Systematic review of the incidence, presentation and management of gastroduodenal artery pseudoaneurysm after pancreatic resection

Background Gastroduodenal artery (GDA) pseudoaneurysm is a serious complication following pancreatic resection, associated with high morbidity and mortality rates. This review aimed to report the incidence of GDA pseudoaneurysm after pancreatic surgery, and describe clinical presentation and management. Methods MEDLINE and Embase were searched systematically for clinical studies evaluating postoperative GDA pseudoaneurysm. Incidence was calculated by dividing total number of GDA pseudoaneurysms by the total number of pancreatic operations. Additional qualitative data related to GDA pseudoaneurysm presentation and management following pancreatic resection were extracted and reviewed from individual reports. Results Nine studies were selected for systematic review involving 4227 pancreatic operations with 55 GDA pseudoaneurysms, with a reported incidence of 1·3 (range 0·2–8·3) per cent. Additional data were extracted from 39 individual examples of GDA pseudoaneurysm from 14 studies. The median time for haemorrhage after surgery was at 15 (range 4–210) days. A preceding complication in the postoperative period was documented in four of 21 patients (67 per cent), and sentinel bleeding was observed in 14 of 20 patients (70 per cent). Postoperative complications after pseudoaneurysm management occurred in two‐thirds of the patients (14 of 21). The overall survival rate was 85 per cent (33 of 39). Conclusion GDA pseudoaneurysm is a rare yet serious cause of haemorrhage after pancreatic surgery, with high mortality. The majority of the patients had a preceding complication. Sentinel bleeding was an important clinical indicator.


Introduction
Mortality from pancreatic resection has fallen significantly over the past few decades, especially in experienced centres 1 -3 . Morbidity, including delayed gastric emptying, anastomotic leak and pancreatic fistula, remains high, affecting around 20-40 per cent of patients 4 -6 . Postoperative haemorrhage is less common, but is a life-threatening event with an estimated mortality rate of 20-50 per cent 7,8 . Early-onset haemorrhage is rare, and generally occurs within 24 h, usually due to technical failures. Delayed haemorrhage occurring days or weeks after surgery occurs for a variety of reasons, but one cause of massive haemorrhage is from the formation of visceral arterial pseudoaneurysms. Although several arteries have been shown to be vulnerable to pseudoaneurysm formation, observational studies indicate that pseudoaneurysms of the gastroduodenal artery (GDA) are the most common 9 -11 . The aim of this systematic review was to synthesize existing evidence regarding the incidence, clinical presentation and management of GDA pseudoaneurysms after pancreatic surgery.

Methods
Studies evaluating GDA pseudoaneurysm formation after pancreatic surgery were identified by means of database searches of MEDLINE and Embase. In Embase the terms used were: 'false-aneurysm' AND 'gastroduodenal artery' AND 'pancreaticoduodenectomy' OR 'pancreatectomy' OR 'distal pancreatectomy' OR 'pylorus preserving pancreaticoduodenectomy' OR 'pancreas surgery'. In MEDLINE the terms used were: 'false aneurysm' OR 'pseudoaneurysm' AND 'gastroduodenal artery' AND 'pancreaticoduodenectomy' OR 'pancreas surgery' OR 'pancreatectomy' OR 'distal pancreatectomy' OR 'pylorus preserving pancreaticoduodenectomy'.
A manual reference search was also performed to identify additional observational studies. No language restrictions were applied. Inclusion criteria were: manuscript published in a peer-reviewed journal until 2017, investigating adult patients aged over 18 years, undergoing pancreatic surgery for any indication, developing GDA pseudoaneurysm, and reporting clinical outcomes of interest. The authors independently reviewed all relevant titles and abstracts, and all disagreements were resolved by consensus. Observational studies that reported both the number of GDA pseudoaneurysms and the total number of pancreatic operations performed were used for quantitative analysis of incidence. Qualitative information also relevant to the clinical presentation and management of GDA pseudoaneurysms was extracted from individual cases and collated. Data extracted included index surgery, sentinel bleeding defined as haemorrhage that occurred in the gastrointestinal tract (intraluminal) or intra-abdominally (through a surgical drain) between 6 h and 10 days before a massive haemorrhage in the postoperative setting, day of postoperative bleeding, diagnostic method, management,  GDA, gastroduodenal artery; PD, pancreatoduodenectomy; PPPD, pylorus-preserving pancreatoduodenectomy; DP, distal pancreatectomy; SR, segmental resection; HPD, pancreatoduodenectomy plus hepatic resection; TP, total pancreatectomy. Hur et al. 36 1  other postoperative complications and mortality. GDA pseudoaneurysms were confirmed either radiologically or during surgery in all studies. Fig. 1. Some 88 studies were initially identified, 80 were screened, and 29 fulfilled the inclusion criteria. Of these, 13 studies 12 -24 were not included in the quantification as they were case series dealing exclusively with GDA pseudoaneurysms.

