Introduction

Acute haemorrhage is the most challenging complication in patients with pancreatitis. Although it is considered uncommon, bleeding into the gastrointestinal tract, pancreatic duct, retroperitoneum, or peritoneal cavity results in a considerable increase in mortality. Because spontaneous major haemorrhage in patients with chronic pancreatitis is rare, limited data have been reported in small, retrospective series or case reports [1–5].

Acute bleeding in chronic pancreatitis from peptic ulcer disease or Mallory–Weiss syndrome might be indirectly related to coincidental chronic pancreatitis. Splenic vein thrombosis with occlusion leads to portal hypertension and gastrointestinal varices, which is a potential source of bleeding [6]. Splenic vascular complications, such as infarction, intrasplenic pseudocysts, and spontaneous splenic rupture, have all been reported [2, 3, 7, 8]. However, both the incidence and clinical outcome remain unclear for these complications.

Another reason for major haemorrhage in chronic pancreatitis is the development of vascular necrosis. Peripancreatic and splanchnic arteries exposed to proteolytic pancreatic enzymes frequently demonstrate local arteritis and destruction of the vessel wall. These pathological changes may be manifested as either pseudoaneurysm formation or bleeding into a pre-existing pseudocyst, which transforms the pseudocyst into a pseudoaneurysm. Very rarely, a pseudoaneurysm might erode into the pancreatic duct, with haemorrhage from the ampulla, which is described as haemosuccus pancreaticus [5, 9].

Acute haemorrhage from pseudoaneurysms is a rare, but frequently lethal, complication in chronic pancreatitis. Mortality rates in the literature range from 10% to 50% [1, 4, 10]. The optimal surgical or interventional treatment modality is controversial, as no evidence-based data are currently available. Most data regarding visceral pseudoaneurysms are obtained from patient series that include diagnoses of both acute and chronic pancreatitis [11–18].

In this prospective cohort study, we report our experience with bleeding complications associated with chronic pancreatitis in a series of 541 patients treated in one surgical department over a period of 9.5 years.

Patients and methods

Between October 1993 and April 2003 a cohort of 541 patients with chronic pancreatitis was admitted for treatment at our surgical department. The diagnosis of chronic pancreatitis was established by the presence of one or more of the following: calcifications seen on abdominal computed tomography (CT), characteristic ductal lesions in endoscopic retrograde cholangiopancreatography (ERCP), and previous histological evidence from a surgical specimen or biopsy. Data from these patients regarding bleeding complications were reviewed from our prospective pancreas database. Patients with postoperative complications after pancreatic surgery during the same hospital stay were excluded.

In total, 36 patients with chronic pancreatitis (6.7% of all admitted patients) presented with severe bleeding complications. Their average age was 47.9 years (range 28–75 years). Eighty-eight percent of the patients were male. Alcohol abuse was the main cause of chronic pancreatitis (87%). Four patients had undergone previous surgery for their pancreatic disease (one distal pancreatectomy, two Whipple resections, one cystojejunostomy).

A severe bleeding complication was defined as a loss of more than 2.0 g/dl of haemoglobin within 24 h and concurrent apparent gastrointestinal blood loss or evidence of blood under radiological examination. Patients who were treated conservatively for minor bleeding into a pseudocyst that resolved spontaneously (six additional cases) were excluded from the study.

For pseudoaneurysms, our approach was to perform angiography whenever possible, depending upon the patient’s haemodynamic condition. In cases of embolization failure or haemodynamic instability, patients underwent surgery immediately. In cases of successful embolization, patients were observed for at least 48 h in the intensive care unit (ICU). Semi-urgent surgery was defined as surgery after angiography, within 48 h, in patients demonstrating intermittent bleeding, or after stabilization of haemodynamic function in the ICU following haemorrhagic shock syndrome. When surgery was found to be necessary, partial pancreatectomy was the preferred procedure in order to prevent a recurrence of bleeding. Alternatively, suture ligation of a bleeding vessel was also undertaken, depending upon the site of haemorrhage. Visceral angiography and embolization were performed by experienced radiologist consultants. The technique of super-selective embolization was performed with stainless steel coils (William Cook, Europe; Bjaeverskov, Denmark), which were placed distally and proximally to the arterial wall defect, or in the pseudoaneurysm itself, thus isolating the pseudoaneurysm from the circulation.

Mortality was defined as death occurring during hospitalization or within 30 days. Data on recurrence of bleeding were obtained from direct contact with the patients, clinic follow-up records, and information from referring physicians, with a median follow-up time of 4.1 years.

The data examined in those patients with bleeding complications included prevalence, clinical presentation, location of bleeding lesions, diagnostic studies, operative findings, amount of blood transfused, and outcome with respect to mortality and recurrence of bleeding. Statistical analysis was carried out with Student’s t-test and the Wilcoxon rank sum test. Statistical significance was defined as P<0.05 (two-sided). Calculations were performed by SAS statistical software, version 6.12 (SAS Institute, Cary, N.C., USA).

