To evaluate the efficacy of early angiography in the diagnosis and treatment of delayed arterial hemorrhage as a significant complication following pancreatectomy.
Retrospective case series.
University hospital; 6-year period between January 1992 and December 1997.
Patients and Intervention
Of a total of 81 patients undergoing pancreatic head resection, 10 (12%) developed massive arterial hemorrhage after surgery. All 10 patients underwent an emergency angiography and 8 of these were managed by transcatheter arterial embolization.
Before onset of major bleeding, 9 patients (90%) had developed intra-abdominal abscess resulting from pancreatic fistula or other anastomotic leak, and all 10 patients had had preliminary minor bleeding. The angiogram demonstrated an exact site of bleeding as a pseudoaneurysm in all 10 patients. Transcatheter arterial embolization achieved temporary control of bleeding in all 8 patients for whom embolization was attempted and complete hemostasis in 5 of the 8 subsequently, yielding a success rate of 63%. Overall, 4 patients (40%) died of complications related directly to the major hemorrhage or the transcatheter arterial embolization technique.
An emergency angiography should be considered in all patients who develop either a sentinel or massive gastrointestinal bleed following pancreatectomy irrespective of suspected intra-abdominal sepsis. Transcatheter arterial embolization allows temporary control of massive hemorrhage and hemodynamic stabilization in most cases, and prevents the need for high-risk emergency surgery.
RECENT ADVANCES in surgical techniques and improvements in perioperative management have drastically reduced morbidity and mortality after pancreatectomy.1-5 However, pancreatic anastomotic leak remains of great concern because it sometimes results in the life-threatening complication of arterial hemorrhage.6-10 Early diagnosis and appropriate management are mandatory because major arterial hemorrhage has a mortality rate of up to 50%.11 Despite the obvious importance of this potentially lethal problem, only a few studies and case reports have focused on massive hemorrhage following pancreatic surgery.11-14 There is no diagnostic and therapeutic modality established. Recently, Reber et al14 described 5 patients with extrahepatic pseudoaneurysms of the hepatic arteries following pancreatobiliary surgery, and highlighted the superselective microcoil embolization as a treatment of choice in high-risk patients. We have treated 10 patients with delayed arterial hemorrhage as a result of anastomotic complications after pancreatectomy using interventional radiology in these 5 years. In this communication, our experience with these 10 patients was analyzed to elucidate the clinical problems. Special attention was focused on the efficacy of early angiography in the diagnosis and management of delayed arterial hemorrhage after pancreatectomy.
From January 1992 through December 1997, a total of 81 patients underwent pancreatic head resection with pancreatoenterostomy at the Department of Surgery I, Kyushu University, Fukuoka, Japan. Of these 81 patients, 10 (12%) developed postoperative massive arterial hemorrhage, defined as major bleeding from the abdominal drains and/or the gastrointestinal tract requiring transfusion of at least 4 U of packed cells within the first 24 hours after the onset of hemorrhage. The medical records of these 10 patients were retrospectively reviewed. There were 4 men and 6 women with a mean age of 64 years (range, 53-84 years). Indications for initial surgery were ampullary cancer in 5 patients, distal bile duct cancer in 2 patients, pancreatic cancer in 1 patient, serous cystadenoma of the pancreas in 1 patient, and intraductal papillary adenoma of the pancreas in 1 patient. All patients had elective surgery under favorable conditions. Operative procedures included pylorus-preserving pancreatoduodenectomy (8 patients), standard pancreatoduodenectomy with partial hepatectomy (1 patient), and duodenum-preserving pancreatic head resection (1 patient). In patients with malignant diseases, extensive skeletonization of the hepatic artery was usually performed to remove neural and lymphoid tissue of the hepatoduodenal ligament. All patients underwent an end-to-side, 2-layer pancreatojejunostomy using a mucosa-to-mucosa anastomosis with a stenting tube (5 patients) or invagination technique with a total tube drainage method (5 patients). In all patients, an end-to-side, single-layer hepaticojejunostomy with a stent was created with interrupted absorbable sutures distal to the pancreatojejunostomy. A gastrostomy tube was placed in all patients using the Witzel technique. Two open silicone drains were routinely placed in the vicinity of the pancreatic anastomosis, but not in direct contact with it. Prophylactic octreotide was not used in any patient. Pancreatic fistula was defined as a high amylase level (>1000 U/L) in the drainage fluid collected from the peripancreatic drains and/or anastomotic disruption demonstrated radiologically. Leakage of the hepaticojejunostomy or duodenojejunostomy was assessed by contrast medium study through the biliary stenting tube or the gastrostomy tube. Criteria for intra-abdominal abscess were defined as intra-abdominal fluid collection proved radiographically, in association with clinical signs such as abdominal pain, temperature of greater than 38°C, or leukocyte counts of more than 15 ×109/L. Sentinel bleeding was considered to be minimal bleeding from either the abdominal drains or the gastointestinal tract without evidence of peripheral circulatory impairment or transfusion requirement.
