- Email: [email protected]
Pancreaticoduodenectomy for selected periampullary neuroendocrine tumors: Fifty patients
Pancreaticoduodenectomy for selected periampullary neuroendocrine tumors: Fifty patients Giao Q. Phan, BA, Charles J. Yeo, MD, John L. Cameron, Hruban, MD, and Robert Udelsman, MD, Baltimore, Md.
MD, Michael
M. Maher,
FRCS, Ralph
H.
Background.
Most resectable pancreatic or peripancreatic neuroendocrine tumors are treated by enuckation or distal pancreatectomy. A minority of tumors may require pancreaticoduodenectomy for complete tumor excision because of their large size, location, or lymph node involvement. Methods. This study reviews the management of 50 patients treated by pancreaticoduodenectomy for periampullary neuroendocm’ne tumors between 1962 and 1996 at a single institution. Results. There were 30 men and 20 women with a mean age of 52 f 2 years. Functional tumors were resected in 17patients: insulinoma, seven tumors; gastrinoma, eight tumors; vipoma, one tumor; and glucagonoma, one tumor Tumors were classified as malignant in 29patients and benign in 21. The median intraoperative blood loss was 800 ml, and the median number of units of blood transfused was zero. The postoperative length of stay was 20 + 2 days. Postoperative morbidity included 11 patients (24 %) with a pancreatic fistula and four patients (8 %) with a biliary fistula. There was one in-hospital death (2 %), in 1967. The actuarial survival rates at 2, 5, and 7 years are 81%, 73 Yo, and 65 %, respectively. Patients with benign tumors had a significantly improved 5year survival rate (94 %) compared with those with malignant tumors (61%; p = 0.03). Conclusions. Selected patients with periampullary neuroendocrine tumors can be managed successfully by pancreaticoduodenectomy, with low mortality and acceptable morbidity rates. (Surgery 1997;122:98997.) From the Departments of Surgery and Pathology, The Johns Hopkins Medical Institutions, Baltimore, Md.
NEUROENDOCBINE TUMORS OF THE pancreas are rare
entities, with an annual incidence of approximately five clinically recognized cases per one million people. The first report of an islet cell tumor of the pancreas was in 1902 by Nichol1s.l A pancreatic endocrine carcinoma was first reported in 1927 by Wilder et a1.,2describing a physician with an insulinoma that had infiltrated most of the pancreas and metastasized to the liver. The first surgical cure of a pancreatic neuroendocrine tumor was reported by Howland et a1.3in 1929, describing the enucleation of a benign insulinoma from the body of the pancreas. Although the first successful pancreatiSupported in part by a grant from the National Institutes of Health (ROlCA56130). Presented at the Eighteenth Annual Meeting of the American Association of Endocrine Surgeons, Baltimore, Md., April 6-8, 1997. Reprint requests: Charles J. Yeo, MD, Department of Surgery, The Johns Hopkins Hospital, Blalock 606, 600 N. Wolfe St., Baltimore, MD 21287-4606. Copyright 0 1997 by Mosby-Year Book, Inc. 0039-6060/97/$5.00 + 0 11/6/84659
coduodenectomy was performed in 1912 by Kausch,4 it was not until 1940 that Whipple5 performed the first pancreaticoduodenectomy for an islet cell carcinoma of the pancreas. After Zollinger and Ellison presented the triad of gastric hypersecretion, peptic ulceration, and non-pcell islet tumors in 1955, Oberhelman et al.’ in 1961 suggested pancreaticoduodenectomy as surgical therapy for duodenal gastrinomas. Although most pancreatic endocrine tumors can be removed by enucleation or distal pancreatectomy, some may require pancreaticoduodenectomy for complete resection and potential cure because of their large size, location, or lymph node involvement in the periampullary region. High morbidity and mortality rates associated with pancreaticoduodenectomy in the 195Os, 1960s and 1970s limited its widespread acceptance and led to suggestions that the operation be abandoned.8 Recent advances in surgical techniques and perioperative care have made pancreaticoduodenectomy safer, particularly at institutions with high voland extensive umes experience,g-13 thus expanding the indications for the procedure to SURGERY
989
Phan et al.
Surgery December 1997
Year of Pancreaticoduodenectomy
Fig. 1. Annual distribution of 50 pancreaticoduodenectomies performed for periampullary neuroendocrine tumors at The Johns Hopkins Hospital.
Table I. Clinical
presentation
No. of patients
Symptoms/signs
Functional tumors (n = 17; 35%) Gastrinoma Insulinoma VIPoma Glucagonoma Nonfunctional tumors (n = 32; 65%) Jaundice Abdominal pain Weight loss Pruritus Diarrhea Nausea/vomiting Anorexia Gastrointestinal bleeding
%
8 7 1 1
47 41 6 6
15 15 14 6 6 3 2 2
47 47 44 19 19 9 6 6
include selected patients with both benign and malignant periampullary diseases. This article reviews the intraoperative and postoperative management, pathologic elements, and survival of 50 consecutive patients who underwent pancreaticoduodenectomy for periampullary neuroendocrine tumors at The Johns Hopkins Hospital. PATIENTS
AND
METHODS
Patient selection and follow-up. Between May 1962 and November 1996, inclusive, 50 patients underwent pancreaticoduodenectomy for pathologically proven periampullary neuroendocrine tumors at The Johns Hopkins Hospital; 42 of these patients underwent resection between 1987 and 1996. From 1991 to 1996, data were recordefi prospectively; for patients operated on before 1991, pathology records were examined retrospec-
tively to 1949 to identify these patients, with 1962 being the first year a pancreaticoduodenectomy was performed for a neuroendocrine tumor. Medical records were reviewed, and follow-up was obtained by review of office records or through telephone contact during October and November 1996. Five patients were lost to follow-up and were excluded from survival analyses. Of note, previous reports from this institution have included some of these patients: Udelsman et a1.14 presented 12, Yeo et aLI described 10, and Cameron et a1.g discussed four. Diagnostic criteria. Patients with appropriate signs, symptoms, and laboratory evidence of hormonal excess were classified into their respective clinical syndromes: insulinoma (Whipple’s triad, neuroglycopenia, and catecholamine surge), gastrinoma (peptic ulcer, diarrhea, and esophagitis) , vipoma (watery diarrhea, hypokalemia, and achlorhydria), and glucagonoma (hyperglycemia, necrolytic migratory erythema, and hypoproteinemia). Tumors of patients without a recognizable clinical syndrome and with normal serum hormone levels (with the exception of pancreatic polypeptide) were classified as nonfunctioning, independent of the results of the immunoperoxidase stains of the tumor specimen. Tumor characteristics. The site of the primary tumor was determined by the operative note or pathology report. Malignancy was determined by the presence of metastases to lymph nodes or the liver at operation or by the development of such metastases during follow-up. In cases in which the primary tumor was not found but pathology reports showed lymph node involvement (as occurred in two patients with Zollinger-Ellison syndrome), the tumor was classified as metastatic. Margin status and tumor diameter were determined from the pathology report. Postoperative course. Complications were defined by the following published criteria16: delayed gastric emptying (need for postoperative nasogastric decompression for more than 10 days), pancreatic flstula (daily drainage of 50 ml or more of amylase-rich fluid after postoperative day 7), intraabdominal abscess (radiographically defined or surgically found fluid collection associated with a septic clinical picture), and wound infection (need for the wound to be partially or wholly opened and packed). Statistical analyses. Survival and prognostic indexes were calculated by the Kaplan-Meier method. Differences among subsets were compared according to the log-rank test. Significance was accepted at the 5% level. Data are presented as
Phan et al. 991
surgery
Volume122, Number 6 Table
II. Intraoperative
factors
Type of resection
17 Classic (with hemigastrectomy) 33 Pylorus preserving
Estimated blood loss
Range 200-28,000 cc Median 800 cc Mean 1610 i- 570 cc
Packed red cell transfusion
Range O-35 units Median 0 units Mean 1.7 * 0.7 units
Operative
Range 3.9-13.4 hr Median 6.8 hr Mean 6.9 f 0.3 hr
time
Type of anastomosis
36 Pancreaticojejunostomies 7 Pancreaticogastrostomies 4 None (3 with concomitant pancreatectomy) 3 Undeterminable
total pancreatectomy,
1 with previous distal
Synchronous liver resection
3 For nonfunctional
tumors
Vessel resection
1 superior mesenteric vein resection for nonfunctional
the median or mean f standard error of the mean. Statistical computations were performed with either SPSS for Windows, version 7.0, or Core1 Quattro Pro 7. RESULTS Patient demographics. The study cohort consisted of 30 men (60%) and 20 women (40%), with ages ranging from 27 to 38 years and a mean age of 52 * 2 years. Forty-three patients (86%) were white, four (8%) were African-American, and three (6%) were of another race. The annual distribution of these patients is depicted in Fig. 1. Clinical presentation. Functional tumors were found in 17 patients (35%), whereas nonfunctional tumors were found in 32 patients (65%) ; the functional status of one patient’s tumor was not determinable (Table I). The most common functional tumor was gastrinoma, which was seen in eight patients; there were also seven insulinomas, one vipoma, and one glucagonoma. No patients with carcinoid syndrome were identified in this series. The majority (84%) of patients with nonfunctional tumors had jaundice, abdominal pain, or weight loss, similar to patients with nonneuroendocrine periampullary tumors. Intraoperative factors. Of the 50 pancreaticoduodenectomies, 1’7 were classic resections with hemigastrectomy and 33 were pylorus preserving
tumor
(Table II). Estimated blood loss ranged from 200 to 28,000 cc (in a patient with a portal vein injury), with a median of 800 cc and a mean of 1610 f 570 cc. The number of units of red blood cells transfused ranged from 0 to 35 (in the same patient with the portal vein injury), with a median of zero units and a mean of 1.7 f 0.7 units. The operative time ranged from 3.9 to 13.4 hours, with a median of 6.8 hours and a mean of 6.9 f 0.3 hours. The type of pancreatic-enteric anastomosis performed was a pancreaticojejunostomy in 36 patients and a pancreaticogastrostomy in seven; four patients required neither anastomosis (three patients underwent total pancreatectomies and one had undergone a previous distal pancreatectomy) ; for three patients the type of pancreaticenteric anastomosis constructed was undeterminable. Three patients underwent synchronous hepatic resection for nonfunctional tumors metastatic to the liver. Superior mesenteric vein resection was required in one patient with a nonfunctional tumor. Six patients with clinical and biochemical evidence of Zollinger-Ellison syndrome without definitive preoperative or intraoperative tumor identification underwent “blind” classic pancreaticoduodenal resection (Table III); all were eugastrinemic at last follow-up. Tumor characteristic. Malignant neoplasms were found in 29 patients (580/o), whereas benign
992
Phan et al.