Results

From a total of 36 patients with bleeding complications the most common causes of severe haemorrhage were bleeding pseudoaneurysms in 25 patients (69.4%). Haemorrhage was indirectly related to chronic pancreatitis in eight patients (22.2%) with gastric or duodenal ulcers or variceal bleeding. Three patients (8.4%) with chronic pancreatitis presented with complications involving the spleen (Table 1).

Table 1 Cause of bleeding in 36 patients with chronic pancreatitis
Full size table

Haemorrhage indirectly related to chronic pancreatitis

Seven patients developed bleeding ulcers (Forrest stages Ia to IIa). One of the patients with a Forrest stage Ia gastric ulcer underwent primary gastrotomy and ligation of the bleeding vessel. All other haemorrhages were controlled with endoscopically guided fibrin clot injections. One patient with recurrent variceal bleeding from the gastric fundus underwent ligation of the short gastric vessels with splenectomy. None of these patients died. In follow-up; two patients underwent elective partial duodenopancreatectomy 8 months and 13 months after the bleeding episodes.

Haemorrhage due to splenic complications

Two patients had spontaneous splenic rupture with haemorrhagic shock syndrome, and one patient had a complete organ infarction. Emergent splenectomy was performed in these cases. In one case, distal pancreatectomy was also performed because of an inflammatory mass on the pancreatic tail. There was no recurrence of bleeding or mortality observed in those patients.

Bleeding pseudoaneurysms

Clinical presentation

The clinical presentation on admission of these 25 patients varied but included one of the following clinical signs alone or in combination: acute onset of haemorrhagic shock with rapidly worsening anaemia (ten patients), gastrointestinal bleeding (12 patients), and sudden onset of severe abdominal pain (12 patients). Ten patients had an antecedent history of chronic upper abdominal and back pain within 4 weeks of admission and were requiring increasing amounts of analgesics. Additionally, five of those patients had acute exacerbation of chronic pancreatitis, diagnosed by clinical and biochemical criteria (acute abdominal pain requiring hospitalization, associated with elevated pancreatic enzyme levels three times above the upper limit of normal values). Table 2 summarizes the most important clinical symptoms at the time of admission. In one-third of all pseudoaneurysms the splenic artery was involved. The arteries and sites involved in the haemorrhages are shown in Table 3.

Table 2 Presentation of 25 patients at onset of bleeding from pseudoaneurysms
Full size table
Table 3 Intraoperative and angiographic findings according to location of pseudoaneurysms
Full size table

Radiological findings and interventional treatment of pseudoaneurysms

In nine of ten patients who underwent contrast-enhanced dynamic CT scan a pseudoaneurysm was suspected because of the presence of an enhancing mass. In one case of a splenic vein aneurysm, clear differentiation between a cystic pancreatic tumour and a thrombosed pseudoaneurysm was not possible by CT scan or angiography. This patient then underwent pylorus-preserving pancreatectomy. Coeliac axis and superior mesenteric angiography was performed in 19 patients, and a pseudoaneurysm was identified in 18 of them (Fig. 1).

Fig. 1
figure 1

Angiography and CT scan showing extravasation (arrows) from the superior mesenteric artery into a pseudocyst (delineated by arrowheads)

Full size image

Embolization as primary treatment was performed in nine patients (47% embolization rate). Of these, eight were successful (Table 4). There was one procedure-related complication with a recurrence of bleeding within 24 h of embolization of the right colic artery due to haemorrhage in the peritoneal cavity and partial necrosis of the right hemicolon. This patient, with multiple co-morbidities, underwent an urgent right hemicolectomy and died in the ICU from a postoperative acute myocardial infarction. Two patients underwent elective surgery after successful primary embolization (Table 5). During follow-up, one patient presented with a secondary pseudoaneurysm involving the left hepatic artery 18 months after successful embolization of the gastroduodenal artery. This new pseudoaneurysm was also embolized, with occlusion of the left hepatic artery and no complications.