A PRISMA flow diagram is shown in
Five further studies were excluded as they did not provide the rate of GDA pseudoaneurysm in the postoperative period 25 -27 , or the total number of pancreatic operations performed 28,29 . One study 30 was excluded because it also included procedures not involving the pancreas (hepatic resection and gastrojejunostomy). Finally, one study 9 was removed as it recruited patients from the same institution as another report 31 , but over a shorter period.
All of the nine manuscripts 10,11,31 -37 selected for quantitative analysis were single-centre observational studies ( Table 1), mostly reporting on GDA pseudoaneurysms following pancreatoduodenectomy. One study 33 included both pancreatoduodenectomy and modified pylorus-preserving pancreatoduodenectomy, while another 10 included several pancreatic procedures. Data from 39 patients with GDA pseudoaneurysms were extracted from 14 studies for systematic review of clinical presentation and management ( Table 2).
A total of 55 GDA pseudoaneurysms were identified in the postoperative period following 4227 pancreatic procedures, with a reported incidence of 1⋅3 (range 0⋅2-8⋅3) per cent ( Table 1) Diagnostic procedures were reported for 21 patients; 18 GDA pseudoaneurysms were detected by angiography.
Thirty-five of the 39 patients (90 per cent) were treated using an endovascular approach. Nineteen (49 per cent) were managed using transarterial embolization (TAE) of the GDA via the common hepatic artery (CHA), and 14 (36 per cent) by selective embolization of the pseudoaneurysm. Stenting was employed in two patients (5 per cent), and only four (10 per cent) were treated by emergency laparotomy. The overall survival rate was 85 per cent (33 of 39).

Discussion
The GDA is the most common site for pseudoaneurysm formation after pancreatic surgery 9 -11 , and its rupture in the postoperative period has long been recognized as a cause of substantial morbidity and mortality 7,8 . GDA pseudoaneurysms are rare. The present analysis suggests that they occur in 0⋅2-8⋅3 per cent of pancreatic resections. It should be noted, however, that studies included in this review were all high-volume resectional centres.
In this series, two-thirds of the patients (4 of 21) had a preceding complication following pancreatic resection. Most authors favoured the hypothesis that lytic, enzyme-rich, pancreatic fluid from a pancreatic anastomotic leak could result in autodigestion of GDA vessel wall owing to its proximity to the pancreatic anastomosis.
Interestingly, a few studies 10,22,38 reported the formation of pseudoaneurysms at distance from the pancreatic anastomosis and in the absence of an overt pancreatic fistula, suggesting that minor iatrogenic injury, such as skeletonization of the vessel wall during extensive lymphadenectomy, may lead to vessel weakening and subsequent pseudoaneurysm formation.
Various techniques have been suggested to reduce the chance of pseudoaneurysm formation, including the 'wrapping' technique 25,39 -45 , in which the exposed retroperitoneal vessels are covered with omentum or the falciform ligament. Others 46 have suggested leaving 1 cm at the origin of the GDA stump to minimize the likelihood of lytic pancreatic juices coming into contact with the vessel.
Recognition of a sentinel bleed may help in early management, this being a feature in most patients 9,37,47 . Although sentinel bleeding was associated with poor outcome in some series 6,35 , few authors discussed the importance of immediate angiography after a sentinel bleed to look for the possibility of a ruptured pseudoaneurysm 6,37 . Although angiography also has the added benefit of allowing transition to endovascular treatment, a number of reports noted that negative findings cannot be used to exclude a bleeding pseudoaneurysm 26 , as bleeding can be intermittent or the rate of bleeding is below the detection limit of the equipment 7,9,37,48,49 .
Surgical intervention has been largely replaced by interventional radiology 50 -54 . Some older studies 26,55 advocated surgery in the context of additional intra-abdominal complications such as pancreatic fistula, but more recent series 11,36 have documented the superiority of endovascular management. A recent meta-analysis 56 of non-randomized studies comparing endovascular management and laparotomy for delayed massive haemorrhage suggested lower complication and mortality rates in the endovascular group.
The endovascular management of pseudoaneurysms varied, reflecting the location and size of the pseudoaneurysm and probably institutional preferences for approach and embolization technique, and materials. TAE of the GDA was conducted via either the CHA or the superior mesenteric artery to achieve both proximal and distal occlusion, to exclude the pseudoaneurysm and prevent backflow from collateral circulation 57 . Such an approach should consider patency of the portal venous system 41 as TAE distal and proximal to the GDA pseudoaneurysm can cause complete occlusion of the CHA, leading to liver infarction (reported range 30-66 per cent) 29,37,58 , as well as hepatic failure and abscess formation 58 . Covered stents represented the alternative to TAE. The key advantage over TAE would be in maintaining patency of the CHA and reducing the risk of hepatic infarction, although accurate stent deployment might be technically more challenging and time-consuming than TAE 36,57 . Despite these issues, stenting seems to be preferred in more recent series, on the grounds that selective embolization of the GDA stump or pseudoaneurysm seems to be associated with high rates of recurrent bleeding 36,42 .
Limitations of this study include heterogeneity of the included studies, the descriptors used and study sizes. The absence of any prospective registers or clinical trials on this topic needs to be addressed.

Disclosure
The authors declare no conflict of interest.