~Surgery December 1997
.I 0.0
.I 0
12
24
36
48
60
72
I 84
0.0 0
12
24
Months
48
60
72
4
Months
Fig. 2. Actuarial survival curve after pancreaticoduodenectomy (n = 45 evaluable patients). Two-, 5-, and T-year survival rates were Sl%, 73%, and 65%, respectively.
Fig. 4. Actuarial survival curves after pancreaticoduodenectomy, comparing benign tumors (n = 18; solid line; median survival not reached) and malignant tumors (12= 27; dotted line; median survival 71 months). Curves are significantly different, with p = 0.03.
aged 4.5 f 1.1 cm, with a median of 3.0 cm. Malignant tumors averaged 3.2 * 0.4 cm, with a median of 3.0 cm. The site of the primary tumor was in the head, neck, or uncinate process of the pancreas in 39 patients (78%), the duodenum in six (12%), and the distal common bile duct in one (2%). The site
.a i .7 .6
36
i
5 1 .4 i 3
could
2
patients because the pancreatic tumor was too large (18 cm) and in two patients because tumor was found only in peripancreatic lymph nodes. Postoperative course. The postoperative length of stay ranged from 1 to 86 days, with a median of 15 days and a mean of 20 f 2 days. Seventeen patients (37%) were discharged home without in-hospital complications. Twenty-nine patients (63%) had some form of complications (Table V) ; the existence of complications could not be determined in four patients because of lack of adequate clinical information. The most common complications were wound infection (11 patients, 24%)) pancreatic fistula (11 patients, 24%), and delayed gastric emptying (seven patients, 15%). Three patients
.I1
0.01 0
12
24
36
48
60
72
84
Months Fig. 3. Actuarial survival curves after pancreaticoduodenectomy, comparing functional tumors (n = 15; solid line; median survival not reached) and nonfunctional tumors (n = 30; dotted line; median survival not reached) (difference not significant).
tumors were found in 21 (42%) (Table IV). Thirteen (76%) of the functional tumors were malignant, whereas 16 (50%) of the nonfunctional tumors were malignant. Of the functional tumors, all eight gastrinomas, the vipoma, the glucagonoma, and three of the insulinomas were malignant; only the remaining four insulinomas were benign. The mean diameter of the primary tumor for the entire cohort was 3.8 f 0.5 cm, with a median of 3.0 cm. The size of the functional tumors averaged 2.5 + 0.5 cm, with a median of 1.8 cm. Nonfunctional tumors were generally larger, with a mean of 4.3 * 0.7 cm and a median of 3.5 cm. Benign tumors aver-
not be determined
in four
patients:
in two
(6%) required reoperation: one for a pancreaticojejunostomy leak, one for repair of a fascial dehiscence, and one for repair of a gastrocutaneous fis-
tula. One patient (2%) with a benign insulinoma died of postoperative hemorrhage on postoperative day 1 as the result of a classic pancreaticoduodenectomy performed in 1967. There have been no postoperative in-hospital deaths in 48 consecutive pancreaticoduodenectomies for neuroendocrine tumors at The Johns Hopkins Hospital since 1967.
Phan
surgery
et al.
993
Volume 122, Number 6
Table III. Blind classic pancreaticoduodenectomy Patient
Gas&in
Preoperative
(Postoperative)
Preoperative
for Zollinger-Ellison studies
syndrome
Intraoperative
U/S
Pathology
F/U
0.5 cm Duodenum +1/6 LNs (-) margins
Alive 41 mo
0.3 cm Duodenum t1/9 LNs (-) margins
Died second cardiac arrest 15 mo
1.7 cm Neck +3/7 LNs (-) margins
Alive 44 mo
UGI: fundic mass Not used CT: no mass AbIG: 1.5 cm HOP mass SSRI, intense activity near gallbladder fossa
1.0 cm Duodenum t1/23 LNs (-) margins
Alive 1 mo
(39)
CT: no mass ANG: tumor blush in HOP area
Not used
Primary not found t4/29 LNs (-) margins
Alive 36 mo
(40)
UGI: wide duodenal Not used c loop EGD: duodenitis; gastritis Endoscopic U/S: no mass CT: no mass ANG: no tumor blush ERCP: normal
Primary not found
Alive 30 mo
UGI: duodenal ulcer EGD: esophagitis CT: hyperdense HOP ANG: no tumor blush
No mass identified
J.D.
Fasting PSS
571 875
(92) (129)
G.J.*
Fasting PSS
250 500
(45)
EGD: marginal ulcers Not used CT: no mass ANG: no tumor blush THPVS: gastrin step-up from HOP/duodenal area
J.M.”
Fasting PSS
800 2800
(41)
UGI: normal CT: 1.5 cm mass in neck of pancreas ANG: no tumor blush
E.O.
Fasting PSS
2520 3200
(40)
RR.
Fasting PSS
658 980
A.R.
Fasting PSS
475 t
No mass identified
+l/ll LNs (-) margins
U/S, Ultrasonography; F/U, follow-up; UGI, upper gastrointestinal series; PSS, post-secretin stimulation; EGD, esophagogastroduodenoscopy; CT, computed tomography; HOE head of pancreas; LNs, lymph nodes; ANG, selective mesenteric angiography; THPVS, transhepatic portal vencw sampling; SSEU, somatostatin receptor imaging; ERCE: endoscopic retrograde cholangiopancreatography. *Patients with history of antrectomy.
Postdischarge follow-up revealed that 11 patients (22%) had late complications, seven of which required operation: two for small bowel obstruction, four for incisional hernia, and one for gastric outlet obstruction. Other late complications included two patients with small bowel obstruction managed medically, three patients with late-identified wound infection, and two patients with intraabdominal abscessmanaged percutaneously. Survival and outcome. Follow-up ranged from 1 to 177 months (mean, 42 k 6 months). Thirty-five patients (78%) were alive at last follow-up, whereas 10 patients (22%) have died; five patients were lost to follow-up. Of the 10 deaths, seven patients died of progression of tumor, one of cardiac arrest, one of an unknown cause, and one in hospital of postoperative hemorrhage.
