Delayed Post Pancreatectomy Hemorrhage: Incidence, Clinical Characteristics, Risk Factors and Management

Pancreatic resection and especially pancreaticoduodenectomy (PD) is considered a complex operation, previously notorious for high perioperative morbidity and mortality rates. However, while the perioperative mortality of PD was over 15% in the 70’s, it is now days less than 4% [1-7]. One of the most important reasons for this improvement is the greater experience of a limited number of surgeons who perform the procedure on a regular basis in high-volume referral centers [8-11]. Combined pancreatic resection and vascular reconstruction prevalence is rising. As expected, these patients have a higher morbidity and mortality rate in comparison to conventional PD (39.9% vs.33.3% and 5.7% vs.2.9%, respectively) [12].


Background
Pancreatic resection and especially pancreaticoduodenectomy (PD) is considered a complex operation, previously notorious for high perioperative morbidity and mortality rates. However, while the perioperative mortality of PD was over 15% in the 70's, it is now days less than 4% [1][2][3][4][5][6][7]. One of the most important reasons for this improvement is the greater experience of a limited number of surgeons who perform the procedure on a regular basis in high-volume referral centers [8][9][10][11]. Combined pancreatic resection and vascular reconstruction prevalence is rising. As expected, these patients have a higher morbidity and mortality rate in comparison to conventional PD (39.9% vs.33.3% and 5.7% vs.2.9%, respectively) [12].
In spite of the impressive progress made in recent years in perioperative morbidity and mortality, a pancreatic fistula as a postoperative complication occurs in as high as 22% of cases [13]. Pancreatic fistulas can lead to sepsis and hemorrhage if not adequately drained. These complications are associated with mortality of 20 to 40%, prolonged hospitalization, and increased hospital expenses [14].
Post pancreatectomy hemorrhage (PPH) is a term used for all bleeding episodes post pancreatic surgery. It is further divided according to the time of onset post operatively (delayed: >24 hrs postoperatively), location (intra/extraluminal, anastomotic origin) and severity of bleeding (severe PPH defined as hemoglobin drop>3 gr/ dl, hemodynamic instability or need for intervention) [15]. Prevalence of PPH is 5.5%, equally distributed among different indications for surgery (malignancies, borderline tumors, and focal pancreatitis). In an era of increased number of PD with vascular reconstruction, it is expected that the incidence of PPH will rise. In recent years an objective, universally accepted definition and clinical grading of PPH is adopted according to the International Study Group on Pancreatic surgery (ISGPS) [15].
PPH-related overall mortality is 16-20% and raises to about 50% among the subgroup of DPPH secondary to pseudoaneurysm (PSA) or vascular erosion [16]. PSA exists when a normal arterial wall is replaced by fibrous tissue due to mechanical (iatrogenic) or chemical (by pancreatic juice) injury. This PSA may further cause a life-threatening bleeding.
Clearly, some of the improvement in overall mortality after PD may be attributed to the post-operative management of complications. In this aspect, the considerable advances in invasive radiological techniques enable control of DPPH and reduce the need for re-laparotomy with its significant sequela [15].
This study describes our experience in the management of DPPH with focus on risk factors and management.

Methods
We retrospectively analyzed our prospectively collected data on all pancreatic resections preformed in our center during the time period of January 2008 to December 2013. This study was approved by the Institutional Ethic Board. All pancreatic resections files (403) were reviewed for preoperative demographics and comorbidities, patient operative course and post-operative complications including postoperative imaging studies and final pathology report.

DPPH
PPH was defined according to the ISGPS commonly used criteria. Briefly DPPH is defined as either intra or extra-luminal bleeding more than 24 hrs post-surgery. When it is accompanied by hemoglobin concentration drop of >3 g/dl and a circulatory impairment requiring medical intervention it is defined a grade C hemorrhage. In the cohort 21 bleeding events were recorded: 7 early bleeding events and 14 late. 3/7 with early PPH had grade C bleeding event and were re-operated in POD 0. The rest of the early PPH had grade B hemorrhage (in nasogastric tube aspirate and external drains). 4 of DPPH were grade B hemorrhages that were proved to bleed from gastro-jejunal anastomosis by endoscopy. Our study which focuses on PPH due to PSA or vascular erosion deals with grade C DPPH according to the ISGPS definitions. We also included cases in which active bleeding were demonstrated by angiography.

Pancreatic fistula
Pancreatic leak was defined by 3 times or more increased amylase content in secretions of external drains (in comparison to plasma) after post-operative day (POD) 3 [17].

Sentinel bleeding
Sentinel bleeding is defined as a preliminary warning bleeding that precedes a major hemorrhage [18,19]. This sentinel bleeding probably indicates local infection and an anastomotic leak [18].

Diagnosis and treatment of choice
The decision upon management of post pancreatectomy bleeding in this study is based on a clinical evaluation and judgment of 4 senior experienced pancreatic surgeons.