Table 4 Treatment-related mortality and rate of recurrence of bleeding in 25 patients with pseudoaneurysm due to chronic pancreatitis
Full size table
Table 5 Treatment-related morbidity and mortality in pseudoaneurysm patients (RCA right colic artery, SA splenic artery, SMA superior mesenteric artery, JA jejunal branch, GDA gastroduodenal artery, AMI acute myocardial infarction, DIC disseminated intravascular coagulation, ARDS adult respiratory distress syndrome, PPPD pylorus-preserving pancreatectomy)
Full size table

Surgical intervention for pseudoaneurysms

From a total of 25 patients with arterial pseudoaneurysms, 16 (64%) underwent surgery. Six of these patients were taken directly to the operating room for urgent surgery, without angiography or CT scan, because of massive haemorrhage and haemodynamic instability (Table 4). Haemostasis was achieved by suture ligation of the culprit vessel in five patients (two in combination with external drainage of a pseudocyst, and two in combination with splenectomy). One patient died postoperatively, having had a massive haemorrhage with erosion of a splenic artery pseudoaneurysm into the stomach and peritoneal cavity (Table 5). Distal pancreatectomy with splenectomy was carried out in a sixth patient with a ruptured splenic artery pseudoaneurysm. During follow-up, one of the patients in this group suffered an upper gastrointestinal haemorrhage from oesophageal varices 9 years after surgery.

Ten patients underwent semi-urgent surgery after angiography and CT scan (Table 4), with six demonstrating intermittent bleeding. Four other patients presenting with haemorrhagic shock were resuscitated in the ICU and then underwent radiological examination. Pylorus-preserving pancreatectomy was performed in two patients, and distal pancreatectomy with splenectomy in two further patients. A transcystic arterial ligation of a jejunal artery pseudoaneurysm was carried out on two patients. Other surgical procedures included cystogastrostomy with ligation of the gastroduodenal artery, transgastric and proximal ligation of the splenic artery, and repair of a ruptured superior mesenteric artery in combination with cystojejunostomy. In this semi-urgent surgery group we observed four recurrences of bleeding (Table 4). One of these patients died after primary cystogastrostomy with ligation of the gastroduodenal artery, due to multiple episodes of re-bleeding (Table 5).

During follow-up after primary surgery, one patient developed a liver abscess after 1 year and a retroperitoneal abscess 2 years after primary surgery, both of which were drained percutaneously.

Survival and blood transfusions

Three of 36 patients met bleeding-associated deaths, representing a hospital mortality rate of 8.3%. The bleeding-related overall mortality rate in our cohort of 541 patients with chronic pancreatitis was 0.6%. All had suffered major haemorrhagic complications due to pseudoaneurysms, with shock as the primary indicator of haemorrhage. Two underwent primary surgical treatment [both with disseminated intravascular coagulation (DIC) after massive blood loss], and one was treated with primary embolization of the right colic artery and died, after secondary surgery, from an acute myocardial infarction.

The average overall blood replacement was 9.3 units of packed red blood cells (range 0–44 units; each containing 300 ml of blood). For patients with pseudoaneurysms, the average blood requirement was lower when the patient was treated angiographically (6.5 vs 17.3 units of blood, P<0.05). Patients with severe blood loss (defined as a transfusion of more than 10 units) had a significantly higher hospital mortality (P<0.05). Patient age, pseudoaneurysm location, and treatment modality had no significant influence on mortality.

Discussion

Bleeding complications in chronic pancreatitis are considered to be very rare. In our cohort of 541 patients we observed an overall prevalence of 6.7%. Haemorrhages indirectly related to pancreatitis from peptic ulcers or varices were found in 22% of our patients with bleeding complications. The higher prevalence of peptic ulcer bleeding in chronic pancreatitis than in the general population is well-established [2]. In these cases, surgery is undertaken only if endoscopic treatment fails or if the exact location of the bleeding site cannot be visualized by endoscopy.

Splenic infarction or spontaneous rupture in chronic pancreatitis occurs infrequently [7, 19]. Only three of 541 patients developed splenic bleeding complications, representing 8% of all patients with bleeding complications. A recently study published by Malka et al. [3] found six patients with intrasplenic subcapsular haematoma or splenic rupture among a cohort of 500 patients with chronic pancreatitis observed over 24 years. Although there are no prospective trials with patients with asymptomatic splenic vein thrombosis, a prophylactic splenectomy should be considered in patients with advanced gastroesophageal varices and splenomegaly, to prevent variceal bleeding and splenic rupture.

Haemorrhage from pseudoaneurysms occurred in 4.6% of patients with chronic pancreatitis and accounted for 69% of bleeding complications in such patients. Although 25 patients is a small cohort, this study represents one of the largest single-institute experiences. Most authors present data regarding pseudoaneurysms caused by both chronic and necrotizing pancreatitis [4, 11–13, 15–17, 20]. However, the different prognoses in those disease groups make it difficult for one to compare outcome and treatment modalities. Comparable studies, regarding bleeding complications caused by pseudoaneurysms in only chronic pancreatitis, reported a prevalence of from 7% to 10% for admitted patients [1, 10]. We consider that a haemorrhage frequency of the cited 10% for hospital-admitted patients should, in reality, be much more infrequent in tertiary centres. The small number of studies published indicates the relative rarity of this vascular complication. For that reason, and because of the urgent nature of the complication, it is unlikely that any randomized trials will be available in the near future for evaluating different treatment modalities. Evidence for optimal treatment will continue to be gathered from observational patient studies.