The overall actuarial survival rates at 2, 5, and 7 years were Sl%, 73%, and 65%, respectively (Fig. 2). The 2-, 5-, and ‘I-year survival rates for patients with functional tumors were 86%, 86%, and 71%, respectively, compared with 78%, 66%, and 66%, respectively, for patients with nonfunctional tumors (difference not significant) (Fig. 3). The 2, 5-, and 7-year survival rates for patients with benign tumors were significantly better (94%, 94%, and 94%) than they were for patients with malignant tumors (73%, 61%, and 41%, respectively; $I = 0.03) (Fig. 4). A univariate analysis of the 27 evaluable patients with malignant tumors was performed to assessthe role of various factors in influencing long-term survival (Table VI). In this analysis, factors such as estimated blood loss, operative time, tumor diameter,
994 Phan et al.
surgzry December 1997
Table IV. Tumor
characteristics
Malignancy 29 (58%) Malignant 20 (69%) Lymph nodes metastases only 4 (14%) Liver metastases only 5 (17%) Both lymph nodes and liver metastases 21 (42%) Benign Diameter of primary tumor (cm) Mean Median Entire group (n = 46) 3.8 + 0.5 3.0 Functional (n = 14) 2.5 + 0.5 1.8 Nonfunctional (n = 32) 4.3 * 0.7” 3.5 Benign (n = 19) 4.5 f 1.1 3.0 Malignant (n = 27) 3.2 + 0.4 3.0 Site of primary tumor HOP/neck/uncinate 39 Entire pancreas 2 Duodenum 6 Distal CBD 1 2 (primary tumor not found) Lymph nodes only HO8 Head of pancreas; CM, common *p = 0.10 compared
with functional
Table V. Postoperative
bile duct.
tumors by analysis of variance.
complications
Complication
DISCUSSION
No. ofpatients
%
17 11 11 7 4 3 3 2 1 1 1 1
37 24 24 15 9 7 7 4 2 2 2 2
None Wound infection Pancreatic fistula (1 re-op) Delayed gastric emptying Biliary fistula Intraabdominal abscess Pleural effusion Sepsis Fascial dehiscence (re-op) Gastrocutaneous fismla (re-op) Perioperative myocardial infarction Postoperative hemorrhage (death) Re-op, Reoperation.
postoperative tive length
complication
the tumor
red cell blood
0.05); patients a median
correlated resection
status of
survival.
transfusions whereas
Both
Patients
had a median
those not receiving
did not reach the median
survival
(p =
who had nonfunctional
tumors
had
survival
had functional survival
affected
of 32 months,
transfusions
survival.
status and the functional
significantly
receiving survival
status, and postopera-
of stay did not influence
the transfusion
of 43 months, tumors
(p = 0.05).
whereas
did not reach
Resection
margin
with survival;
patients
margins
a median
months,
compared
negative
margins
had
Range 0.3-18.0 0.3-8.0 0.6-18.0 0.5-18.0 0.3-8.0
with 71 months (fi = 0.07).
those who the median status also
who had positive survival in patients
of 13 with
In 1968 Howard17 published a personal series of 41 consecutive pancreaticoduodenectomies without an operative death. More recent reports have documented even larger consecutive series without deaths.ga12 Although pancreaticoduodenal resection remains a formidable procedure, there is a growing consensus that it is indicated in appropriately selected patients with periampullary adenocarcinoma, as well as selected patients with benign diseases18 such as chronic pancreatitis and less aggressive neoplasms such as neuroendocrine tumors.14 In this series the postoperative mortality rate was an acceptable 2%. There have been no postoperative deaths in 48 patients who have undergone morbidity rates resection since 1967. Postoperative have continued to be high (630/o), but most of the complications were not life-threatening and were managed nonoperatively. The relatively high rate of pancreatic fistula (24%) may be attributable to the fact that only one patient (2%) had a history of chronic pancreatitis, and the majority of patients had a soft, nonfibrotic pancreas at the site of pancreatic transection, which is known to be associated with an increased rate of anastomotic leakage.lg The relatively low rate of delayed gastric emptying observed in this series (15%) may be a reflection of our routine prophylactic use of the prokinetic agent erythromycin.16 Although neuroendocrine tumors are more indolent than pancreatic exocrine adenocarcinoma, they can cause significant morbidity and death.
Phan et al.
surgery
995
Volume 122, Number 6
Most of the nonfunctional tumors in this series caused significant obstructive and constitutional symptoms, as evidenced by the fact that 84% of the patients had jaundice, abdominal pain, or weight loss. All patients with functional tumors had symptoms of the corresponding hormone at presentation. At the time of operation the decision to proceed to pancreaticoduodenectomy was made cautiously and only after careful assessment of the resection options. There is undoubtedly a role for less aggressive resection in selected patients with periampullary neuroendocrine neoplasms. For example, we have performed local excision of benign duodenal gastrinomas, and we have enucleated benign insulinomas from the region of the head and uncinate process of the pancreas. Such local procedures are generally reserved for small tumors, without evidence of nodal metastases. In this series the mean tumor diameter was 3.8 f 0.5 cm (median, 3.0 cm), and 58% of all tumors were malignant, with 86% of the malignant tumors having lymph node metastases. Even in the group of six patients treated by blind classic pancreaticoduodenectomy (Table III), in which three of the tumors were found by the pathologist to be 1.0 cm or less in the duodenum, all patients had nodal metastases in the resected specimen. These nodal metastases could have been left unresected if a local resection of the duodenal primary tumor had been performed. Additional support for the complete excision of malignant tumors comes from the data (Table VI) comparing median survival between patients who had positive margins (13 months) with those patients who had negative margins (71 months). In fact, many studies have concluded similarly that aggressive resection of malignant neuroendocrine tumors significantly improves long-term survival.20~2’ The six patients with Zollinger-Ellison syndrome for which a blind pancreaticoduodenectomy was performed had biochemical evidence of and symptoms from a gastrinoma. Their symptoms included abdominal pain, diarrhea, peptic ulcer disease, and multiple gastrointestinal bleeds; two patients had previously undergone partial gastrectomy and vagotomy without symptomatic relief. Preoperative imaging studies failed to localize these tumors definitively. When a mass was not found after exploring the pancreas and duodenum, intraoperative ultrasonography was used in two cases, both without success in locating the suspected tumor. In all cases the decision to perform a blind pancreaticoduodenectomy was made on strong clinical suspicion that the tumor was in the gastrinoma triangle. 22 In five of the six cases at least one
Table VI. Univariate survival analysis for malignant tumors (fi = 27) Parameter
Median
EBL Z300 ml <800 ml
NR
Transfusion Yes No Operative time 26.8 hr ~6.8 hr
Tumor function Nonfunctional Functional Tumor diameter 23.0 cm C3.0 cm
71
Survival (mo) Mean p Value
83 + 23 47 5 7
0.43
66 i 22 51+5
0.05
NR NR NR
104 i 25 43 f 7
0.82
43 NR
37 + 6 128 + 29
0.05
71
92 * 30 78 + 25
0.37
43 13 71
26 + 9 105 rt 27
0.07
NR 71
113 f 22 53 zk10
0.81
43 71
88 f 27 58 + 7
0.36
32
Surgical margins Complications Yes No Postoperative LOS 215 days ~15 days
EBL, Estimated blood loss; NT?, not reached; LOS, length of stay.
preoperative study had suggested that the tumor was localized to the head of the pancreas or the duodenum. In the remaining case, intraoperative lymph node biopsy in the gastrinoma triangle revealed metastatic neuroendocrine tumor. Fortunately, these blind resections have rendered all of these patients symptom free and eugastrinemic at last follow-up. Because this series encompasses a 35-year period, much has evolved in the management of patients with neuroendocrine neoplasms. Preoperative imaging modalities such as dynamic thin-slice contrast-enhanced computed tomography, endoscopic ultrasonography, octreotide scanning, and selective arterial secretin stimulation testing have improved the ability to localize these tumors before laparotomy. In addition, intraoperative techniques including ultrasonography and duodenal transillumination further improve tumor localization. None of these techniques was available 20 years ago. In summary, this single-institution experience documents a low mortality rate and acceptable morbidity rate in a highly selected group of 50 patients treated by pancreaticoduodenectomy for
996
Phan et al.
