Annals of Surgical Oncology 14(1):50–60
DOI: 10.1245/s10434-006-9136-3
Predictors for Patterns of Failure after Pancreaticoduodenectomy in Ampullary Cancer Hui-Ping Hsu, MD,1 Ta-Ming Yang, MD,2 Yu-Hsiang Hsieh, PhD,3 Yan-Shen Shan, MD, PhD,1 and Pin-Wen Lin, MD1
1
Department of Surgery, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, 138, Sheng-Li Road, Tainan, 70428, Taiwan (ROC) 2 Department of Surgery, Tainan Municipal Hospital, Tainan, Taiwan 3 Department of Emergency Medicine, The Johns Hopkins University School of Medicine, 1830 E. Monument Street, Suite 6-100, Baltimore, Maryland 21205, USA
Background: Ampullary cancer has the best prognosis in periampullary malignancy but unpredicted early recurrence after resection is frequent. The current study tried to find the predictors for recurrence to be used as determinative for postoperative adjuvant therapy. Methods: Information was collected from patients who underwent pancreaticoduodenectomy with regional lymphadenectomy for ampullary cancer in high-volume hospitals between January 1989 and April 2005. Recurrence patterns and survival rates were calculated and predictors were identified. Results: A total of 135 eligible patients were included. The 30-day operative mortality was 3%. Median followup for relapse-free patients was 52 months. Disease recurred in 57 (42%) patients, including 31 liver metastases, 26 locoregional recurrences, 9 peritoneal carcinomatoses, 7 bone metastases, and 6 other sites. Pancreatic invasion (P = 0.04) and tumor size (P = 0.05) were the predictors for locoregional recurrence, while lymph node metastasis was the sole predictor for liver metastasis (P = 0.01). The 5-year disease-specific survival rate was 45.7%; 77.7% for stage I, 28.5% for stage II, and 16.5% for stage III; and 63.7% for nodenegative versus 19.1% for node-positive patients. Pancreatic invasion and lymph node involvement were both predictors for survival of patients with ampullary cancer. Conclusion: Pancreaticoduodenectomy with regional lymphadenectomy is adequate for early-stage ampullary cancer but a dismal outcome can be predicted in patients with lymph node metastasis and pancreatic invasion. Lymph node metastasis and pancreatic invasion can be used to guide individualized, risk-oriented adjuvant therapy. Key Words: Ampulla of Vater—Pancreaticoduodenectomy.
ectomy is the main treatment.4,5 However, pancreaticoduodenectomy with regional lymphadenectomy is a technically demanding procedure with high rates of morbidity and mortality. In the older literature, the 5-year survival rate ranges from 6% to 38%.2,4 After improvements in preoperative evaluation, surgical technique, and postoperative care, the last ten years now report a 5-year survival rate of 35.0%-62.7%.6,7 Even after apparently successful operations, some patients still relapse earlier without apparent causes or clinical clues. In the literature, we found that even postoperative adjuvant chemoradiotherapy does not
Carcinoma of the ampulla of Vater is a rare neoplasm accounting for 0.063%-0.210% of all routine autopsy cases1 and 12.7%-32.2% of surgical resectable periampullary carcinomas.2,3 The resection rate of ampullary cancer is about 76.5%-88.0%2,4 and pancreaticoduodenectomy with regional lymphadenReceived June 16, 2006; accepted June 26, 2006; published online October 20, 2006. Address correspondence and reprint requests to: Yan-Shen Shan, MD, PhD; E-mail:
[email protected] Published by Springer Science+Business Media, Inc. 2006 The Society of Surgical Oncology, Inc.
50
PREDICTOR FOR RECURRENCE IN AMPULLARY CANCER
improve long-term survival after radical surgery. The available postoperative histopathologic factors seem not to be able to give a reliable prognosis.8,9 Because the procedure varies greatly among surgeons, and hospital volume also influences in-hospital mortality, analysis of the outcome of ampullary cancer in relation to the various histopathologic factors and other indicators is complicated. Therefore, we collected information on ampullary cancer patients who underwent pancreaticoduodenectomy with regional lymphadenectomy by volume surgeon from highvolume hospitals who used the same surgical procedure to exclude the surgical factor and analyzed the predictors for recurrence and dismal survival as the indicators for future postoperative adjuvant therapy to improve outcome.