Visceral angiography
Transfemoral visceral angiography and embolization were performed by experienced IRs. The preferred technique of super selective embolization was employed with stainless steel coils placed proximal to the arterial wall defect or within the aneurysm itself. Coils were placed using a coil pusher or via a flushing method. Alternatively, bridging stent grafts were used.

Incidence and clinical characteristics of DPPH
We identified 10 patients with DPPH, which yields a 2.5% incidence of this dreadful complication. Patient's characteristics are described in table 1. Interestingly, all cases were male, in comparison to equal gender distribution (49% males and 51% females) in the whole cohort of pancreatic resections. In all DPPH cases, pancreatic leak was diagnosed before hemorrhage. The leak was classified as grade C in 80% of the patients (due to septic course) and grade B in 20% (due to peri-pancreatic collection) [17]. The bleeding event reclassified all our DPPH patients to grade C. In the cohort an overall 25% of pancreatic leak was diagnosed; 10% were grade A leak and the rest were B and C. Drain culture was positive in all cases classified as grade C cases prior to bleeding.

Diagnosis
Principally, the diagnosis was pure clinical. Of note are 3 cases in which CTA following sentinel bleeding failed to demonstrate active bleeding or PSA, later found during angiography.

Management
In 7 patients, interventional radiology (IR) was the primary initial treatment, 2 were rushed to the operating room and one died in the ICU before any definitive therapeutic effort could be done. The decision to operate was influenced by a combination of bleeding and enteric leak in one case, and vascular reconstruction and presumed venous bleeding in the other. Four out of 7 patients were transferred to the IR suite while hemodynamically unstable and on vasopressor support. A decision which may contradict basic surgical principles, but was found to be effective (3/4 survived vs. 0/2 survivors when surgery was applied). Figure 1 shows the different invasive radiology techniques employed for bleeding control. These included coiling of the bleeding branch or of the feeding source and stenting (usually with bridging stents). Angiographic findings, i.e. bleeding source, and procedures performed are detailed in table 2. 2/8 patients needed re-angiography due to recurrent bleeding after the first procedure.

Outcomes
The outcomes of delayed PPH are quite gloom as shown in table 2. Overall, this complication harbor 50% mortality rate (n=5) with prolongation of hospital stay among the entire group (average of 38 days, in comparison to other pancreatic resections, 9.5 days). Among the survivors of DPPH: 2 year survival of 80% (comparable to the cohort), readmission rate of 40% in the first year after operation (cholangitis, small bowel obstruction due to post-operative ventral hernia).   All cases in our series demonstrated an event of sentinel bleeding, which actually pose a diagnostic opportunity. When it was promptly investigated by formal angiography or computed tomography angiography (CTA) an improvement in outcome is noticed (2/2 cases survived, with only minor morbidity of delayed gastric emptying in one of the two). However, all bleeding events in our patients rose from the celiac trunk which makes CTA useless for bleeding localization. Moreover, in 3 cases CTA failed to demonstrate active bleeding or a PSA. Altogether, these data raise the possibility to directly evaluate sentinel bleeding by angiography, omitting CTA, especially in high risk setting (like in pancreatic fistula). It may also be prudent to evaluate for PSA in high risk cases (e.g. pancreatic fistula and vascular reconstruction) in the appropriate time course even in the absence of sentinel bleeding. In summary, due to the high mortality rate of DPPH, we suggest a liberal use of angiography or imaging in high risk settings. Clearly, further research should be carried before a firm conclusion is reached.
As mentioned above, all cases of DPPH had a pancreatic fistula. At least 4 of the patients had low output fistula so that a link between volume of drainage and DPPH could not be established. Organisms cultured in our patients represent common bowel flora. This stands in contrast to a previous report which stressed the role of S. Aureus on delayed PPH [21].
Interestingly, in this series all cases are males while in the cohort of pancreatic surgery they consist only 49% of all patients. This male tendency had not been previously reported though higher male ratio is found in previous series [4,19,22]. However, in light of the small sample size, this finding may be a sampling error.
The significant role of the IR in the management of DPPH is demonstrated once again in this work. None of the operated patients survived. Only patients who are bleeding could be controlled angiographically survived. Even hemodynamically unstable patients treated by IR approach had 75% survival. This was also reported by others [22,23]. Culminating evidence suggests that angiography is the preferred management option for DPPH. Under the limitations of small patient number and possible selection bias, the surgical dogma that hemodynamically instability mandates operation is once again confronted with the better outcome presented here by IR.
Our and others experience highlight again the key role of multidisciplinary expertise, experience, availability and resources needed to manage DPPH in the effort to minimize mortality following pancreatic surgery.

Conclusion
DPPH is a major source of post pancreatectomy mortality in recent years. This complication has a typical clinical scenario: a pancreatic leak and a sentinel bleeding event. This scenario affords an opportunity for early intervention by angiographic techniques. It seems that hemodynamic instability does not preclude effective angiographic intervention and according to published studies, is the only way to save patient life.