periampullary neuroendocrine tumors. For the entire cohort the 5-year actuarial survival rate was 73%. Survival analysis of the patients with malignant tumors indicated that three factors correlated with decreased survival: blood transfusion,23a24 nonfunctional tumor status,25 and the presence of positive surgical margins. REFERENCES 1. Nicholls AG. Simple adenoma of the pancreas arising from an island of Langerhans. J Med Res 1902;8:385-95. 2. Wilder RM, Allan FN, Power MH, Robertson HE. Carcinoma of the islands of the pancreas: hyperinsulinism and hypoglycemia. JAMA 1927;89:348-55. 3. Howland G, Campbell WR, Maltby EJ, Robinson WL. Dysinsulism: convulsions and coma due to islet cell tumor of the pancreas, with operation and cure. JAMA 1929;93:6749. 4. Kausch W. Das carcinom der papilla duodeni und seine radikale entfernung. Beitr Klin Chir 1912;78:439-86. 5. Whipple AO. Pancreaticoduodenectomy for islet cell carcinoma: a five-year follow-up. Ann Surg 1945;121:847-52. 6. Zollinger RM, Ellison EH. Primary peptic ulcerations of the jejunum associated with islet cell tumors of the pancreas. Ann Surg 1955;142:709-28. 7. Oberhelman HA Jr, Nelsen TS, Johnson AN Jr, Dragstedt LR II. Ulcerogenic tumors of the duodenum. Ann Surg 1961;153:21427. 8. Crile G Jr. The advantages of bypass operations over radical pancreatoduodenectomy in the treatment of pancreatic cancer. Surg Gynecol Obstet 1970;130:1049-53. 9. Cameron JL, Pitt HA, Yeo CJ, Lillemoe KD, Kaufman HS, Coleman J. One hundred fort@ve consecutive pancreaticoduodenectomies without mortality. Ann Surg 1993;217:430-8. 10. Miedema BW, Sarr MG, van Heerden JA, Nagorney DM, McIlrath DC, Ilstrup D. Complications following pancreaticoduodenectomy: current management. Arch Surg 1992;127:945-50. 11. Delcore R, Friesen SR. Role of pancreatoduodenectomy in the management of primary duodenal wall gastrinomas in patients with Zollinger-Ellison syndrome. Surgery 1992;112:1016-23. 12. Trede M, Schwa11 G, Saeger H-D. Survival after pancreatoduodenectomy: 118 consecutive patients without an operative mortality. Ann Surg 1990;211:447-58. 13. Gordon TA, Burleyson GP, Tielsch JM, Cameron JL. The effects of regionalization on cost and outcome for one general high-risk surgical procedure. Ann Surg 1995;221:43-9. 14. Udelsman R, Yeo CJ, Hruban RH, Pitt HA, NiederhuberJE, Coleman J, et al. Pancreaticoduodenectomy for selected pancreatic endocrine tumors. Surg Gynecol Obstet 1993;177:269-78. 15. Yeo CJ, Wang BH, Anthone GJ, Cameron JL. Surgical experience with pancreatic islet-cell tumors. Arch Surg 1993;128:1143-8. 16. Yeo CJ. Management of complications following pancreaticoduodenectomy. Surg Clin North Am 1995;75:913-24. 17. Howard JM. Pancreatico-duodenectomy: forty-one consecutive Whipple resections without an operative mortality. Ann Surg 1968;168:629-40. 18. Barnes SA, Lillemoe KD, Kaufman HS, Sauter PK, Yeo CJ, Talamini MA, et al. Pancreaticoduodenectomy for benign disease. Am J Surg 1996;171:131-5.
surgery December 1997 19. Yeo CJ, Cameron JL, Maher MM, Sauter PK, Zahurak ML, Talamini MA, et al. A prospective randomized trial of pancreaticogastrostomy versus pancreaticojejunostomy following pancreaticoduodenectomy. Ann Surg 1995;222:580-92. 20. Legaspi A, Brennan ME Management of islet carcinoma. Surgery 1988;104:1018-23. 21. Carty SE, Jensen RT, Norton JA. Prospective study of aggressive resection of metastatic pancreatic endocrine tumors. Surgery 1992;112:102432. 22. Stabile BE, Morrow DJ, Passaro E Jr. The gastrinoma triangle: operative implications. Am J Surg 1984;147:25-31. 23. Cameron JL, Crist DW, SitzmannJV, Hruban RH, Boitnott JI3, Siedler AJ, et al. Factors influencing survival following pancreaticoduodenectomy for pancreatic cancer. Am J Surg 1991;161:120-5. 24. Wu H, Little A. Perioperative blood transfusions and cancer recurrence. J Clin Oncol 1988;6:1348-54. 25. Eckhauser FE, Cheung PS, Vinik AI, Strode1 WE, Lloyd RV, Thompson NW. Nonfunctioning malignant neuroendocrine tumors of the pancreas. Surgery 1986;100:978-88. DISCUSSION Dr. Richard Prinz (Chicago, Ill.). This is a 35-year review, yet 42 of the patients have been seen in the past
9 years. During this g-year period how many patients with neuroendocrine tumors of the pancreas with distal resections or enucleations?
were treated Focusing on
lesions in the head, do you treat anybody with enucleation in Baltimore? Current treatment of small gastrinomas in the duodenum and the head of the pancreas is enucleation and a nodal dissection. Have you done that procedure? If so, what were your results? Are poor results with this approach the reason you have turned to doing pancreaticoduodenectomy? Positive nodes were found in all of your resections for Zollinger-Ellison syndrome. You claim that all of these patients were eugastrinemic afterward. Have you done secretin stimulation tests in addition to basal serum gas&in levels to prove that? Not all of our patients have become eugastrinemic after enucleation and node dissection. Most studies with other types of tumors, such as colon cancer and even pancreatic adenocarcinoma, showed that survival is better ifyou do not have to transfuse patients. Are you showing the opposite, or am I misreading your results? Dr. Yeo. We actually perform operations in Baltimore other than pancreaticoduodenectomy; we do distal pancreatectomies and enucleations. Obviously, much of this pancreaticoduodenectomy experience is related to our referral patients
basis, being a tertiary-care center, and many are sent to us who require pancreaticoduo-
denectomy. For example, during the same period we reviewed 125 islet cell tumor resections and performed 50 pancreaticoduodenectomies, 39 distal pancreatectomies,
and 21 enucleations.
I think
a lot of the bias is
related to the referral pattern. As to the issue of gastrinomas,
in this particular
series we have eight
gastrino-
mas. We do perform enucleations and nodal dissections as well. We have done fewer of those than Whipple procedures for gastrinoma. For the “blind” resections we have one patient who is eugastrinemic who has a negative secretin stimulation, and the other four are eugastrinemic and are apparently cured in relatively short-
Phan et al.
surgery Volume 122, Ahmber 6 term follow-up. As to the blood transfusion data, there is better survival when the patient received no blood transfusions during the operation. Dr. Peter Angeles (Chicago, Ill.). I am curious in particular about the two patients who had nodal disease only. How was the decision made to do a pancreaticoduodenectomy in these patients in whom only nodal disease was found? Dr. Yeo. Both of the patients with nodal disease only were patients with gastrinoma. One had undergone a previous vagotomy and antrectomy at another institution. We never received the pathology slides, and we suspected that residual disease was making the patient hypergastrinemic. In the other patient we never found the primary tumor. Perhaps the primary tumor was just too small to be localized by our pathologists. However, that patient also remained eugastrinemic and symptom free at long-term follow-up. Dr. Geoffrey B. Thompson (Rochester, Minn.). Did you look at nodal status and what effect it had on overall prognosis? Were any of the Whipple procedures performed for Zollinger-Ellison syndrome in patients with multiple endocrine neoplasia type 1 (MEN l)? If so, what were the long-term results in that particular subgroup? One of the main concerns about doing a Whipple pro-
997
cedure for islet cell tumors is that the surrounding pancreas is normal and often very soft, making pancreaticojejunostomy a major concern. Are you doing anything special for these patients, such as using octreotide (Sandostatin) before and after the operation or fibrin glue during the operation? Dr. Yeo. The nodal status of the patients was not predictive of survival because nodal status was used to define the malignant state. Nodal status is not a significant variable in the univariate analysis. None of these Whipple procedures was done for patients with MEN 1, although we have operated on a handful of patients with MEN 1 and gastrinomas. We typically do a distal gastrectomy and enucleate the lesions from the duodenum. We have a 24% fistula rate, which is quite high. Most patients had normal glands at the pancreatic neck, and the 24% fistula rate is by a very strict definition of fistula in patients with normal glands. We now use octreotide prophylactically for a soft pancreas, although I am still not convinced that it is as useful as is touted. Only one of the 11 patients who had a fistula required reexploration for completion pancreatectomy. But there definitely is a higher risk of pancreatic fistula in these patients than in those with pancreatic adenocarcinoma.