PATIENTS AND METHODS Patients with ampullary cancer who underwent pancreaticoduodenectomy and regional lymphadenectomy between January 1989 to April 2005 in National Cheng Kung University Hospital and Tainan Municipal Hospital, both in Tainan, Taiwan, were included in the study. Patients who underwent palliative procedures or medical treatments were excluded. Age, gender, clinical presentation, laboratory findings on admission and before operation, preoperative biliary drainage procedures, type of resection, operative findings, postoperative morbidity, operative mortality, histopathologic findings, recurrent patterns, and survival were recorded from a retrospective chart review. Percutaneous transhepatic biliary drainage or endoscopic nasobiliary drainage was used routinely to relieve obstructive jaundice before operation if the bilirubin level was higher than 10 mg/dl. The use of either standard (SPD) or pylorus-preserving pancreaticoduodenectomy (PPPD) was determined by the surgeons. The surgical techniques have been described previously.10,11 In the SPD, the vagus nerves were preserved and the distal two-thirds of the stomach was transected. In PPPD, the right gastric artery was preserved, unless the artery restricted the mobility of the stomach. The duodenum was dissected and divided at least 2 cm distal to the pylorus. Reconstruction was accomplished by singlelayer end-to-side choledochojejunostomy without Ttube drainage and end-to-side pancreaticojejunostomy with appropriate pediatric nasogastric tube as a diversion stent in the pancreatic duct for three weeks. The resected lymph nodes included the anterior and posterior pancreaticoduodenal lymph nodes, nodes
51
along the right lateral aspect of the superior mesenteric artery and vein, and nodes in the lower hepatoduodenal ligament. If local invasion of the portal vein or superior mesenteric vein was encountered during the operation without further macroscopic distant metastasis, resection of the vein with end-toend reconstruction was performed. After completing reconstruction, one or two suction drains were placed through separate skin incisions in the right upper quadrant of the abdomen relative to the pancreatic and biliary anastomoses. During the operation, pancreatic consistency was evaluated by the surgeon. The diameters of the stumps of the pancreatic duct and common bile duct (CBD) were measured at the stump end. Operative time, intraoperative blood loss, and units used for blood transfusion were recorded precisely. Only eight patients received adjuvant postoperative chemoradiotherapy. Four patients received intravenous fluorouracil-based chemotherapy (fluorouracil/leucovorin + mitomycin or fluorouracil + epirubicin + mitomycin) and two patients took oral tegafur (Futraful, FT-207). Morbidity and Mortality All general and procedure-related complications were recorded. Pancreatic fistula was defined as amylase-rich fluid with drain fluid volume greater than 10 ml/day, persistent elevation of the drain amylase level, and three times higher than the serum level for longer than seven days.12 A biliary fistula was defined as the presence of bile drainage that persisted after postoperative day 7. Postoperative pancreatitis was defined as an increase in serum lipase with clinical presentation. Intra-abdominal abscess was defined as when drainage fluid was dirty with positive bacterial culture. Hemorrhagic complications included intra-abdominal bleeding after surgical dissection and gastrointestinal tract bleeding due to marginal ulcer. Delayed gastric emptying was defined as occurring when the nasogastric tube was left in place for ten days or more, plus one of the following: (1) emesis after removal of the nasogastric tube, (2) reinsertion of a nasogastric tube, or (3) failure to progress with diet.13 Wound infection was defined as a positive wound culture and the presence of pus necessitating opening of the wound. The pulmonary complications included pneumonia, pleural effusion, empyema, or respiratory failure. The operative mortality was defined as 30-day mortality, but hospital mortality included those who died from other causes associated with hospitalization, even more than 30 days after the operation. Ann. Surg. Oncol. Vol. 14, No. 1, 2007
52
H.P. HSU ET AL.