As observed in other series, the splenic, pancreaticoduodenal, and gastroduodenal artery are the most commonly affected vessels in pseudoaneurysms [21, 22]. None of our patients was in an asymptomatic state. Acute onset of pain was the most important symptom. Ten of our patients also had an antecedent history of chronic pain within 4 weeks of admission, as a typical warning sign prior to the bleeding attack, and five of these patients suffered additional acute exacerbation superimposed on chronic pancreatitis. Approximately half of our patients had gastrointestinal bleeding as the first manifestation of this complication, and 40% were haemodynamically unstable. In our patients, haemorrhage following rupture of a pseudoaneurysm directly into the gastrointestinal tract was much more severe than in those patients who bled into the pancreatic duct (four patients with haemosuccus pancreaticus) or pseudocysts where tamponade provided haemostasis.

Over the past 20 years, selective visceral angiography and contrast-enhanced CT scans have improved diagnostic accuracy for visceral pseudoaneurysms [2, 23]. Moreover, angiography has the advantage of providing simultaneous therapy via embolization. In our series, angiography was able to detect the pseudoaneurysm in 18 (95%) of 19 cases and CT scan in nine of ten cases, which reflects the high sensitivity of both methods. Nevertheless, a positive publication bias must be taken into consideration with regard to published series that study the sensitivity of angiography [12]. Despite the high diagnostic sensitivity of this method, the indication for angiography will always be dictated by a patient’s haemodynamic situation, as in our series, where we were unable to transport six patients for radiological examination because of their persistent haemodynamic instability.

In the past decade, many authors have reported the reliability of trans-catheter embolization in the treatment of arterial pseudoaneurysms for all kinds of pancreatitis, with embolization rates from 20% to 60%, as well as achieving haemostasis in 60%–80% of all cases (including both surgical treatment and intervention) [11–13, 17, 24, 25]. The number of patients who had undergone embolization, in these studies, ranged from 3 to 32, and the re-bleeding rate varied from 25% to 67%, with an average mortality rate of 20%. Our findings in nine embolized patients (embolization rate of 47%) show an 89% success rate (8/9), and an 11% re-bleeding and mortality rate, but only patients with chronic pancreatitis were included in our study. However, from the few data that are available for patients with chronic pancreatitis only, one can see that the optimal treatment remains an open question. There are recently published reports that have emphasized an aggressive operative approach [1, 2, 26]. Bresler et al. [1] recommended surgical treatment only, arguing from their remarkably low observed mortality rate of one in ten patients. They comment that embolization does not cure a chronically inflamed pancreas, although the results from embolization therapy are encouraging. In our opinion, an important strategy is to treat the bleeding without unnecessary major surgery. Nevertheless, the operating mortality rate in several series varies from 20% to 50%, depending upon the patient’s haemodynamic state, the bleeding location, and timing of the operation [2, 20, 23, 27]. With our multi-disciplinary approach we observed an overall mortality rate for pseudoaneurysms of 12% of 25 patients. Two patients died after surgery, and one after primary embolization. While there is no difference in the mortality rate between embolization and operative treatment, the haemodynamic state and the amount of blood transfused both had a significant influence on survival. However, surgical treatment remains a challenge and is associated with a higher morbidity and re-bleeding rate than embolization, although this is probably partly due to selection bias. Our data support the notion that ligation or repair of the bleeding vessel is complicated by higher re-bleeding and re-intervention rates than partial pancreatectomy. All patients who suffered recurrence of bleeding after primary surgery underwent ligation or repair of the bleeding vessel, without partial pancreatectomy. However, pancreatic resection was limited, in our study, by the patients’ clinical state and the presence of peripancreatic inflammation, which precluded a safe anatomical resection.

Conclusion and treatment recommendations for pseudoaneurysms

Patients with bleeding complications due to chronic pancreatitis seem to have a more favourable prognosis, with lower mortality, than those who suffer bleeding due to necrosis or abscess [28]. A necessary condition for improved outcome is an adequately differentiated approach, especially in arterial pseudoaneurysms. Any haemodynamically stable patient should undergo emergency angiography with embolization when technically possible. If there are no other pancreas-related indications for surgery (suspicion of malignancy or chronic pain), embolization might be the definitive treatment, which is what we observed in six of nine embolized patients. If there are indications for surgery, embolization can serve as a preliminary measure until surgery can be performed under more favourable conditions. If embolization is not available or has failed, urgent or semi-urgent surgery is indicated, although perioperative morbidity might be higher. In this case, depending upon the patient’s general condition and anatomic considerations, a partial pancreatic resection will prevent the recurrence of haemorrhage more safely than surgical control of the bleeding vessel alone.