ANNOUNCEMENT Effective October 1, 199’7, please send all manuscripts and other submissions for Surguy to the following address: Kim Fons Managing Editor surgery 11830 Westline Industrial Drive St. Louis, MO 63146-3318
MD, Michael
M. Maher,
FRCS, Ralph
H.
Background.
Most resectable pancreatic or peripancreatic neuroendocrine tumors are treated by enuckation or distal pancreatectomy. A minority of tumors may require pancreaticoduodenectomy for complete tumor excision because of their large size, location, or lymph node involvement. Methods. This study reviews the management of 50 patients treated by pancreaticoduodenectomy for periampullary neuroendocm’ne tumors between 1962 and 1996 at a single institution. Results. There were 30 men and 20 women with a mean age of 52 f 2 years. Functional tumors were resected in 17patients: insulinoma, seven tumors; gastrinoma, eight tumors; vipoma, one tumor; and glucagonoma, one tumor Tumors were classified as malignant in 29patients and benign in 21. The median intraoperative blood loss was 800 ml, and the median number of units of blood transfused was zero. The postoperative length of stay was 20 + 2 days. Postoperative morbidity included 11 patients (24 %) with a pancreatic fistula and four patients (8 %) with a biliary fistula. There was one in-hospital death (2 %), in 1967. The actuarial survival rates at 2, 5, and 7 years are 81%, 73 Yo, and 65 %, respectively. Patients with benign tumors had a significantly improved 5year survival rate (94 %) compared with those with malignant tumors (61%; p = 0.03). Conclusions. Selected patients with periampullary neuroendocrine tumors can be managed successfully by pancreaticoduodenectomy, with low mortality and acceptable morbidity rates. (Surgery 1997;122:98997.) From the Departments of Surgery and Pathology, The Johns Hopkins Medical Institutions, Baltimore, Md.
NEUROENDOCBINE TUMORS OF THE pancreas are rare
entities, with an annual incidence of approximately five clinically recognized cases per one million people. The first report of an islet cell tumor of the pancreas was in 1902 by Nichol1s.l A pancreatic endocrine carcinoma was first reported in 1927 by Wilder et a1.,2describing a physician with an insulinoma that had infiltrated most of the pancreas and metastasized to the liver. The first surgical cure of a pancreatic neuroendocrine tumor was reported by Howland et a1.3in 1929, describing the enucleation of a benign insulinoma from the body of the pancreas. Although the first successful pancreatiSupported in part by a grant from the National Institutes of Health (ROlCA56130). Presented at the Eighteenth Annual Meeting of the American Association of Endocrine Surgeons, Baltimore, Md., April 6-8, 1997. Reprint requests: Charles J. Yeo, MD, Department of Surgery, The Johns Hopkins Hospital, Blalock 606, 600 N. Wolfe St., Baltimore, MD 21287-4606. Copyright 0 1997 by Mosby-Year Book, Inc. 0039-6060/97/$5.00 + 0 11/6/84659
coduodenectomy was performed in 1912 by Kausch,4 it was not until 1940 that Whipple5 performed the first pancreaticoduodenectomy for an islet cell carcinoma of the pancreas. After Zollinger and Ellison presented the triad of gastric hypersecretion, peptic ulceration, and non-pcell islet tumors in 1955, Oberhelman et al.’ in 1961 suggested pancreaticoduodenectomy as surgical therapy for duodenal gastrinomas. Although most pancreatic endocrine tumors can be removed by enucleation or distal pancreatectomy, some may require pancreaticoduodenectomy for complete resection and potential cure because of their large size, location, or lymph node involvement in the periampullary region. High morbidity and mortality rates associated with pancreaticoduodenectomy in the 195Os, 1960s and 1970s limited its widespread acceptance and led to suggestions that the operation be abandoned.8 Recent advances in surgical techniques and perioperative care have made pancreaticoduodenectomy safer, particularly at institutions with high voland extensive umes experience,g-13 thus expanding the indications for the procedure to SURGERY
989
Phan et al.
Surgery December 1997
Year of Pancreaticoduodenectomy
Fig. 1. Annual distribution of 50 pancreaticoduodenectomies performed for periampullary neuroendocrine tumors at The Johns Hopkins Hospital.
Table I. Clinical
presentation
No. of patients
Symptoms/signs
Functional tumors (n = 17; 35%) Gastrinoma Insulinoma VIPoma Glucagonoma Nonfunctional tumors (n = 32; 65%) Jaundice Abdominal pain Weight loss Pruritus Diarrhea Nausea/vomiting Anorexia Gastrointestinal bleeding
%
8 7 1 1
47 41 6 6
15 15 14 6 6 3 2 2
47 47 44 19 19 9 6 6
include selected patients with both benign and malignant periampullary diseases. This article reviews the intraoperative and postoperative management, pathologic elements, and survival of 50 consecutive patients who underwent pancreaticoduodenectomy for periampullary neuroendocrine tumors at The Johns Hopkins Hospital. PATIENTS
AND
METHODS
Patient selection and follow-up. Between May 1962 and November 1996, inclusive, 50 patients underwent pancreaticoduodenectomy for pathologically proven periampullary neuroendocrine tumors at The Johns Hopkins Hospital; 42 of these patients underwent resection between 1987 and 1996. From 1991 to 1996, data were recordefi prospectively; for patients operated on before 1991, pathology records were examined retrospec-
tively to 1949 to identify these patients, with 1962 being the first year a pancreaticoduodenectomy was performed for a neuroendocrine tumor. Medical records were reviewed, and follow-up was obtained by review of office records or through telephone contact during October and November 1996. Five patients were lost to follow-up and were excluded from survival analyses. Of note, previous reports from this institution have included some of these patients: Udelsman et a1.14 presented 12, Yeo et aLI described 10, and Cameron et a1.g discussed four. Diagnostic criteria. Patients with appropriate signs, symptoms, and laboratory evidence of hormonal excess were classified into their respective clinical syndromes: insulinoma (Whipple’s triad, neuroglycopenia, and catecholamine surge), gastrinoma (peptic ulcer, diarrhea, and esophagitis) , vipoma (watery diarrhea, hypokalemia, and achlorhydria), and glucagonoma (hyperglycemia, necrolytic migratory erythema, and hypoproteinemia). Tumors of patients without a recognizable clinical syndrome and with normal serum hormone levels (with the exception of pancreatic polypeptide) were classified as nonfunctioning, independent of the results of the immunoperoxidase stains of the tumor specimen. Tumor characteristics. The site of the primary tumor was determined by the operative note or pathology report. Malignancy was determined by the presence of metastases to lymph nodes or the liver at operation or by the development of such metastases during follow-up. In cases in which the primary tumor was not found but pathology reports showed lymph node involvement (as occurred in two patients with Zollinger-Ellison syndrome), the tumor was classified as metastatic. Margin status and tumor diameter were determined from the pathology report. Postoperative course. Complications were defined by the following published criteria16: delayed gastric emptying (need for postoperative nasogastric decompression for more than 10 days), pancreatic flstula (daily drainage of 50 ml or more of amylase-rich fluid after postoperative day 7), intraabdominal abscess (radiographically defined or surgically found fluid collection associated with a septic clinical picture), and wound infection (need for the wound to be partially or wholly opened and packed). Statistical analyses. Survival and prognostic indexes were calculated by the Kaplan-Meier method. Differences among subsets were compared according to the log-rank test. Significance was accepted at the 5% level. Data are presented as
Phan et al. 991
surgery
Volume122, Number 6 Table
II. Intraoperative
factors
Type of resection
17 Classic (with hemigastrectomy) 33 Pylorus preserving
Estimated blood loss
Range 200-28,000 cc Median 800 cc Mean 1610 i- 570 cc
Packed red cell transfusion
Range O-35 units Median 0 units Mean 1.7 * 0.7 units
Operative
Range 3.9-13.4 hr Median 6.8 hr Mean 6.9 f 0.3 hr
time
Type of anastomosis
36 Pancreaticojejunostomies 7 Pancreaticogastrostomies 4 None (3 with concomitant pancreatectomy) 3 Undeterminable
total pancreatectomy,
1 with previous distal
Synchronous liver resection
3 For nonfunctional
tumors
Vessel resection
1 superior mesenteric vein resection for nonfunctional
the median or mean f standard error of the mean. Statistical computations were performed with either SPSS for Windows, version 7.0, or Core1 Quattro Pro 7. RESULTS Patient demographics. The study cohort consisted of 30 men (60%) and 20 women (40%), with ages ranging from 27 to 38 years and a mean age of 52 * 2 years. Forty-three patients (86%) were white, four (8%) were African-American, and three (6%) were of another race. The annual distribution of these patients is depicted in Fig. 1. Clinical presentation. Functional tumors were found in 17 patients (35%), whereas nonfunctional tumors were found in 32 patients (65%) ; the functional status of one patient’s tumor was not determinable (Table I). The most common functional tumor was gastrinoma, which was seen in eight patients; there were also seven insulinomas, one vipoma, and one glucagonoma. No patients with carcinoid syndrome were identified in this series. The majority (84%) of patients with nonfunctional tumors had jaundice, abdominal pain, or weight loss, similar to patients with nonneuroendocrine periampullary tumors. Intraoperative factors. Of the 50 pancreaticoduodenectomies, 1’7 were classic resections with hemigastrectomy and 33 were pylorus preserving
tumor
(Table II). Estimated blood loss ranged from 200 to 28,000 cc (in a patient with a portal vein injury), with a median of 800 cc and a mean of 1610 f 570 cc. The number of units of red blood cells transfused ranged from 0 to 35 (in the same patient with the portal vein injury), with a median of zero units and a mean of 1.7 f 0.7 units. The operative time ranged from 3.9 to 13.4 hours, with a median of 6.8 hours and a mean of 6.9 f 0.3 hours. The type of pancreatic-enteric anastomosis performed was a pancreaticojejunostomy in 36 patients and a pancreaticogastrostomy in seven; four patients required neither anastomosis (three patients underwent total pancreatectomies and one had undergone a previous distal pancreatectomy) ; for three patients the type of pancreaticenteric anastomosis constructed was undeterminable. Three patients underwent synchronous hepatic resection for nonfunctional tumors metastatic to the liver. Superior mesenteric vein resection was required in one patient with a nonfunctional tumor. Six patients with clinical and biochemical evidence of Zollinger-Ellison syndrome without definitive preoperative or intraoperative tumor identification underwent “blind” classic pancreaticoduodenal resection (Table III); all were eugastrinemic at last follow-up. Tumor characteristic. Malignant neoplasms were found in 29 patients (580/o), whereas benign
992
Phan et al.