Histopathology All surgical specimens were checked by a pathologist. The site of the tumorÕs origin was determined at the time of gross and histolopathologic examinations. Only adenocarcinoma originating in the ampulla of Vater was included. Histologic differentiation was recorded as well, moderately, or poorly differentiated. Tumor size was measured from surgical specimen before formalin fixation. Tumor stage and TNM stage were defined according to the American Joint Committee on Cancer (AJCC) classification of 2002.14 Pancreatic invasion was defined as a tumor nest that invaded the parenchyma of the pancreas grossly or microscopically. The lymphovascular/perineural invasion or morphologic type of intestinal or pancreaticobiliary mucosa was not routinely recorded by a pathologist and was not included in study.
between the two groups was done by independentsample t test for continuous variables with normal distribution. Continuous variables that did not follow normal distribution were compared by a nonparametric two-independent-sample test. The association of categorical variables and survival was assessed using the Kaplan-Meier method, and significance was tested using the log-rank test. Those factors with P < 0.2 in univariate analysis were defined as significant. A multivariate analysis using Cox proportional hazards regression model was used to determine significant influence on survival. Risk factors associated with nonrecurrence or recurrence were determined by multivariate logistic regression. Each model included age and gender as covariates. The results were expressed as odds ratio (OR) with 95 percent confidence interval (CI) and corresponding two-tailed p values. Statistical significance was determined for P < 0.05.
Followup Followup at three-month intervals comprised physical examination and laboratory tests for tumor markers (carcinoembryonic antigen, CEA; cancer antigen 125 and 199, CA-125 and CA-199). Abdominal sonography was done every three months in the first year and then every six months in the second year. The computerized tomography of the abdomen was done annually or if there was suspicion of intra-abdomnal metastasis. Radiography of the thorax, bone scan, and computerized tomography of the brain were performed if clinical examination reported suspicion of metastasis. Recurrence was categorized as either locoregional or metastatic. Disease relapse was defined as biopsyproven disease or radiologic evidence of recurrence. Local recurrence was defined as recurrent retroperitoneal mass or regional nodes. Metastasis was defined as relapse of disease at a distant site, either a visceral organ or nonregional lymph nodes. The endpoint of the study was death and the primary criterion of followup was survival time. The overall survival rate was defined as the total survival ratio in these patients, including those who died from causes other than ampullary cancer. The disease-specific survival rate was limited to the effect of ampyllary cancer.
Statistical Analysis All statistical analyses were performed using SAS v9.13 (SAS Institute, Cary, NC). Univariate analysis was performed using the v2 test or FisherÕs exact test for categorical variables. Statistical comparison Ann. Surg. Oncol. Vol. 14, No. 1, 2007
RESULTS Demographic Data There were 146 patients diagnosed with adenocarcinoma of ampulla of Vater. A total of 135 patients who underwent pancreaticoduodenectomy with regional lymphadenectomy were included, the resection rate was 93%. Table 1. shows the demographic data comparing survivor and nonsurvivor patients. There were no statistical differences in age, gender, operative methods, stump of common bile duct, and consistency of pancreas between the two groups. Higher levels of preoperative bilirubin and tumor markers were noted in nonsurvivor patients (P < 0.05). Nonsurvivor patients had longer operative time, more blood loss and blood transfusion, and larger stump pancreatic duct size. Histopathologic Findings The nonsurvivors had more advanced histopathologic conditions, including larger tumor size, higher ratio of pancreatic invasion and lymph node involvement, poorer differentiation, and more patients with T3/T4 lesions or stage III/IV disease (Table 2). The median tumor size was 2.0 cm (range = 0.5-8.0 cm). The size of the primary tumor associated with pancreatic invasion was significantly larger than the sizes of those without pancreatic invasion (median = 2.9 cm, range = 1.0–8.0 cm with pancreatic invasion; median = 2.0 cm, range
53
PREDICTOR FOR RECURRENCE IN AMPULLARY CANCER
TABLE 1. Demographics, preoperative data, and operative findings in this series of 135 patients with ampullary cancer who underwent pancreaticoduodenectomy: results of univariate analysis Survivors a
No. of patients (%) Age, mean ± SD (range) Gender Male Female Total bilirubin (mg/dl), mean ± SD (range) Operative methods Whipple PPPD Operative time (min), mean ± SD (range) Blood loss (ml), mean ± SD (range) Blood transfusion (units), mean ± SD (range) Stump of pancreatic duct (mm), mean ± SD (range) Stump of common bile duct (mm), mean ± SD (range) Consistency of pancreas Soft Firm Hard CEA (U/dl), mean ± SD (range) CA-125 (U/dl), mean ± SD (range) CA-199, U/dl, mean ± SD (range) a
Nonsurvivors
P
64 (47%) 60 ± 11 (36–80)
60 (44%) 62 ± 12 (32–90)
33 (52%) 31 (48%) 3.7 ± 3.7 (0.4–16.5)
32 (53%) 28 (47%) 6.3 ± 5.8 (0.4–26.4)
0.49
33 (52%) 31 (48%) 275 ± 130 (150–765) 501 ± 326 (100–1700) 0.85 ± 1.48 (0–6) 3 ± 2 (1–10) 13 ± 5 (2–25)
39 (65%) 21 (35%) 312 ± 131 (180–735) 884 ± 701 (150–3100) 2.66 ± 2.47 (0–8) 4 ± 3 (1–15) 15 ± 6 (7–27)
0.09
24 (80%) 4 (13%) 2 (7%) 2.2 ± 2.3 (0.01–13.0) 14.8 ± 8.3 (0.5–35.1) 176 ± 384 (0–1860)
20 (65%) 7 (23%) 4 (13%) 14.2 ± 51.4 (0.1–296.3) 35.8 ± 41.7 (5.4–172.5) 580 ± 1393 (0.1–7513)
0.40
0.58
0.02
0.04 0.01 0.001 0.008 0.10
0.02 0.001 0.003
Excluding patients who died from causes other than ampullary cancer.