Massive arterial hemorrhage occurred in 10 patients from 19 to 206 days (median, 27.5 days) after surgery. Before onset of the hemorrhage, pancreatic fistula was identified in 6 patients (60%), leakage of the duodenojejunostomy in 2, and hepaticojejunostomy in 1. Moreover, 9 patients (90%) had developed an intra-abdominal abscess. Two of the 9 patients required operative drainage to control the intra-abdominal abscess. In 1 of the 2 patients, resection of the duodenum was performed for duodenal necrosis and severe pancreatic fistula 20 days after duodenum-preserving pancreatic head resection. Another patient had developed acute necrotizing pancreatitis and required drainage of the pancreatic bed 7 days after pylorus-preserving pancreatoduodenectomy. The other 7 patients were treated conservatively, with maintenance of the abdominal drains placed intraoperatively and use of broad-spectrum antibiotics. All 10 patients had had preliminary minor bleeding from the abdominal drains or the gastrointestinal tract 12 hours to 10 days (mean, 3 days) prior to the major hemorrhage. Each episode of these sentinel bleeds was associated with abdominal pain or high fever.
After restoration of hemodynamic stability by fluid loading, all patients underwent emergency angiography using the standard Seldinger technique and preshaped nonoccluding femorovisceral catheters. Panabdominal, celiac, and superior mesenteric artery angiography was performed. The angiogram demonstrated a pseudoaneurysm with (n=7) or without (n=3) extravasation in all 10 patients (Figure 1). The exact site of bleeding and/or pseudoaneurysm was located in the proper (3 patients) or common (3 patients) hepatic artery in 6 patients, the stump of the gastroduodenal artery in 2 patients, the splenic artery in 1 patient, and the branch of the superior mesenteric artery in the other patient. Before the angiographic investigation, gastrointestinal endoscopy was performed in 3 patients because of a sentinel bleed from the gastrointestinal tract; however, it failed to determine the exact site of bleeding because the gastrointestinal tract was full of blood clots.
Transcatheter arterial embolization (TAE) was attempted in 8 patients (80%) by a senior radiologist, using various coil occlusion devices. The choice of TAE technique was dependent on the site of arterial hemorrhage and the ability to selectively catheterize the vessel. The occluded vessels were common hepatic artery in 4 patients, proper hepatic artery in 2 patients, gastroduodenal artery in 1 patient, and branch of the superior mesenteric artery in the other patient. Control of hemorrhage, as evidenced by cessation of extravasation and obliteration of pseudoaneurysms, was confirmed by repeated angiography. Temporary control of the bleeding was obtained in all the 8 patients in whom embolization was attempted (Figure 2). Despite the technical success, 1 patient died of multiple organ failure resulting from hemorrhagic shock immediately after the embolization. In 1 patient whose portal vein was revealed to be stenotic on the venous phase, occlusion of the common hepatic artery resulted in acute hepatic failure, leading to the patient's death 2 days after TAE. One patient experienced rebleeding 20 days following the procedure from an arterial branch different from the one originally occluded and required another TAE. This patient died of multiple organ failure 4 days after the second embolization of the common hepatic artery. Therefore, complete occlusion of the bleeding artery was obtained in 5 patients subsequently, yielding a short-term success rate of 63%. After the successful embolization, 2 patients died during the hospital stay of fulminant sepsis (2 and 2.5 months after embolization). However, hemodynamic stability had been attained in both patients following embolization, and neither had a recurrence of hemorrhage.
In 2 patients, an alternative management strategy was employed and no embolization was attempted. In 1 of these 2 patients, prolonged arterial spasm during catheter manipulation stopped visible evidence of bleeding, and no further treatment was required. In another patient who was found to have a pseudoaneurysm of the common hepatic artery, thorombotic occlusion of the portal vein precluded TAE and required relaparotomy for hemostasis. However, the surgical approach to the bleeding pseudoaneurysm was unsuccessful because of dense adhesion and the patient died of recurrent hemorrhage 2 months after the reoperation.