~Surgery December 1997
.I 0.0
.I 0
12
24
36
48
60
72
I 84
0.0 0
12
24
Months
48
60
72
4
Months
Fig. 2. Actuarial survival curve after pancreaticoduodenectomy (n = 45 evaluable patients). Two-, 5-, and T-year survival rates were Sl%, 73%, and 65%, respectively.
Fig. 4. Actuarial survival curves after pancreaticoduodenectomy, comparing benign tumors (n = 18; solid line; median survival not reached) and malignant tumors (12= 27; dotted line; median survival 71 months). Curves are significantly different, with p = 0.03.
aged 4.5 f 1.1 cm, with a median of 3.0 cm. Malignant tumors averaged 3.2 * 0.4 cm, with a median of 3.0 cm. The site of the primary tumor was in the head, neck, or uncinate process of the pancreas in 39 patients (78%), the duodenum in six (12%), and the distal common bile duct in one (2%). The site
.a i .7 .6
36
i
5 1 .4 i 3
could
2
patients because the pancreatic tumor was too large (18 cm) and in two patients because tumor was found only in peripancreatic lymph nodes. Postoperative course. The postoperative length of stay ranged from 1 to 86 days, with a median of 15 days and a mean of 20 f 2 days. Seventeen patients (37%) were discharged home without in-hospital complications. Twenty-nine patients (63%) had some form of complications (Table V) ; the existence of complications could not be determined in four patients because of lack of adequate clinical information. The most common complications were wound infection (11 patients, 24%)) pancreatic fistula (11 patients, 24%), and delayed gastric emptying (seven patients, 15%). Three patients
.I1
0.01 0
12
24
36
48
60
72
84
Months Fig. 3. Actuarial survival curves after pancreaticoduodenectomy, comparing functional tumors (n = 15; solid line; median survival not reached) and nonfunctional tumors (n = 30; dotted line; median survival not reached) (difference not significant).
tumors were found in 21 (42%) (Table IV). Thirteen (76%) of the functional tumors were malignant, whereas 16 (50%) of the nonfunctional tumors were malignant. Of the functional tumors, all eight gastrinomas, the vipoma, the glucagonoma, and three of the insulinomas were malignant; only the remaining four insulinomas were benign. The mean diameter of the primary tumor for the entire cohort was 3.8 f 0.5 cm, with a median of 3.0 cm. The size of the functional tumors averaged 2.5 + 0.5 cm, with a median of 1.8 cm. Nonfunctional tumors were generally larger, with a mean of 4.3 * 0.7 cm and a median of 3.5 cm. Benign tumors aver-
not be determined
in four
patients:
in two
(6%) required reoperation: one for a pancreaticojejunostomy leak, one for repair of a fascial dehiscence, and one for repair of a gastrocutaneous fis-
tula. One patient (2%) with a benign insulinoma died of postoperative hemorrhage on postoperative day 1 as the result of a classic pancreaticoduodenectomy performed in 1967. There have been no postoperative in-hospital deaths in 48 consecutive pancreaticoduodenectomies for neuroendocrine tumors at The Johns Hopkins Hospital since 1967.
Phan
surgery
et al.
993
Volume 122, Number 6
Table III. Blind classic pancreaticoduodenectomy Patient
Gas&in
Preoperative
(Postoperative)
Preoperative
for Zollinger-Ellison studies
syndrome
Intraoperative
U/S
Pathology
F/U
0.5 cm Duodenum +1/6 LNs (-) margins
Alive 41 mo
0.3 cm Duodenum t1/9 LNs (-) margins
Died second cardiac arrest 15 mo
1.7 cm Neck +3/7 LNs (-) margins
Alive 44 mo
UGI: fundic mass Not used CT: no mass AbIG: 1.5 cm HOP mass SSRI, intense activity near gallbladder fossa
1.0 cm Duodenum t1/23 LNs (-) margins
Alive 1 mo
(39)
CT: no mass ANG: tumor blush in HOP area
Not used
Primary not found t4/29 LNs (-) margins
Alive 36 mo
(40)
UGI: wide duodenal Not used c loop EGD: duodenitis; gastritis Endoscopic U/S: no mass CT: no mass ANG: no tumor blush ERCP: normal
Primary not found
Alive 30 mo
UGI: duodenal ulcer EGD: esophagitis CT: hyperdense HOP ANG: no tumor blush
No mass identified
J.D.
Fasting PSS
571 875
(92) (129)
G.J.*
Fasting PSS
250 500
(45)
EGD: marginal ulcers Not used CT: no mass ANG: no tumor blush THPVS: gastrin step-up from HOP/duodenal area
J.M.”
Fasting PSS
800 2800
(41)
UGI: normal CT: 1.5 cm mass in neck of pancreas ANG: no tumor blush
E.O.
Fasting PSS
2520 3200
(40)
RR.
Fasting PSS
658 980
A.R.
Fasting PSS
475 t
No mass identified
+l/ll LNs (-) margins
U/S, Ultrasonography; F/U, follow-up; UGI, upper gastrointestinal series; PSS, post-secretin stimulation; EGD, esophagogastroduodenoscopy; CT, computed tomography; HOE head of pancreas; LNs, lymph nodes; ANG, selective mesenteric angiography; THPVS, transhepatic portal vencw sampling; SSEU, somatostatin receptor imaging; ERCE: endoscopic retrograde cholangiopancreatography. *Patients with history of antrectomy.
Postdischarge follow-up revealed that 11 patients (22%) had late complications, seven of which required operation: two for small bowel obstruction, four for incisional hernia, and one for gastric outlet obstruction. Other late complications included two patients with small bowel obstruction managed medically, three patients with late-identified wound infection, and two patients with intraabdominal abscessmanaged percutaneously. Survival and outcome. Follow-up ranged from 1 to 177 months (mean, 42 k 6 months). Thirty-five patients (78%) were alive at last follow-up, whereas 10 patients (22%) have died; five patients were lost to follow-up. Of the 10 deaths, seven patients died of progression of tumor, one of cardiac arrest, one of an unknown cause, and one in hospital of postoperative hemorrhage.