TABLE 2. Histopathologic findings in 135 patients with ampullary cancer who underwent pancreaticoduodenectomy: results of univariate analysis
a
No. of patients (%) Tumor size (cm), mean ± SD (range) Pancreatic invasion Negative Positive Tumor stage T1 T2 T3 T4 Lymph node status Negative Positive Differentiation Well Moderate Poor AJCC TNM stage Stage I Stage II Stage III Stage IV a b
Survivors
Nonsurvivors
P
64 (47%) 2.0 ± 1.8 (1 – 7)
60 (44%) 2.6 ± 1.5 (1 – 8)
0.04
50 (79%) 13 (21%)
24 (41%) 34 (59%)
< 0.0001
13 35 11 4
3 20 23 13
(5%) (34%) (39%) (22%)
< 0.0001
49 (78%) 14 (22%)
23 (38%) 37 (62%)
< 0.0001
29 (50%) 25 (43%) 4 (7%)
20 (35%) 28 (49%) 9 (16%)
0.008b
38 21 4 0
10 34 12 2
(21%) (56%) (17%) (6%)
(60%) (33%) (6%) (0%)
(17%) (59%) (21%) (3%)
< 0.0001
Excluding patients who died from causes other than ampullary cancer. The P value was the result of comparison of survival (log-rank test).
= 0.5–7.0 cm without pancreatic invasion; P = 0.001). Also, the primary tumor size was correlated with lymph node status (median = 2.0 cm in two groups; P = 0.114). However, there was no association between tumor size and histologic differentiation (median = 2.0 cm in well- and moderately
differentiated tumor and median = 1.75 cm in poorly differentiated tumors; P = 0.805). After further examination of the association of tumor size and clinical histopathology, the tumor size was not correlated with preoperative bilirubin level, diameter of pancreatic duct, operative time, intraoperative Ann. Surg. Oncol. Vol. 14, No. 1, 2007
54
H.P. HSU ET AL.
