this subgroup of patients. If N13a is positive, the risk of pT3 or pT4 is 50.0%, and a more extended liver resection may be required to achieve complete clearance of the tumor.
The diagnostic accuracy for image-T3 or -T4 cases is high, and the chance for overstaging (pT2 or less) is only 19.6% (Table 4). The risk of pN2 is 42.9% (24/56) in this subgroup. If the N13a node is negative for cancer, the risk of pN2 is 20.0% and extended liver resection to clear any local extension is worth a try in such cases.5 If necessary, resection of adjacent organs is also justified in patients with up to pN1 disease. If N13a is positive, ie, pN2, extended liver resection with concomitant pancreaticoduodenectomy is a controversial surgical challenge.30,31
Patients who have been preoperatively diagnosed as having benign disease were not included in our strategies. Carcinoma tends to be missed when gallstones are present.20,32 Preoperative US or EUS may be unable to describe the lesions clearly in patients with gallstones or cholecystitis.10 Notably, 50.0% (22/44) of such patients have pT2 or pT3 disease (Table 4), and 25.0% (11/44, data not shown) are positive for lymph node metastasis. The rate for the preoperative diagnosis of gallbladder cancer was 71.1% in the present series. This rate is high compared with previous studies,4,33 probably because of advances in diagnostic imaging and the nature of our institutions, which are tertiary referral centers. When intraoperative or postoperative histopathologic examination reveals a cancerous lesion in a specimen resected from a patient who was preoperatively diagnosed as having benign disease, the strategy for secondary radical intervention should follow the same algorithms.
Because there were no long-term survivors among patients with bile duct invasion (Table 7), bile duct invasion was not included in our strategy. Together with invasion to adjacent organs, bile duct invasion has been classified as noncurable by surgery.26 In accord with a previous study,34 survival rates were similar between patients with or without bile duct resection among those without bile duct invasion. Bile duct resection simply to facilitate lymph node dissection may not be necessary.
The algorithms proposed herein are a practically useful guide for choosing an appropriate extent of local resection and lymph node dissection for patients with up to pN1 disease without bile duct invasion. By following these algorithms, unnecessary lymph node dissection2 can be avoided in patients with up to stage IVa gallbladder cancer. The suitable strategy for stage IVb (pN2 or greater) gallbladder cancer is a controversial surgical challenge (Figure 7 and Figure 8). None of the patients with pN2 or greater disease in the present series survived for more than 5 years. These patients appear to be beyond the reach of curative surgical resection.19
In conclusion, precise preoperative staging for gallbladder carcinoma was difficult, despite advances in hepatobiliary imaging. The overall accuracy of image-T was 52.6%; however, image-T was a significant predictor of lymph node metastasis and patient outcome. We also identified key nodes for the lymphatic spread of gallbladder cancer: the cystic, pericholedochal, and posterosuperior peripancreatic nodes. By integrating image-T factor and data from the intraoperative examination of frozen sections of key lymph nodes, the most accurate staging before resection may be possible. Based on this staging, we proposed algorithms for the surgical treatment of gallbladder carcinoma. These algorithms are useful in patients with up to stage IVa disease. An extended lymph node dissection plus or minus an extended liver resection should be performed in some patients with more advanced disease. However, there was no survival advantage to more radical procedures, including bile duct resection or pancreaticoduodenectomy.
Corresponding author and reprints: Norihiro Kokudo, MD, Hepato-Biliary-Pancreatic Surgery Division, Department of Surgery, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan 113-8655 (e-mail: KOKUDO-2SU@h.u-tokyo.ac.jp).
Accepted for publication December 22, 2002.