The overall actuarial survival rates at 2, 5, and 7 years were Sl%, 73%, and 65%, respectively (Fig. 2). The 2-, 5-, and ‘I-year survival rates for patients with functional tumors were 86%, 86%, and 71%, respectively, compared with 78%, 66%, and 66%, respectively, for patients with nonfunctional tumors (difference not significant) (Fig. 3). The 2, 5-, and 7-year survival rates for patients with benign tumors were significantly better (94%, 94%, and 94%) than they were for patients with malignant tumors (73%, 61%, and 41%, respectively; $I = 0.03) (Fig. 4). A univariate analysis of the 27 evaluable patients with malignant tumors was performed to assessthe role of various factors in influencing long-term survival (Table VI). In this analysis, factors such as estimated blood loss, operative time, tumor diameter,
994 Phan et al.
surgzry December 1997
Table IV. Tumor
characteristics
Malignancy 29 (58%) Malignant 20 (69%) Lymph nodes metastases only 4 (14%) Liver metastases only 5 (17%) Both lymph nodes and liver metastases 21 (42%) Benign Diameter of primary tumor (cm) Mean Median Entire group (n = 46) 3.8 + 0.5 3.0 Functional (n = 14) 2.5 + 0.5 1.8 Nonfunctional (n = 32) 4.3 * 0.7” 3.5 Benign (n = 19) 4.5 f 1.1 3.0 Malignant (n = 27) 3.2 + 0.4 3.0 Site of primary tumor HOP/neck/uncinate 39 Entire pancreas 2 Duodenum 6 Distal CBD 1 2 (primary tumor not found) Lymph nodes only HO8 Head of pancreas; CM, common *p = 0.10 compared
with functional
Table V. Postoperative
bile duct.
tumors by analysis of variance.
complications
Complication
DISCUSSION
No. ofpatients
%
17 11 11 7 4 3 3 2 1 1 1 1
37 24 24 15 9 7 7 4 2 2 2 2
None Wound infection Pancreatic fistula (1 re-op) Delayed gastric emptying Biliary fistula Intraabdominal abscess Pleural effusion Sepsis Fascial dehiscence (re-op) Gastrocutaneous fismla (re-op) Perioperative myocardial infarction Postoperative hemorrhage (death) Re-op, Reoperation.
postoperative tive length
complication
the tumor
red cell blood
0.05); patients a median
correlated resection
status of
survival.
transfusions whereas
Both
Patients
had a median
those not receiving
did not reach the median
survival
(p =
who had nonfunctional
tumors
had
survival
had functional survival
affected
of 32 months,
transfusions
survival.
status and the functional
significantly
receiving survival
status, and postopera-
of stay did not influence
the transfusion
of 43 months, tumors
(p = 0.05).
whereas
did not reach
Resection
margin
with survival;
patients
margins
a median
months,
compared
negative
margins
had
Range 0.3-18.0 0.3-8.0 0.6-18.0 0.5-18.0 0.3-8.0
with 71 months (fi = 0.07).
those who the median status also
who had positive survival in patients
of 13 with
In 1968 Howard17 published a personal series of 41 consecutive pancreaticoduodenectomies without an operative death. More recent reports have documented even larger consecutive series without deaths.ga12 Although pancreaticoduodenal resection remains a formidable procedure, there is a growing consensus that it is indicated in appropriately selected patients with periampullary adenocarcinoma, as well as selected patients with benign diseases18 such as chronic pancreatitis and less aggressive neoplasms such as neuroendocrine tumors.14 In this series the postoperative mortality rate was an acceptable 2%. There have been no postoperative deaths in 48 patients who have undergone morbidity rates resection since 1967. Postoperative have continued to be high (630/o), but most of the complications were not life-threatening and were managed nonoperatively. The relatively high rate of pancreatic fistula (24%) may be attributable to the fact that only one patient (2%) had a history of chronic pancreatitis, and the majority of patients had a soft, nonfibrotic pancreas at the site of pancreatic transection, which is known to be associated with an increased rate of anastomotic leakage.lg The relatively low rate of delayed gastric emptying observed in this series (15%) may be a reflection of our routine prophylactic use of the prokinetic agent erythromycin.16 Although neuroendocrine tumors are more indolent than pancreatic exocrine adenocarcinoma, they can cause significant morbidity and death.
Phan et al.
surgery
995
Volume 122, Number 6
Most of the nonfunctional tumors in this series caused significant obstructive and constitutional symptoms, as evidenced by the fact that 84% of the patients had jaundice, abdominal pain, or weight loss. All patients with functional tumors had symptoms of the corresponding hormone at presentation. At the time of operation the decision to proceed to pancreaticoduodenectomy was made cautiously and only after careful assessment of the resection options. There is undoubtedly a role for less aggressive resection in selected patients with periampullary neuroendocrine neoplasms. For example, we have performed local excision of benign duodenal gastrinomas, and we have enucleated benign insulinomas from the region of the head and uncinate process of the pancreas. Such local procedures are generally reserved for small tumors, without evidence of nodal metastases. In this series the mean tumor diameter was 3.8 f 0.5 cm (median, 3.0 cm), and 58% of all tumors were malignant, with 86% of the malignant tumors having lymph node metastases. Even in the group of six patients treated by blind classic pancreaticoduodenectomy (Table III), in which three of the tumors were found by the pathologist to be 1.0 cm or less in the duodenum, all patients had nodal metastases in the resected specimen. These nodal metastases could have been left unresected if a local resection of the duodenal primary tumor had been performed. Additional support for the complete excision of malignant tumors comes from the data (Table VI) comparing median survival between patients who had positive margins (13 months) with those patients who had negative margins (71 months). In fact, many studies have concluded similarly that aggressive resection of malignant neuroendocrine tumors significantly improves long-term survival.20~2’ The six patients with Zollinger-Ellison syndrome for which a blind pancreaticoduodenectomy was performed had biochemical evidence of and symptoms from a gastrinoma. Their symptoms included abdominal pain, diarrhea, peptic ulcer disease, and multiple gastrointestinal bleeds; two patients had previously undergone partial gastrectomy and vagotomy without symptomatic relief. Preoperative imaging studies failed to localize these tumors definitively. When a mass was not found after exploring the pancreas and duodenum, intraoperative ultrasonography was used in two cases, both without success in locating the suspected tumor. In all cases the decision to perform a blind pancreaticoduodenectomy was made on strong clinical suspicion that the tumor was in the gastrinoma triangle. 22 In five of the six cases at least one
Table VI. Univariate survival analysis for malignant tumors (fi = 27) Parameter
Median
EBL Z300 ml <800 ml
NR
Transfusion Yes No Operative time 26.8 hr ~6.8 hr
Tumor function Nonfunctional Functional Tumor diameter 23.0 cm C3.0 cm
71
Survival (mo) Mean p Value
83 + 23 47 5 7
0.43
66 i 22 51+5
0.05
NR NR NR
104 i 25 43 f 7
0.82
43 NR
37 + 6 128 + 29
0.05
71
92 * 30 78 + 25
0.37
43 13 71
26 + 9 105 rt 27
0.07
NR 71
113 f 22 53 zk10
0.81
43 71
88 f 27 58 + 7
0.36
32
Surgical margins Complications Yes No Postoperative LOS 215 days ~15 days
EBL, Estimated blood loss; NT?, not reached; LOS, length of stay.
preoperative study had suggested that the tumor was localized to the head of the pancreas or the duodenum. In the remaining case, intraoperative lymph node biopsy in the gastrinoma triangle revealed metastatic neuroendocrine tumor. Fortunately, these blind resections have rendered all of these patients symptom free and eugastrinemic at last follow-up. Because this series encompasses a 35-year period, much has evolved in the management of patients with neuroendocrine neoplasms. Preoperative imaging modalities such as dynamic thin-slice contrast-enhanced computed tomography, endoscopic ultrasonography, octreotide scanning, and selective arterial secretin stimulation testing have improved the ability to localize these tumors before laparotomy. In addition, intraoperative techniques including ultrasonography and duodenal transillumination further improve tumor localization. None of these techniques was available 20 years ago. In summary, this single-institution experience documents a low mortality rate and acceptable morbidity rate in a highly selected group of 50 patients treated by pancreaticoduodenectomy for
996
Phan et al.