The most common and difficult complication after pancreatoduodenectomy is pancreatic anastomotic leakage; the incidence ranges from 8% to 25% in most series.1-10 Most patients with pancreatic fistula can be managed by conservative measures, including maintenance of the peripancreatic drains placed intraoperatively and the use of antibiotics and octreotide.7,9 However, when the patient develops an intra-abdominal abscess, percutaneous or operative drainage is mandatory because regional sepsis may cause the life-threatening complication of arterial hemorrhage. Arterial bleeding occuring after pancreatic surgery has a dismal clinical outcome, with a mortality rate of 20% to 50%.8,11,13 In our series, delayed arterial hemorrhage occurred in 10 (12%) of 81 patients undergoing pancreatectomy with pancreatoenterostomy. Four of the 10 patients died of causes related directly to the arterial hemorrhage; thus, the mortality rate in these patients was 40%. Because delayed arterial hemorrhage is such a dreaded complication after pancreatic surgery, we retrospectively analyzed our experience of the 10 patients to evaluate the clinical problems.
Of these 10 patients, 9 (90%) had developed intra-abdominal abscess resulting from the pancreatic fistula or the dehiscence of other anastomotic sites before the onset of bleeding. These findings provide more evidence to support the theory that local sepsis erodes the major vessels in the pancreatic bed denuded by regional lymphadenectomy and may result in a pseudoaneurysm formation and/or sloughing of a ligature at an arterial stump.11,13 A relatively longer time interval (median, 27.5 days) between initial surgery and the onset of bleeding is another factor suggesting this mechanism. Notably, 1 patient had arterial hemorrhage 206 days after surgery. This patient had suffered from relapsing fulminant sepsis resulting from uncontrolled intra-abdominal abscess during a prolonged and complicated hospital stay. In patients with intra-abdominal sepsis, therefore, a possibility of arterial hemorrhage should be kept in mind even in the late phase after pancreatic resection.
Shankar and Russell12 described 9 patients with major bleeding after pancreatic resection and first referred to a preliminary warning bleed, which preceded a major hemorrhage by 6 hours to 10 days. Brodsky and Turnbull13 also emphasized the importance of sentinel bleed as a prelude to arterial hemorrhage in their series of 5 cases. In this series, a warning minor bleed from either the abdominal drains or gastrointestinal tract was present in all patients as an indicator of imminent arterial hemorrhage. A careful review of records of 81 patients undergoing pancreatic resection during this period identified only 1 patient with a sentinel bleed that had not been followed by a major bleeding. Based on our experience that the sentinel bleed was followed by massive arterial hemorrhage in the majority of cases, our current policy is to attempt early angiography as soon as the sentinel bleed is recognized.
Precise identification of the bleeding artery is essential for accurate treatment. When gastrointestinal bleeding is suspected after pancreatic resection, endoscopy is usually employed as the first diagnostic procedure. However, gastrointestinal endoscopy often fails to determine the bleeding site accurately in patients with delayed arterial hemorrhage.12 Moreover, positive endoscopic findings such as erosive gastritis can be misleading and may delay treatment of patients with a precarious hemodynamic state.13 In this series, emergency endoscopy was performed in 3 patients who had demonstrated sentinel bleeding from the gastrointestinal tract, but we could not find the bleeding site because of massive blood clots in the gastrointestinal tract. The difficulty with endoscopy may be related to the fact that arterial hemorrhage occurs near the pancreatojejunal anastomosis and passes through the gastrointestinal tract. On the other hand, angiography may be the first-step diagnostic method that will elucidate the exact site of bleeding. In this series, angiography demonstrated the bleeding point as a pseudoaneurysm in all 10 patients, irrespective of the presence or absence of continuous bleeding. Although it has been reported that major bleeding sometimes could not be identified during angiography, probably due to the intermittent nature of the hemorrhage,12,13 we reconfirmed the diagnostic value of angiography, which identified the site of bleeding as a pseudoaneurysm.
The traditional approach to treatment for patients with delayed arterial bleeding after pancreatectomy has been surgical; however, emergency surgery for high-risk patients with hemodynamic instability and poor general condition has a high mortality rate.8,15 The surgical principle recommended was proximal and distal suture ligation of the involved artery, evacuation of hematoma, drainage of septic collection, and repair of anastomotic dehiscence.13 However, a surgical approach to the bleeding artery is often hazardous or even unsuccessful owing to the anatomical inaccessibility of these arteries and associated inflammatory reaction, especially in patients who have undergone multiple previous abdominal operations. In 1 patient who underwent relaparotomy, we could not reach the bleeding pseudoaneurysm because the tissues were hyperemic and friable, and the tissue planes were difficult to dissect owing to dense adhesions.