blood loss, or numbers of positive lymph nodes in graphs of scatterplots (data not shown). After pancreaticoduodenectomy, a median of 9 lymph nodes (range = 0–39) were removed from each patient. In survivors, a median of 7 lymph nodes (range = 1–39) were removed compared with a median of 10 lymph nodes (range = 0–25) in nonsurvivors without statistical significance (P = 0.137). The median number of lymph nodes involved was zero in survivors (range = 0–8) and one in nonsurvivors (range = 0–9) with P < 0.0001. The median number of dissected lymph nodes did not have any significant difference between patients with or without liver metastases (median = 8 vs. 10.5 lymph nodes; P = 0.094). However, the patients with liver metastases had a higher ratio of lymph node metastases than those without liver metastases (median = 0 vs. 2 lymph nodes; P = 0.001). The median number of dissected and involved lymph nodes did not have any significant difference between patients with or without locoregional recurrence (median = 8 vs. 10 dissected lymph nodes; P = 0.235; median = 0 vs. 0 involved lymph node; P = 0.163). Morbidity and Mortality Operative morbidity was 59% without any significant difference between SPD and PPPD (53% vs. 66%). There was more subjective delayed gastric emptying developed in PPPD patients. Incidence of pancreaticojejunostomy leakage was 10.7% (11 patients) in our series. In these 11 patients, the pancreaticojejunostomy leakage caused intra-abdominal fluid accumulation or abscess, which could be drained percutaneously, i.e., no surgical intervention needed. There was no mortality associated with pancreaticojejunostomy leakage in our series. Thirtyday operative mortality was 3% and hospital mortality was 6%. Mortality did not correlate with the operative method. In the patients with 30-day operative mortality or hospital mortality, the most frequent causes of mortality were pulmonary complications and intra-abdominal abscess. No one had evidence of pancreaticojejunostomy leakage. Recurrence Patterns Fifty-seven patients (42%) had disease recurrence during followup, including 31 (23%) liver metastases, 26 (19%) locoregional recurrences, 9 (7%) peritoneal carcinomatosis, 7 (5%) bone metastases, and 6 (4%) metastases at other sites (Table 3). There were 10 patients who developed liver Ann. Surg. Oncol. Vol. 14, No. 1, 2007
TABLE 3. Recurrence patterns in 135 patients with ampullary cancer who underwent pancreaticoduodenectomy No. patients (%) Liver metastasis Local recurrence Peritoneal carcinomatosis Bone metastasis Other metastasisa Total a
31 26 9 7 6 57
(23%) (19%) (7%) (5%) (4%) (42%)
Including brain, lung, and ovary metastases.
metastasis and locoregional recurrence simultaneously. In the univariate analysis of risk factors for recurrence, preoperative CA-125 level, diameter of the common bile duct stump, tumor size, pancreatic invasion, tumor stage, lymph node involvement, and AJCC TNM stage were statistically significant (data not shown). We further analyzed the risk factors for liver metastasis and local recurrence. In univariate analysis, liver metastasis was associated with preoperative bilirubin level and tumor marker, operative time and blood loss, diameter of the common bile duct stump, and lymph node involvement (Table 4). In multivariate analysis, only lymph node involvement had statistical significance to liver metastasis (Table 5). Patients with locoregional recurrence had different risk factors. Gender, tumor size, pancreatic invasion, tumor stage, and AJCC TNM stage had statistical significance in univariate analysis, but only pancreatic invasion was significant in multivariate analysis. Tumor size had borderline significance to locoregional recurrence (P = 0.05) (Tables 4 and 5). Survival During followup, 30 patients received chemoradiotherapy for recurrent disease. Six patients received intravenous fluororacil-based chemotherapy, four patients received gemcitabine-based chemotherapy, and two patients received another regimen. Only one patient received radiotherapy for locoregional recurrence. The overall 5-year survival rate in the series was 40.9% and the actuarial disease-specific 5-year survival rate was 45.7% (Table 6). There was no difference in the 5-year survival rate between patients with or without postoperative adjuvant therapy or between those with or without chemoradiotherapy for recurrent disease. For patients alive at the time of the followup study, the mean followup was 56 ± 39 months (median = 52 months, range = 5–145 months). Kaplan-Meier disease-specific survival
55
PREDICTOR FOR RECURRENCE IN AMPULLARY CANCER
TABLE 4. Univariate analysis of risk factors in different recurrence patterns in 135 patients with ampullary cancer who underwent pancreaticoduodenectomy Liver metastasis No. patients (%) a
Age (yr) Gender Female Male Total bilirubin (mg/dl)a GOT (unit)a CEA (U/dl)a CA-125 (U/dl)a CA-199 (U/dl)a Operative time (min)a Blood loss (ml) Stump of CBD (mm)a Tumor size (cm)a Pancreatic invasion Negative Positive Tumor stage T1 T2 T3 T4 Lymph node Negative Positive Differentiation Well Moderate Poor AJCC stage I II III IV a
Local recurrence P
No. patients (%)
P
60 ± 9
0.59
58 ± 11
0.23
16 (52%) 15 (48%) 5.2 ± 4.3 112 ± 85 27.2 ± 78.2 45.6 ± 47.4 1060 ± 1934 368 ± 159 946 ± 635 16 ± 7 2.5 ± 1.4
0.37
8 (31%) 18 (69%) 4.8 ± 4.5 121 ± 110 4.0 ± 6.5 30.6 ± 19.8 359 ± 640 303 ± 125 694 ± 466 15 ± 5 2.9 ± 1.8
0.04
0.01 0.77