periampullary neuroendocrine tumors. For the entire cohort the 5-year actuarial survival rate was 73%. Survival analysis of the patients with malignant tumors indicated that three factors correlated with decreased survival: blood transfusion,23a24 nonfunctional tumor status,25 and the presence of positive surgical margins. REFERENCES 1. Nicholls AG. Simple adenoma of the pancreas arising from an island of Langerhans. J Med Res 1902;8:385-95. 2. Wilder RM, Allan FN, Power MH, Robertson HE. Carcinoma of the islands of the pancreas: hyperinsulinism and hypoglycemia. JAMA 1927;89:348-55. 3. Howland G, Campbell WR, Maltby EJ, Robinson WL. Dysinsulism: convulsions and coma due to islet cell tumor of the pancreas, with operation and cure. JAMA 1929;93:6749. 4. Kausch W. Das carcinom der papilla duodeni und seine radikale entfernung. Beitr Klin Chir 1912;78:439-86. 5. Whipple AO. Pancreaticoduodenectomy for islet cell carcinoma: a five-year follow-up. Ann Surg 1945;121:847-52. 6. Zollinger RM, Ellison EH. Primary peptic ulcerations of the jejunum associated with islet cell tumors of the pancreas. Ann Surg 1955;142:709-28. 7. Oberhelman HA Jr, Nelsen TS, Johnson AN Jr, Dragstedt LR II. Ulcerogenic tumors of the duodenum. Ann Surg 1961;153:21427. 8. Crile G Jr. The advantages of bypass operations over radical pancreatoduodenectomy in the treatment of pancreatic cancer. Surg Gynecol Obstet 1970;130:1049-53. 9. Cameron JL, Pitt HA, Yeo CJ, Lillemoe KD, Kaufman HS, Coleman J. One hundred fort@ve consecutive pancreaticoduodenectomies without mortality. Ann Surg 1993;217:430-8. 10. Miedema BW, Sarr MG, van Heerden JA, Nagorney DM, McIlrath DC, Ilstrup D. Complications following pancreaticoduodenectomy: current management. Arch Surg 1992;127:945-50. 11. Delcore R, Friesen SR. Role of pancreatoduodenectomy in the management of primary duodenal wall gastrinomas in patients with Zollinger-Ellison syndrome. Surgery 1992;112:1016-23. 12. Trede M, Schwa11 G, Saeger H-D. Survival after pancreatoduodenectomy: 118 consecutive patients without an operative mortality. Ann Surg 1990;211:447-58. 13. Gordon TA, Burleyson GP, Tielsch JM, Cameron JL. The effects of regionalization on cost and outcome for one general high-risk surgical procedure. Ann Surg 1995;221:43-9. 14. Udelsman R, Yeo CJ, Hruban RH, Pitt HA, NiederhuberJE, Coleman J, et al. Pancreaticoduodenectomy for selected pancreatic endocrine tumors. Surg Gynecol Obstet 1993;177:269-78. 15. Yeo CJ, Wang BH, Anthone GJ, Cameron JL. Surgical experience with pancreatic islet-cell tumors. Arch Surg 1993;128:1143-8. 16. Yeo CJ. Management of complications following pancreaticoduodenectomy. Surg Clin North Am 1995;75:913-24. 17. Howard JM. Pancreatico-duodenectomy: forty-one consecutive Whipple resections without an operative mortality. Ann Surg 1968;168:629-40. 18. Barnes SA, Lillemoe KD, Kaufman HS, Sauter PK, Yeo CJ, Talamini MA, et al. Pancreaticoduodenectomy for benign disease. Am J Surg 1996;171:131-5.
surgery December 1997 19. Yeo CJ, Cameron JL, Maher MM, Sauter PK, Zahurak ML, Talamini MA, et al. A prospective randomized trial of pancreaticogastrostomy versus pancreaticojejunostomy following pancreaticoduodenectomy. Ann Surg 1995;222:580-92. 20. Legaspi A, Brennan ME Management of islet carcinoma. Surgery 1988;104:1018-23. 21. Carty SE, Jensen RT, Norton JA. Prospective study of aggressive resection of metastatic pancreatic endocrine tumors. Surgery 1992;112:102432. 22. Stabile BE, Morrow DJ, Passaro E Jr. The gastrinoma triangle: operative implications. Am J Surg 1984;147:25-31. 23. Cameron JL, Crist DW, SitzmannJV, Hruban RH, Boitnott JI3, Siedler AJ, et al. Factors influencing survival following pancreaticoduodenectomy for pancreatic cancer. Am J Surg 1991;161:120-5. 24. Wu H, Little A. Perioperative blood transfusions and cancer recurrence. J Clin Oncol 1988;6:1348-54. 25. Eckhauser FE, Cheung PS, Vinik AI, Strode1 WE, Lloyd RV, Thompson NW. Nonfunctioning malignant neuroendocrine tumors of the pancreas. Surgery 1986;100:978-88. DISCUSSION Dr. Richard Prinz (Chicago, Ill.). This is a 35-year review, yet 42 of the patients have been seen in the past
9 years. During this g-year period how many patients with neuroendocrine tumors of the pancreas with distal resections or enucleations?
were treated Focusing on
lesions in the head, do you treat anybody with enucleation in Baltimore? Current treatment of small gastrinomas in the duodenum and the head of the pancreas is enucleation and a nodal dissection. Have you done that procedure? If so, what were your results? Are poor results with this approach the reason you have turned to doing pancreaticoduodenectomy? Positive nodes were found in all of your resections for Zollinger-Ellison syndrome. You claim that all of these patients were eugastrinemic afterward. Have you done secretin stimulation tests in addition to basal serum gas&in levels to prove that? Not all of our patients have become eugastrinemic after enucleation and node dissection. Most studies with other types of tumors, such as colon cancer and even pancreatic adenocarcinoma, showed that survival is better ifyou do not have to transfuse patients. Are you showing the opposite, or am I misreading your results? Dr. Yeo. We actually perform operations in Baltimore other than pancreaticoduodenectomy; we do distal pancreatectomies and enucleations. Obviously, much of this pancreaticoduodenectomy experience is related to our referral patients
basis, being a tertiary-care center, and many are sent to us who require pancreaticoduo-
denectomy. For example, during the same period we reviewed 125 islet cell tumor resections and performed 50 pancreaticoduodenectomies, 39 distal pancreatectomies,
and 21 enucleations.
I think
a lot of the bias is
related to the referral pattern. As to the issue of gastrinomas,
in this particular
series we have eight
gastrino-
mas. We do perform enucleations and nodal dissections as well. We have done fewer of those than Whipple procedures for gastrinoma. For the “blind” resections we have one patient who is eugastrinemic who has a negative secretin stimulation, and the other four are eugastrinemic and are apparently cured in relatively short-
Phan et al.
surgery Volume 122, Ahmber 6 term follow-up. As to the blood transfusion data, there is better survival when the patient received no blood transfusions during the operation. Dr. Peter Angeles (Chicago, Ill.). I am curious in particular about the two patients who had nodal disease only. How was the decision made to do a pancreaticoduodenectomy in these patients in whom only nodal disease was found? Dr. Yeo. Both of the patients with nodal disease only were patients with gastrinoma. One had undergone a previous vagotomy and antrectomy at another institution. We never received the pathology slides, and we suspected that residual disease was making the patient hypergastrinemic. In the other patient we never found the primary tumor. Perhaps the primary tumor was just too small to be localized by our pathologists. However, that patient also remained eugastrinemic and symptom free at long-term follow-up. Dr. Geoffrey B. Thompson (Rochester, Minn.). Did you look at nodal status and what effect it had on overall prognosis? Were any of the Whipple procedures performed for Zollinger-Ellison syndrome in patients with multiple endocrine neoplasia type 1 (MEN l)? If so, what were the long-term results in that particular subgroup? One of the main concerns about doing a Whipple pro-
997
cedure for islet cell tumors is that the surrounding pancreas is normal and often very soft, making pancreaticojejunostomy a major concern. Are you doing anything special for these patients, such as using octreotide (Sandostatin) before and after the operation or fibrin glue during the operation? Dr. Yeo. The nodal status of the patients was not predictive of survival because nodal status was used to define the malignant state. Nodal status is not a significant variable in the univariate analysis. None of these Whipple procedures was done for patients with MEN 1, although we have operated on a handful of patients with MEN 1 and gastrinomas. We typically do a distal gastrectomy and enucleate the lesions from the duodenum. We have a 24% fistula rate, which is quite high. Most patients had normal glands at the pancreatic neck, and the 24% fistula rate is by a very strict definition of fistula in patients with normal glands. We now use octreotide prophylactically for a soft pancreas, although I am still not convinced that it is as useful as is touted. Only one of the 11 patients who had a fistula required reexploration for completion pancreatectomy. But there definitely is a higher risk of pancreatic fistula in these patients than in those with pancreatic adenocarcinoma.
ANNOUNCEMENT Effective October 1, 199’7, please send all manuscripts and other submissions for Surguy to the following address: Kim Fons Managing Editor surgery 11830 Westline Industrial Drive St. Louis, MO 63146-3318