With the recent advances in interventional radiology, TAE has been used widely and has gained acceptance for the treatment of the visceral aneurysm or pseudoaneurysm in the last decade.16-20 Mandel et al16 treated 19 patients with bleeding peripancreatic arterial aneurysms with embolization and showed a 79% success rate. Gambiez et al17 described 14 patients with bleeding pseudocysts complicating chronic pancreatitis treated by arterial embolization with a success rate of 79%. Angiographic treatment has now been attempted in a few patients with arterial hemorrhage after pancreatic and/or biliary surgery.11,13,14,16 We have treated 8 patients using TAE alone, with a short-term success rate of 63%. One patient, in whom the intra-abdominal abscess had been uncontrollable after TAE, experienced rebleeding 20 days following the procedure. Despite an immediate success of another TAE, this patient died of multiple organ failure resulting from hemorrhagic shock. This case illustrates that patients with persistent intra-abdominal septic condition still require surgical drainage even after TAE to avoid a possible recurrent hemorrhage. Regarding the complication of TAE, fatal hepatic necrosis occurred in 1 patient (13%) immediately after embolization of the common hepatic artery; in this patient the stenosis of the portal vein was evident on angiography. We did not observe other complications related to the embolization treatment. It is not obvious whether TAE is superior to other therapeutic options because no data are presented, nor are direct comparisons made with alternate management strategies such as surgery or other transcatheter techniques. We can safely say that angiographic embolization allows temporary control of massive hemorrhage and hemodynamic stabilization in the majority of cases, and prevents the need for high-risk emergency surgery.
In summary, we would like to emphasize the efficacy of early angiography as a first-step diagnostic method in patients with major arterial hemorrhage after pancreatectomy. Transcatheter arterial embolization, when indicated, can achieve temporary control of hemorrhage and hemodynamic stabilization and even permanent hemostasis. After a successful TAE, however, further surgery should be reserved for patients with uncontrolled intra-abdominal septic conditions to prevent possible recurrent bleeding.
Reprints: Koji Yamaguchi, MD, Department of Surgery I, Kyushu University Faculty of Medicine, Fukuoka 812-8582, Japan (e-mail: email@example.com).
J One hundred and forty-five consecutive pancreaticoduodenectomies without mortality. Ann Surg.
1993;217430- 435Google ScholarCrossref
H Survival after pancreatoduodenectomy: 118 consecutive resections without an operative mortality. Ann Surg.
1990;211447- 458Google ScholarCrossref
WP Decreased morbidity and mortality after pancreatoduodenectomy. Am J Surg.
1986;151141- 149Google ScholarCrossref
LW An analysis of the reduced morbidity and mortality rates after pancreaticoduodenectomy. Arch Surg.
1989;124778- 781Google ScholarCrossref
PF Pyloric and gastric preserving pancreatic resection: experience with 87 patients. Ann Surg.
1986;204411- 418Google ScholarCrossref
DM Pancreatic anastomotic leakage after pancreaticoduodenectomy: incidence, significance, and management. Am J Surg.
1994;168295- 298Google ScholarCrossref
PJ Pancreaticoduodenal resection: surgical experience and evaluation of risk factors in 103 patients. Ann Surg.
1984;199432- 437Google ScholarCrossref
D Complications following pancreaticoduodenectomy: current management. Arch Surg.
1992;127945- 950Google ScholarCrossref
JG Pancreatic fistula complicating pancreatectomy for malignant disease. Br J Surg.
1981;68238- 240Google ScholarCrossref
van Berge Henegouwen
DJ Delayed massive haemorrhage after pancreatic and biliary surgery. Br J Surg.
1995;821527- 1531Google ScholarCrossref
AD Arterial hemorrhage after pancreatoduodenectomy: the "sentinel bleed.". Arch Surg.
1991;1261037- 1040Google ScholarCrossref
MW Superselective microcoil embolization: treatment of choice in high-risk patients with extrahepatic pseudoaneurysms of the hepatic arteries. J Am Coll Surg.
1998;186325- 330Google ScholarCrossref
JA Completion pancreatectomy following pancreaticoduodenectomy: clinical experience. World J Surg.
1992;16521- 524Google ScholarCrossref
MA Nonoperative management of peripancreatic arterial aneurysms: a 10-year experience. Ann Surg.
1987;205126- 128Google ScholarCrossref
PA Arterial embolization for bleeding pseudocysts complicating chronic pancreatitis. Arch Surg.
1997;1321016- 1021Google ScholarCrossref
LW Control of major haemorrhage associated with pancreatic pseudocysts by transcatheter arterial embolization. Br J Surg.
1984;71133- 136Google ScholarCrossref
RB Nonoperative management of visceral aneurysms and pseudoaneurysms. Am J Surg.
1992;164215- 219Google ScholarCrossref
JST Current management of visceral artery aneurysms. Surgery.
1996;120627- 634Google ScholarCrossref